Most listing guides copy from each other, and half the numbers in them are years old. This page works the other way round: every claim carries its source and an evidence grade. Whatever failed our cross-check sits in its own list at the end.
Eleven chapters to read, three reference sections to come back to. You do not have to start at the top: jump to where your problem is.
Amazon publishes no ranking documentation. Anyone who still quotes percentages for individual factors has estimated them. That is why every claim on this page states where it comes from.
Stated in Amazon's own documentation: Seller Central, style guides, official posts. This is what you can cite in a dispute.
Follows from German or European law, not from Amazon's rules. Amazon does not check this — your competitors do. Relevant if you sell on the European marketplaces.
From a named study, usually by a tool vendor. Usable, but never neutral: whoever measures usually sells something alongside.
Broad practitioner consensus without a hard single source. Mostly right, but nothing you should base an appeal on.
And a fifth stamp, disputed: the sources contradict each other there. We write down both sides instead of picking one.
Since July 27, 2026, product titles are limited to 75 characters. If you don't shorten yours, Amazon's AI may write one for you — and you can lose the exact keyword that carried your ranking.
The title is the field Amazon reuses everywhere: in search results, in ads and in order confirmations. Amazon cut its space from 200 to 75 characters and introduced a new 125-character field at the same time. Title work is now a decision about priorities: which two or three search terms absolutely have to be in there. If you do nothing, your title can gradually be replaced by an AI suggestion. Brand owners see the suggestion in advance. Everyone else may only notice when the ranking collapses.
Together that is 200 characters again. They are just split differently: up front the sentence the customer reads, behind it the attribute list they skim.
Amazon writes, verbatim: from July 27, 2026, titles in all categories except media may contain at most 75 characters including spaces. The announcement sits in the Seller Central forum and is readable without a login, so you can verify it yourself. Onlinehändler News by Händlerbund reports the same for the European marketplaces, including amazon.de.
Premium Cotton Towel Set 4-Piece Towels Bath Towels 100% Cotton 500 gsm Fluffy Absorbent Quick-Drying for Bathroom Sauna Gym Hotel Quality Oeko-Tex Certified White Grey Anthracite (200 characters)KURT HOME Towel Set 4-Piece, Cotton 500 gsm, White (50 characters)In old titles, the distinctive part usually sits at the back: the differentiating features and the niche keywords. The front holds the brand and the generic product type. If you cut mechanically at character 75, you keep exactly the interchangeable part and lose the buying argument.
Premium Cotton Towel Set 4-Piece Towels Bath Towels 100% CottonKURT HOME Towel Set 4-Piece, 500 gsm Cotton, Oeko-Tex, AnthraciteAmazon writes that item highlights offer an additional 125 characters to state materials or recommended uses. The text appears under the title, in search results and on the detail page.
The best towel for your home: grab it now and experience quality.100% cotton, 500 gsm, 20x40 in, 4-piece set, for bathroom sauna gym, washable at 60 °CAmazon does not disclose which of your keywords the AI preserves when rewriting. Suppose your title currently ranks for "tent stakes sandy soil", and the AI turns it into "stakes for camping": the ranking string is then physically no longer in the field. Trellis describes exactly this as the risk, because the rewrite draws on Amazon's own relevance model, not on your revenue view.
Amazon names the 14-day review period for AI title suggestions explicitly for brand owners: brand owners get 14 days before implementation. Whether sellers without Brand Registry get the same period, Amazon has not said. Trellis reads this silence as a no; other sources report the period without restriction. Interpreting silence is not evidence. Without Registry, do not rely on any review period.
Amazon phrases the limit as "75 characters including spaces". That is a character count, not a byte count. An accented letter counts as one character in the title, exactly like an unaccented one. Anyone writing ASCII workarounds into the title out of old backend habit (where bytes are counted) gives away space and gains nothing. On the German marketplace, that mistake looks like "Kuechenmaschine" instead of "Küchenmaschine".
KURT HOME Kuechenmaschine 1200 W, Ruehrschuessel 5 l, weissKURT HOME Küchenmaschine 1200 W, Rührschüssel 5 l, weißAmazon's title policy of January 2025 reads, verbatim: "Titles may not contain the same word more than twice. Prepositions, articles, and conjunctions are exceptions" and "The special characters !, $, ?, _, {, }, ^, ¬, and ¦ are not allowed, unless they are part of the brand name." Two of these characters are often confused: banned are ¬ and ¦. The vertical bar | and the hyphen are not on this list. At 75 characters, every needless repetition takes space you need for a keyword.
Two independent trade sources report that the 75-character limit applies to every parent ASIN and every child ASIN individually, and that bundles are not exempt. Many sellers only shorten the variant titles. The overlong parent title then gets rewritten by Amazon, and the damage hits the whole variant family.
Zentail quotes Amazon saying that sellers remain responsible for their offers, including AI-generated titles. The original quote sits in a forum post we could not fully verify. The consequence is clear regardless: if the automatically generated title contains a wrong measurement, it stands in your offer under your name.
Your copy decides whether your product enters the candidate set for a search at all. The order after that is shaped mainly by how customers react to your offer. How strongly, Amazon does not document.
Amazon publishes no documentation of its search algorithms. What can be substantiated are the fields and their rules, not their weighting. What can also be substantiated is that search does not run in one step: first a set of matching products is drawn from the catalog, then that set is sorted. Your copy acts on the first stage. The second is driven mainly by how customers react to your offer. How strongly, is written nowhere. If you don't know this split, you polish copy and wonder why the ranking never moves. Or you work on conversion for search terms you are not even indexed for. So keep this order: secure indexing first (a yes/no question largely within your control), then image, price and reviews.
Which leads to the uncomfortable truth of this page: copy work decides whether you get to play. Where you place is decided by your customers. If you only tweak copy and wonder about frozen positions, you are working on the wrong stage.
An Amazon community manager wrote in the official seller forum: 'Amazon does not disclose information or release official documentation around search/discovery algorithms.' That is a forum post, not a help page. It remains the clearest publicly readable statement on the matter. So whoever quotes a number like 30 percent for conversion rate did not get it from an Amazon publication.
Conversion rate makes up 30 percent of the ranking.The number 30 percent circulates for conversion rate. It does not come from Amazon, so treat it as an estimate.A9.com was an Amazon subsidiary that worked on search technology. For an algorithm called 'A10' there is no Amazon source; the name only appears in service-provider posts. On top of that, Amazon describes its own product search in research publications as an interplay of several methods for matching and ranking, not as a single versioned algorithm. Whoever writes 'A10 favors X' cites a name with no source behind it.
The A10 algorithm weights external traffic more than A9 did.There is no Amazon source for an 'A10'. If you have an observation, call it an observation — not an algorithm.Amazon's search glossary says Amazon search cannot perform partial matching. A term therefore has to appear as a whole word somewhere in your catalog data; part of a word is not enough. Two caveats belong here. First, Amazon recognizes word forms itself: by Amazon's own guidance, singular or plural is enough, you don't need both. Second, Amazon does not recognize synonyms automatically — Amazon explicitly tells you to enter them yourself. How far the search maps semantically beyond that is not publicly documented. The practical question is therefore not how often your main keyword appears, but which term is still missing anywhere.
Title, bullets and search terms only say 'towel XXL'. For the search 'towel xxl cotton' your odds are poor, because the word cotton appears nowhere.'Cotton' appears at least once in your catalog data, for example in the item highlights. Whether you write 'towel' or 'towels' does not matter — Amazon recognizes both forms.Amazon writes on its public seller page: once you have used a word combination, you generally do not need it a second time — work with variants instead. For the backend search terms Amazon recommends using them as an overflow for terms that had no room up front, and avoiding redundant entries. A density multiplier as in classic web SEO is documented nowhere. Repetition brings no extra coverage; it only spends space you then lack for a synonym.
Title: 'Towel XXL Cotton'. Bullet: 'XXL towel made of cotton'. Search terms: 'towel xxl cotton'. The same words three times.Title: 'Towel XXL Cotton'. The search terms then hold synonyms like 'bath sheet' and 'sauna towel', not 'towel' again.The ASIN is itself an indexed term and narrows the results to your one product. Add a keyword, and your product only appears if it is indexed for both terms. That way you measure pure indexing without competing against anyone. Caveat: the test does not tell you which field the word sits in, and it is unreliable for very generic single words and for long compound words. Amazon describes the test nowhere; it is practitioner knowledge.
Type 'bath sheet xxl' alone and hunt for your own product in the results. That only shows who ranks better.B012345678 bath sheet xxlThe report shows you, per search query, impressions, clicks, add-to-carts and purchases — each also as your share of the total. That narrows the diagnosis without proving it. No impressions can be an indexing problem, but can also mean your position is so poor nobody ever sees you. Impressions without clicks point to title, main image or price. Clicks without purchases point to the offer and the detail page. About your position within the search results, the report itself says nothing.
No impressions for 'bath sheet xxl', so I am not indexed. Period.No impressions for 'bath sheet xxl'. First check with the ASIN test whether it is indexing or position.The US competition authority FTC sued Amazon on September 26, 2023, together with 17 states. The suit accuses Amazon of putting ever more poorly matching ads into the search results to grow ad revenue. The details became public in November 2023 with a less redacted version of the complaint. Amazon disputes this. These are allegations, not judicially established facts, and they concern the US market. Independently of that: how much advertising sits on page 1 is a business-policy decision by Amazon, not an algorithmic constant. Measure your success by your position within the organic results, not by your place on the screen.
I am on page 1 at position 8, so I am the eighth-best organic result.Subtract the ad slots and measure your position within the organic results.How good are your title and copy, really? The free AI analysis rates your listing in about 40 seconds — with a score from 0–100 and your biggest problem areas.
An instant result. No sign-up.
A missing attribute can remove you entirely from filtered searches — not just push you a few places down. The matching word in your title won't help you then.
Some searches on Amazon are constrained. The classic example is "red dresses": the query returns results that match the browse node "dresses" and the refinement "red". If your color attribute is empty, or holds a fantasy value, the word "red" in your title helps little. You are filtered out, not poorly placed. Attribute care is therefore not mere form-filling. It is visibility work, and it only costs you time. One note up front so you read the evidence grades correctly: almost all Amazon references on this topic sit in the seller help behind the login. As a seller you can read them; a reader without a seller account cannot. We therefore grade such points PRACTICE rather than AMAZON, even where we consider the content correct.
In the seller help Amazon describes product attributes as the basis for indexing and for the refinement options — the filter checkboxes to the left of the search results. A field that feeds a filter in your category costs you visibility if it stays empty. Which fields those are depends on the category. The help page is only reachable after login, hence the PRACTICE grade.
The value from the list is what the filter latches onto. "Blue" is in the list; "midnight blue" usually is not. Whoever enters only the marketing term risks being missing from the color filter. What the field is called and which values apply depends on the category. This reference too sits behind the login.
Color: Midnight blue melangeColor: Blue (the free-text "midnight blue melange" belongs in the title or the item highlights)Amazon describes in the seller help that ASINs without an assigned browse node are not indexed by Amazon search. Without a category assignment, the best copy is useless. We paraphrase the content because we cannot link the page publicly. It is only reachable after login.
Amazon has publicly described a knowledge graph called COSMO. According to Amazon, COSMO derives relationships from customer behavior such as "used for" and "suitable for", and it is deployed in search applications. Important for you: COSMO draws those relationships from behavioral data, not from your listing copy. There is no evidence that an occasion in your copy files you there. Writing down occasion and use case is still worth it — because that is exactly how customers search. That is the reason for this point, not COSMO.
100% cotton, 500 gsm100% cotton, 500 gsm: for sauna, pool and gym, also a gift for athletesAmazon writes publicly that its AI shopping assistant draws answers from product information, customer reviews and community questions. In the seller help Amazon additionally names complete product attributes as a prerequisite for more accurate context. A clean attribute value is easier for a language model to use than an ad sentence. On the name: the assistant was called Rufus; since May 2026 Amazon bundles Rufus and Alexa+ under "Alexa for Shopping". The part of the reasoning that only sits behind the login is the reason for the PRACTICE grade.
Attribute field "Material" stays empty; the bullet says: "Made of 100% cotton"Attribute field "Material": cotton — plus the bullet: "Made of 100% cotton"Amazon's attribute table has two separate columns: "Indexed by search?" and "Used by browse?". Target audience (target_audience_keywords) and other attributes (thesaurus_attribute_keywords) are not listed there as indexed by search. But not indexed by search does not mean without effect — otherwise the second column would not exist. Your time is better spent on the required and filter fields from point 1. The table is only reachable after login.
Changes to the offer and the category assignment do not take effect immediately. Amazon points to a latency in the seller help. We cannot publicly document a concrete period for you, which is why we name no number of hours here. Do not take the numbers from service-provider blogs as fixed either. Whoever sees nothing after a short while and changes again ends up measuring a jumble of several versions.
Accented characters count double, punctuation is unnecessary, and whatever is already in the title does not belong here again.
The backend search terms are one of the few fields your customer never sees that Amazon still evaluates for search. This is where synonyms, spelling variants and terms that would clutter the visible copy belong. Two traps hide in it. First, Amazon measures the limit in bytes, not characters. Second, accented characters like ä, ö, ü and ß take two bytes each in UTF-8 instead of one — a trap for every non-English marketplace. If you plan with a character counter, you plan past reality.
Suppose the title already reads: KURT HOME Handtuch-Set 4-teilig Baumwolle weiß. Words from the title get struck through below, because they add nothing here.
Amazon counts bytes, not characters. For a–z and 0–9, one character equals one byte. Accented characters like ä, ö, ü and ß take two bytes each in UTF-8. Every accented character therefore costs you an extra byte without any character counter showing it. You can verify it with this example: "handtücher badetücher saunatücher frottee" is 41 characters but 44 bytes.
Amazon publicly names three different numbers. In the US seller forum an Amazon employee writes 250 bytes. In the European seller forum an Amazon employee writes 249 bytes. In an answer on flat-file error 97779 the same employee writes 200 bytes. Amazon contradicts itself here. What exactly happens when you exceed the limit, none of these sources say. If you stay clearly below, you never have to settle the question.
Amazon's own seller page says: "Use all lowercase letters and separate words with spaces only", and further that you can leave out articles and prepositions. The German wording sellers quote from the help in the Amazon forum says the same: separate words with spaces, you can write everything in lowercase, and you don't have to enter punctuation like ';', ':', '-'. Every punctuation mark you leave out saves a byte.
Towel, Bath sheet; Sauna towel, for the saunatowel bath sheet sauna towel beach towel terry cottonAmazon describes this field on its own seller page as an overflow for terms that fit nowhere else on the product page, and advises against superfluous terms. In the seller forum Amazon puts it more directly: do not repeat information already captured in other fields such as title or brand. Every byte you spend on a repetition is missing for a synonym that appears nowhere yet.
Amazon's guidance for this field reads: use either singular or plural — there is no need to enter both forms. And: include spelling variants; spelling mistakes are not necessary. Entering both forms only costs you bytes.
towel towels towle bath sheet bath sheetstowels bath sheets sauna towels terryAmazon's guidance for this field explicitly names: do not enter your brand name or other brand names in your search terms; do not enter ASINs as search terms; do not use subjective claims such as 'best', 'cheapest', 'very nice'; and do not use time-limited claims such as 'new' or 'on sale now'.
best towels cheapest quality new on sale nowtowels bath sheets terry cottonAmazon merged what used to be several separate search term lines into a single field. Sellers report in the public Amazon forum that they had to shorten and re-cut their old keyword lines themselves afterwards. What exactly Amazon did with the contents of long-grown catalogs during the migration is not publicly documented. A look into your old listings costs little and settles it for your own catalog.
Are you missing search terms your niche actually types? The free check finds your keyword gaps — calculated on real Amazon search volume.
An instant result. No sign-up.
Each bullet answers exactly one customer question, starts with a short label, and stays understandable on its own.
Your bullet points are not prose. Amazon treats them as a catalog attribute with individual values, and each value has its own formal requirements. As a seller you get five fields for them. On the phone, the customer initially sees only the beginning of each bullet. The few words before the colon therefore co-decide whether they expand it at all. On top of that, Amazon's AI assistant uses catalog data as a source for answers. Both pull in the same direction: short, concrete, verifiable.
Our premium towel set impresses with its …
Washable at 60 °C: Colorfast even after …
Amazon's requirements for bullet points name at least three bullets. Whoever provides fewer risks warnings, automatic interventions in the listing, or removal of the bullets. Sellers get five fields; vendors have up to ten in some categories.
Amazon's updated bullet requirements have applied since August 15, 2024. The wording the sources relay: more than 10 and fewer than 255 characters per bullet. In some categories the limit is lower; 150 characters are reported. Which value applies to you is in the style guide of your category.
In its seller blog Amazon recommends starting with a one-to-two-word descriptor and then explaining in at most 100 characters. Amazon explicitly names sentence fragments instead of full sentences, and a capital letter at the start of each bullet. The page is Amazon's English-language seller blog; there is no separate version per marketplace.
Our premium towel set impresses with its exceptional quality and is guaranteed to delight you!Washable: Colorfast even after 50 washes at 60 °C; tumble-dryer safeOn mobile devices only a preview of each bullet is shown; the rest hides behind the expand. A single agency observation names about 80 characters for it. That number is not confirmed by Amazon and can change with any app update. Only the direction is solid: whatever sits at the back, the phone customer does not see at first.
So that you enjoy your towel set even after many washes, we made it washable at up to 60 °CWashable: Colorfast at 60 °C, even after 50 washes, and tumble-dryer safeAmazon writes that its AI assistant draws on the product catalog, customer reviews, community questions and information from the web. Which sentences from a listing it picks up, Amazon does not say. It stands to reason that a verifiable statement works better as an answer than an advertising adjective. This tip is therefore a conclusion from the Amazon statement, not an Amazon requirement.
Ideal on the go and perfect for every trip, so you are always well preparedCarry-on compliant: outer dimensions 21.7 x 15.7 x 7.9 in including wheelsAmazon's requirements list as not allowed in bullets, among other things: special characters such as trademark and copyright signs, emojis including check marks and crosses, placeholder text, ASINs, claims such as "eco-friendly" or "antibacterial", warranty statements, contact details and external links, as well as repetitive content. Formal rules come on top: no punctuation at the end, semicolon as the separator within a bullet, a space between digit and unit. Violations can lead to Amazon changing or removing the bullet.
Amazon requires avoiding overlaps with other product attributes such as title, product description or product overview. Each individual bullet should carry standalone product information. What already sits structured in the attribute table, you do not need to repeat in the bullet. The original page in the Seller Central help is only reachable after login, so you cannot read it without a seller account. That is why this point sits at PRACTICE and rests on the public retelling.
The main image holds only the product on a pure white background. All arguments, dimensions and infographics belong on the other images of the gallery.
Your main image does not only run on your product page. Amazon also shows it in search, in ads, in comparison modules and in the app. That is why tight requirements apply to this one image. They do not apply to the other gallery images — text there is normal and welcome. The main-image requirements are phrased concretely and are verifiable. You can tick them off yourself before uploading.
At full size, every detail is readable. That is exactly how you look at your images while uploading.
Amazon writes on its own seller page: 'In most cases, product shots should be taken against a white background (RGB color values: 255, 255, 255)'. The German forum post names the same as a pure white background. Soft, semi-transparent cutout edges and missing color profiles shift the white. A tolerance appears in neither source.
Towel with soft, semi-transparent edges on a light-gray background, exported without a color profile.Towel cut out with a hard edge, background as a solid RGB 255, 255, 255, exported as JPEG in sRGB.Amazon writes 'Have the product fill 85% or more of the frame'. The German forum post names the same number, at least 85% of the frame. In the search tile your image is scaled down to a few hundred pixels. The smaller the product sits in the frame, the less of it stays recognizable.
The towel sits small in the middle of the image, with lots of white space around it.The towel fills the frame almost entirely; only a narrow white margin remains.The Amazon seller page names text, logo and watermark as what has no place on the product image, and advises against dominating accessories, models and backgrounds. The German forum post phrases it more tightly: no text, no logos or watermarks — and show only the product being sold, no props or accessories. On the other gallery images, text is normal.
Main image with a brand logo top left and a red badge '100% cotton'.The main image shows only the towel; the note '100% cotton' sits on image 3.Amazon names 500 to 10,000 pixels on the longest side and, as formats, JPEG, TIFF, PNG and non-animated GIF. For the zoom function Amazon recommends on the same page images with more than 1,000 pixels per side. With 2,000 px you are on the safe side of every circulating figure without having to rely on any of them.
1,000 x 800 pixels, exported straight from the phone.2,000 x 2,000 pixels as JPEG.Amazon requires at least one image per product and recommends at least six. How many of those images the product page actually serves in the end, Amazon does not document publicly. Count it on your own live listing instead of relying on a circulating number.
Two images: main image and one detail photo.Main image, in use, infographic, dimensions, material and what's included.In the search results and in the mobile gallery your image is scaled down to a small tile. One vendor source names 160 to 200 pixels of width for it. That number is not confirmed by Amazon, but you can see the underlying problem yourself as soon as you view your image that small. An infographic with 14-point type that looks fine on a monitor is just a gray bar there. So: three to five words per image statement at most, high contrast, and no thin typefaces.
Four lines of body text in thin, light-gray type on the infographic.Two words in large, bold, high-contrast type: 'Quick drying'.For safety and compliance images there are dedicated image variants with the codes PS01 through PS06. A service-provider blog and several sellers in the German seller forum describe them consistently, named after the pattern ASIN.PS01.jpg. We found no publicly accessible Amazon page about it. This way you show legally required markings without touching the text rule on the main image.
The warning notice is montaged into the main image as a text box.Photo of the back of the packaging with warning notice and CE mark, uploaded as ASIN.PS01.jpg.Amazon announced this on July 23, 2026 — for images and videos in the listing as well as for A+ content. As the scope, Amazon names its own stores worldwide. The trigger is New York's law on synthetic performers, General Business Law 396-b, in force since June 9, 2026. You do not have to label real people (even after AI editing), characters from films, series, documentaries and games, non-photorealistic depictions, and images without people. The rule is only days old — check the current state before uploading.
A photorealistic AI model holds your towel; the file is uploaded without metadata.exiftool -XMP-dc:Subject+=contains-synthetic-performer file.jpgThe German forum post names as consequences of image violations the removal of the images, the hiding of the offers from search, and possible account restrictions. We found no publicly accessible Amazon help page with exactly this list, hence PRACTICE and not AMAZON. Without a main image, your offer is practically unreachable through search. The replacement should therefore be uploadable within hours.
The main image gets removed, and a compliant file still has to be built.A cut-out version on RGB 255, 255, 255 sits finished in the folder, ready to upload immediately.Bullets and images under scrutiny: get the honest score for your listing — seven categories, from title to Amazon compliance.
An instant result. No sign-up.
The best copy is useless when "Add to Cart" has turned into "See All Buying Options". Look for the cause in price and availability first, not in your copy.
The featured offer is the fastest way into the cart. If it drops away, the customer needs an extra click and has to pick an offer themselves. Fewer customers do that. Unlike a ranking problem, this can often be reversed quickly: fix the price, and in many cases the button comes back. What matters is the order of diagnosis. If your revenue drops without you having changed the listing, check buy button, price and availability first — keywords after.
For the featured offer, the total of item price and shipping counts — not the item price alone that the customer sees in the list. The source calculates it with an example in US dollars: 19.99 USD plus 3.99 USD shipping makes 23.98 USD and loses against a Prime offer at 22.99 USD with free shipping. If you push the item price down and recover the margin through shipping, you look cheaper in the list and get rated more expensive on the landed price.
You lower the item price of your towel set and put the difference on shipping. In the list that looks cheap.You add item price and shipping together and only then set that sum against the competitor's Prime offer.Amazon is removing the separate eligibility check for the featured offer. In the announcement's wording: "Starting in July 2026, we'll begin removing seller eligibility requirements for the Featured Offer. This change will roll out gradually across all Amazon stores globally, completing by the end of 2026." With that, the yes/no pre-filter that used to be a clear signal is gone. Whether your offers still sit up front, you now read mainly from the trend of your featured-offer percentage. A single value says little.
You only look at the current value and find it fine.You keep the value as a curve and see the slow decline before it reaches your revenue.Amazon writes in the same announcement that nothing changes about how the featured offer is selected. Selection continues "based on criteria most important to customers, such as competitive pricing, delivery speed, and performance". And: "Being considered for the Featured Offer doesn't guarantee your offer will be featured." Being considered does not guarantee you the selection.
You let delivery time and tracking slide because the eligibility check is going away anyway.You keep order defect rate and delivery time clean, because Amazon keeps selecting by them.The most common trigger of a suppression is considered to be price: against the historical purchase price of the ASIN, and against cheaper offers outside Amazon. Also named are low sales velocity, weak seller metrics, and problems with listing quality and item condition. This can also hit private labels where nobody else offers on the ASIN.
The buy button is gone, and the first thing you do is rewrite the bullet points.The buy button is gone, and the first thing you do is compare your price with the ASIN's past purchase price and with your offers outside Amazon.A sold-out offer cannot win the featured offer. At the same time, the sales history your organic ranking rests on breaks off. In the experience of many sellers, the recovery takes longer than the gap itself. A clean proof for this is missing, but practitioners largely agree on the direction.
You let the stock of your bedding set run to zero because the resupply is on its way anyway.You reorder so that a remainder always stays in the warehouse, even if that means a bit more stock on hand.An ad sale only helps your organic ranking if it is incremental — if it would not have happened without the ad. On your own brand name, many customers find you without an ad. If the ASIN's total revenue stays almost flat during the bid test, that is a hint of cannibalization. It is not proof, because season, competitor prices and stock change over the same period. The trade articles clearly contradict each other here: one side considers cannibalization the exception, the other sees a strong ranking effect from ads.
You keep bidding on your own brand name and book the sales as the ad's success.You lower the bid there for two to three weeks and note season, price and stock alongside. If total revenue stays flat, repeat the test in a second period before changing course.The circulating averages come from service providers. The often-quoted benchmark article names its ranges explicitly for the US market and does not disclose the methodology behind them. For your towel set on your marketplace, that says little. What is meaningful: the comparison with yourself over time, and the purchase share from Search Query Performance.
You read an average conversion rate for your category and measure your towel set against it.You set your towel set's conversion rate against its own previous period and add the purchase share from Search Query Performance.Your goal is not the 5.0. Your goal is to sit stably above 4.0 and to get there faster than organic reviews allow.
Reviews act on your conversion, and conversion is considered one of the factors behind your placement. The effect is real, but smaller than the numbers in circulation. And it has an unpleasant property: it arrives with a delay, and with few reviews it swings hard. Two one-star reviews among five push the average to 3.4 even if the other three give five stars each. At 200 reviews, the same two still leave 4.96. That is why you plan your reviews before launch, not after.
The dip on the right is not a measurement error. A spotless 5.0 with few reviews reads to the customer as "too good to be true". A 4.5 with three hundred votes looks more credible than a 5.0 with seven.
The Spiegel Research Center at Northwestern University measured in 2017, together with PowerReviews: "purchase likelihood typically peaks at ratings in the 4.0 - 4.7 range, and then begins to decrease as ratings approach 5.0." The authors attribute this to a flawless rating appearing implausible to many buyers. The 4.0 in the key sentence is the lower bound of that range. Important: the study was not conducted on Amazon data, so the direction is more reliable than the exact thresholds.
The service provider Pattern measured, in its own Amazon portfolio, roughly 4 to 5 percentage points more conversion per additional star — from 3.5 to 4.5 stars, 24 to 29 percent. In relative terms that is about 21 percent. Pattern itself writes "percent" here; by its own numbers, percentage points are meant. It is a vendor analysis without sample size, period or significance data, so only the direction is usable, not the amount. Whoever optimizes from 4.4 to 4.6 gains little. Whoever climbs from 3.7 to above 4.0 works their way into the range where the Spiegel study puts the highest purchase likelihood.
Your towel set sits at 4.4. You sink weeks into reaching 4.6.Your towel set sits at 3.7. You work on it until it sits above 4.0.On amazon.de, Vine costs in tiers: 0 euros (up to 2 units), 70 euros (up to 10) and 170 euros (up to 30) per parent ASIN. The often-quoted 200 is the US price in US dollars. You can enroll an ASIN only once, the chosen tier cannot be changed afterwards, and child ASINs cannot be added later. The fee is only charged if at least three units are enrolled and the first review arrives within 90 days of enrollment. The evidence is Amazon's own announcement in the seller forum of October 19, 2023, not a help page. The help page on it is only readable after login, hence PRACTICE and not AMAZON here. Check the conditions yourself in the Vine section before enrolling — prices differ per marketplace.
A bedding set from your core range, enrolled with 2 units for 0 euros. You cannot top up afterwards.A bedding set from your core range, enrolled right away with 30 units for 170 euros.Vine reviews count into your star average, and you cannot get them removed on request. With an unfinished product, Vine therefore causes exactly the damage a low star average causes — just faster.
The seam on the pillowcase does not hold reliably yet. You enroll it in Vine anyway.Only once the seam holds do you enroll the pillowcase in Vine.Amazon's review policy bans incentives for reviews; violations can go as far as your account being deactivated. On top of that, in Germany the law applies: annex no. 23c to section 3(3) of the German unfair competition act (UWG) bans submitting or commissioning fake consumer reviews. That is a ban without any balancing of interests and without a materiality threshold — only the act itself has to be proven. Bought reviews fall under it. Whether a genuine review you prompted with a coupon also falls under it, the wording does not say. For you it changes nothing: Amazon's rule applies worldwide either way.
A card in the package: a 5-euro coupon for your review.The "Request a Review" button in Seller Central, nothing else.Amazon has restricted the automatic sharing of reviews between variants. The rollout ran category by category from February 12 to May 31, 2026, and is complete; affected sellers got an email 30 days ahead. Reviews continue to be shared for minor differences without impact on function — color, pattern, same function in another size, or pack sizes. For variants that differ in function, no longer. Amazon writes this can affect the overall rating of your products and the number of their reviews. The announcement stood in the seller forum and thus applies across marketplaces. Because it is a forum post and not a help page, it sits at PRACTICE here.
You assume the 300 reviews of your towel family still hang on every child ASIN.You check which child ASIN has how many reviews today, and plan the weak children individually.Amazon states that customers who use the AI assistant Rufus are over 60 percent more likely to buy in that session, and that the assistant draws its answers from customer reviews, among other sources. The 60 percent is Amazon's own figure and a correlation without proof of causation — whoever asks an assistant is closer to the buying decision anyway. Amazon names as further sources the product catalog, questions and answers, and information from the web. A use case that appears neither in your reviews nor in your attributes has a harder time with this answering machine.
Several reviews praise that the towel works as a guest towel. The listing says nothing about it.You add the guest-towel use case to your listing.Where is your listing giving away visibility right now? The AI analysis calculates your potential in real search contacts per month — broken down by area.
An instant result. No sign-up.
A+ content costs no extra fee, but requires a professional account and, as a rule, a brand enrolled in Brand Registry. You still never leave the description field empty.
A+ content lays itself over the product description on the detail page. The field itself remains in Seller Central; an Amazon employee confirms this publicly in the seller forum. Whether Amazon keeps using the covered text for search is not publicly documented. Whether the A+ text itself counts for search, likewise. Both are open, and that is exactly why the math is simple: keeping the field filled costs you nothing. Emptying it can cost you rankings and takes away your fallback if A+ is rejected or removed again.
Product description
The 4-piece towel set in 100% cotton with a 500 gsm weight consists of two hand
towels (20x40 in) and two bath towels (28x55 in). Washable at 60 °C,
tumble-dryer safe, Oeko-Tex certified.
A single block of text many customers skip.
Image-and-text modules instead of a text block.
A+ is live, so I delete the description field.
A+ only covers the description in the display. In the catalog, the field remains and keeps being used.
A+ content lays itself over the product description on the detail page; the field remains in Seller Central. An Amazon employee writes in the publicly readable seller forum: 'A+ content will cover up parts of product detail page information, but you should still have the option to leave the existing product descriptions within Seller Central.' Whether Amazon keeps using the covered text for search, Amazon says nowhere publicly. Keeping it filled costs you nothing, and you keep a fallback in case A+ is rejected or removed again.
Description field empty, because the A+ module sits on top of it anyway.Description field filled and in sync with the A+ module: terry towel, 100 percent cotton, 20 x 40 in.Amazon announced through the official forum account News_Amazon that after June 8, 2021, HTML tags are no longer supported on detail pages. Lines with an HTML tag are removed or displayed without formatting. The announcement is readable without a login. That bold tags still work on a few old listings changes nothing about the announcement.
<b>100% cotton</b><br>Machine washable at 60 °C100% cotton. Machine washable at 60 °C.Amazon writes: 'A+ Content is free for sellers who meet eligibility requirements' and names as prerequisites a professional account plus the Brand Representative or Reseller role for a brand enrolled in Brand Registry. If no create button appears in the A+ Content Manager, you lack the permission. Then the trademark filing is the next step, not the image design.
Amazon's sales page names as an alternative 'or have generic products in your catalog'. An Amazon employee answers in the seller forum, however: 'A+ Content can only be used on ASINs associated with one of your Brand Registered Brands and Generic Listings cannot be changed to another brand.' He points out that for branded products you have to create new listings and ASINs. As long as both statements stand side by side, do not plan with the generic route.
Amazon describes the reseller role as 'for external, third-party sellers who are authorized to sell a brand's goods and services in the Amazon store'. According to the same page it grants access to A+ content, Brand Stores and Sponsored Brands — just not Brand Analytics. For the brand owner, the assignment is one click and costs nothing. Check the scope again yourself before asking; role permissions change.
No public Amazon document names the A+ body text as a ranking factor. The most frequently cited counter-source contains no controlled test and no data. It concedes itself: 'While you won't find an Amazon help article spelling it out for you'. Treat A+ as an instrument for presentation, not as a keyword field.
'Towel set' cut from the title and written into the A+ text instead.'Towel set' stays in the title; the A+ text explains material and care.Amazon explicitly advises against it in its public design guide: 'Avoid embedding text in images so mobile customers can read all the details on smaller screens.' Scaled-down images blur, and the text inside becomes unreadable. Module text stays readable on every screen.
The size 20 x 40 in only appears as print inside the module image.The size 20 x 40 in sits in the module's text field; the image shows only the towel.Amazon writes in the design guide: 'Don't include pricing, promotional, shipping details, QR codes, hyperlinks, directions to other web pages, or contact information' and 'Never reference or compare products with competitors'. The reason is structural: a module reviewed once stays online unchanged for a long time, while prices, promotions and shipping conditions change.
Now only 19.99 instead of 29.99, free shipping, grab it now.Terry in 100 percent cotton, 500 grams per square meter, in four colors.Amazon names as testable 'product images, titles, bullet points, descriptions, and A+ Content', brand story included, and recommends 8 to 10 weeks for robust results. The prerequisites are a professional account and the Brand Representative role for an enrolled brand. A plain before-and-after measurement without a control group is hard to attribute, because season, ad budget, price and competition fluctuate at the same time. That is our assessment, not an Amazon statement.
That your listing is online does not mean it is legally safe. Amazon enforces a few legal requirements technically — most of German and EU competition and labelling law, nobody there checks for you. This chapter covers EU and German law: it applies when you sell on the European marketplaces, amazon.de above all.
Your listing is attacked from two directions. Amazon suppresses it when catalog, title or image rules are violated. Competitors and competition associations send cease-and-desist letters (in Germany: "Abmahnung" — a formal warning letter with a cost note that must be taken seriously) when legally required information is missing. The two overlap in places, because Amazon enforces some legal requirements itself: the GPSR information, the EPR numbers and the unit price. The rest stays unchecked until someone sends a letter. The unit price shows the dual nature most clearly: Amazon technically enforces a German legal norm there, but never names the legal basis in the error message.
The fallacy behind it: That a listing has been online for two years only proves it meets Amazon's catalog rules. It says nothing about German or EU competition and labelling law. The sender of a cease-and-desist letter is never Amazon.
This section informs and does not replace legal advice. When in doubt, and before larger changes, better ask someone who is liable for the answer.
Since December 13, 2024, the EU General Product Safety Regulation applies. Article 19 requires that the distance-selling offer itself contains four blocks of information: manufacturer contact; for manufacturers without an establishment in the EU, additionally the responsible person in the EU; product identification with an image; and warnings in the language of the destination country. Because the regulation's text turns on the content of the offer, the prevailing reading is unambiguous. Händlerbund phrases it: a link to this information is not sufficient. The information belongs in the offer, not on a page behind it.
Amazon informed sellers in June 2025: if required offer information is missing or wrong, or the product is unbranded, the offer is deactivated within 24 hours of finalization. A newly created listing without GPSR data is therefore offline again right at the start. Two caveats you must know: the original notice sits in Seller Central behind the login and is only verifiable for you through the retelling in a trade medium. And whether Amazon still applies the 24-hour period unchanged today is not publicly documented. Report dated June 4, 2025.
Under art. 3 no. 8 GPSR, whoever markets a product under their own name or trademark counts as the manufacturer, regardless of who produced it. Practically every private-label seller is therefore the manufacturer in the legal sense. You cannot retreat to "I am only a retailer". You only need an additional responsible person in the EU if the manufacturer has no establishment in the EU. If your company sits in the EU, you are that person yourself. If you sit outside the EU, you have to appoint one.
Amazon can only calculate the unit price required by section 4 of the German price indication ordinance (PAngV) if quantity and unit exist as structured data. In 2021 Amazon announced it would remove offers of consumables without this information from search. Whether and in which categories Amazon still does that today is not publicly documented. The unit-price obligation itself does not hang on Amazon anyway: it applies regardless of category, and you are the one liable.
Net content: empty, unit: empty. Amazon cannot form a unit price.Net content: 1.5, unit: liters. Amazon calculates the unit price itself.Under art. 16 of the EU textile labelling regulation, the fiber composition must already be visible in the online offer — before the contract is concluded. The sewn-in label in the delivery is not enough. You may only state the composition using the names from Annex I. A brand name like "Lycra" or the US term "spandex" does not replace the official name; at most it may stand directly next to it. Even with correct percentages, the labelling is defective if the fiber name is not from Annex I.
95% cotton, 5% Lycra95% cotton, 5% elastaneFor packaging in Germany you need the LUCID number from the packaging register; for electrical and electronic devices, the WEEE registration number under section 6 of the German electronics act. Since July 1, 2022, a marketplace operator may no longer enable the offering of packaging subject to system participation as long as the producer is not registered. Amazon is fulfilling a legal duty of its own there, and blocks accordingly. A support case changes nothing; the offer only goes live with the number. And each registration is valid for one country only.
Since August 18, 2025, extended information duties apply to batteries and battery-containing devices. Required are statements on waste prevention, on the separate collection and treatment of used batteries, safety notes, and information on the environmental impact of hazardous substances. For distance selling, section 24(3) of the German battery act additionally points to article 74(4) and (5) of the EU battery regulation. That includes showing the costs of extended producer responsibility separately to the end user. Under the old battery directive, exactly that was prohibited. Whoever keeps using an old template can therefore have the opposite of what is required standing in the listing today.
The health claims regulation inverts the usual logic: only what is explicitly approved is allowed. Everything else is banned, even if it is true. Claims about curing or alleviating diseases are not permitted. Permitted are only the explicitly approved claims about reducing a disease risk, and only in the approved wording. The regulation counts as a market-conduct rule under section 3a of the German unfair competition act, so any competitor can send you a cease-and-desist letter directly. That your supplier prints a claim on the packaging is no defense: you are liable for the copy in the listing.
Advertising with test results is only permissible if the source and reference of the test are stated in an easily recognizable and findable way. If the reference is not easily recognizable, courts treat it as missing entirely. This also applies when you do not advertise the verdict yourself but the seal is merely visible on the pictured product. The seal itself is usually protected as a trademark and may only be used with a license.
From September 27, 2026, the rules of the EmpCo directive apply. The German implementation in the unfair competition act has been passed. General environmental claims such as "eco-friendly", "climate-neutral" or "sustainable" will then only be permissible if you can demonstrate recognized excellent environmental performance for the product. Also becoming impermissible: claims resting solely on offsetting, self-made seals without independent certification, and future targets without a verifiable and monitored implementation plan. The bans move into the blacklist of the unfair competition act and thus apply without case-by-case balancing.
Sustainably and climate-neutrally producedOuter fabric made of 80% recycled polyester (GRS-certified, certificate number in the product data sheet)From August 2, 2026, the transparency duties of art. 50 of the EU AI Act apply. There is no general labelling duty for all AI content. For you as a deployer, the visible duty kicks in for deepfakes within the meaning of art. 3 no. 60 — depictions that resemble real persons, objects, places or events and could falsely appear genuine. What matters is not the use of AI but the impression of authenticity. AI-generated product copy in commercial advertising is generally not covered. As a rule of thumb: the more photorealistic the image, the more likely it needs a label. The notice must be clearly recognizable at the first perception of the content; a note in terms and conditions or footnotes is not enough.
Independently of the AI Act, section 5 of the German unfair competition act remains in place — and misleading-advertising law exists in every market. Whoever invents colors, sizes or included accessories by AI deceives about essential characteristics of the goods. That is misleading, whether AI was involved or not. Purely technical optimization such as exposure and sharpness is uncritical; changing what the image asserts is not.
The AI image shows the towel in a stronger blue than what is delivered.The photo of the real towel, optimized only in exposure and sharpness.Is your listing violating Amazon rules without you noticing? The free check also tests compliance and flags risky wording.
An instant result. No sign-up.
The most expensive listing mistakes do not announce themselves: they trigger no account warning and only show up in your revenue — unless you look in the right places.
Four mistakes have something in common: search suppression, an overwritten title, a vanished buy button, and a keyword that is not indexed. None of them necessarily produces a warning in your account. So you cannot rely on a notification. If you know where to look, the walkthrough is done quickly.
With a search suppression your offer stays active but disappears from the search index. You do not necessarily get an active notification. What is reliable: Amazon lists the affected offers in this dedicated view. A large share of organic traffic runs through the search bar — revenue falls without any violation notice appearing in the account. The error field in this view names the concrete trigger.
Your fitted sheet has barely sold for days. There is no warning in the account, so you look for the fault in your ads first.You open the view for search-suppressed offers. The fitted sheet is there, complete with the error field naming the trigger.Amazon writes that titles over 75 characters will be "gradually updated to the AI recommendation", with the offers staying active. Brand owners get 14 days to review the recommendation before implementation. If you miss that message, you only notice the rewritten title through the ranking loss. Without a saved original version you cannot attribute the loss and cannot steer against it.
Your towel set slides in the ranking. What the title looked like before, you no longer know.You have a copy of all titles and see in the comparison which words disappeared from your towel set's title.Only that way do you see what your customer sees: whether the buy button is there, whether all images are being served, whether the GPSR information really sits on the live detail page, and whether a gap now yawns where a deactivated module used to be. The A+ Content Manager and the offer overview do not show you this reliably.
In the A+ Content Manager your module looks complete, so you assume it is on the detail page.You open the ASIN logged out in the browser and see an empty area where the module should be.If Amazon shortened or rewrote your title, keywords have dropped out. The ASIN test measures pure indexing, independent of competition, and thus answers the question with a clear yes or no. After the title migration it is especially urgent.
You type your keyword into search, don't find your bed sheet on page 1, and conclude it is not indexed.You run the ASIN test. It tells you, independent of competition, whether the keyword is indexed.If the ranking collapses without you having changed the listing, the cause is rarely in the copy. If the buy button is missing, conversion falls; with conversion, placement falls — and as a rule your Sponsored Products ads stop running too. That is how a price or account problem turns into a ranking problem within days.
Your bedding set slides, so that same evening you write new keywords into the title.You check in order: buy button there, price fine, stock available, account metrics clean. Only then do you look at the keywords.A large share of the circulating Amazon rules stems from US sources that name no marketplace. Some of it so far applies only to amazon.com — the renaming of the AI assistant, for instance: in the US it has been called Alexa for Shopping since May 2026, in Germany it is still Rufus. Other things very much apply elsewhere too; the end of cross-variant review sharing, for example, was announced by Amazon in the German Seller Central as well. So check in your own marketplace's Seller Central whether and from when a change applies to you.
You read about a change in a US blog and restructure your towel variants because of it.You verify the statement in your marketplace's Seller Central and only act once it says there that it applies to you.The eleven chapters are done. What follows, you do not have to read in one go: the 45 misconceptions clear out widespread assumptions, the limits table is the reference for characters, bytes and deadlines, and the 92 checked claims show which circulating numbers failed the cross-check.
Almost every one of them still appears in guides that rank well today. A few were right once and have not been since an Amazon change. Click to expand.
There is no Amazon source for an algorithm called A10. The name only appears in posts by service providers. A9 is not an algorithm name either: A9.com was an Amazon subsidiary that worked on search technology. That it built the amazon.de search is not what Amazon's 2004 press release says. An Amazon community manager wrote in the official seller forum that Amazon publishes no information and no documentation about its search and discovery algorithms. That is a forum post, not a help page, but it is the clearest publicly readable statement on the matter. In its research publications Amazon also describes its product search as an interplay of several methods for matching and ranking, not as a single versioned algorithm. If you have an observation, call it an observation — not an algorithm.
The question is open, and the trade articles clearly contradict each other. One side considers cannibalization the exception; the other sees a strong ranking effect from ads. There is no Amazon source describing the relationship. The only traceable path runs through the sale the ad triggers — and only if that sale is incremental. On keywords where you already rank near the top organically, the paid click mostly replaces a free one. Check this against your own numbers instead of trusting either side.
No Amazon document names external traffic as a ranking factor. Only the indirect path through sessions and sales is plausible. The catch is rarely mentioned: external traffic often converts worse than search traffic inside Amazon. It can therefore worsen exactly the signal you wanted to improve.
That is only partly true. Part of the space on page 1 is ad inventory and cannot be reached organically. How large that share is, Amazon decides as a matter of business policy — it is not an algorithmic constant. In their lawsuit of September 26, 2023, the FTC and 17 US states additionally accuse Amazon of deliberately showing worse-matching ads. Amazon disputes this. These are allegations, not judicially established facts, and they concern the US market. Measure your success by your position within the organic results, not by your place on the screen.
Amazon names several criteria side by side: competitive price, delivery speed and performance. Amazon does not publish a weighting, and being considered does not guarantee you the selection. More expensive offers with faster delivery can win the spot. That the landed price of item price plus shipping counts, rather than the item price alone, is not in Amazon's announcement but in trade articles. Compare your price as a landed price anyway. Otherwise you look cheap in the list and get rated as more expensive on the landed price.
That is the exact inversion of what Amazon wrote. Amazon explicitly stresses that it is not changing how the featured offer is selected and continues to select by price, delivery speed and performance. What was dropped is only the upstream yes/no check, which used to be a clear signal. How your offers stand is now best read from the trend of your featured-offer percentage in the business report "Detail Page Sales and Traffic". A single value says little.
Search suppression and account violations are two different things. With a suppression your offer stays active but disappears from the search index. You do not necessarily get an active notification, and a violation notice usually does not appear in your account. The error is not invisible though: Amazon lists the affected offers in Seller Central under Inventory, Manage Inventory in a dedicated view, including an error field with the trigger. If you never look there, you may only notice through your revenue.
That is outdated. Since July 27, 2026, a limit of 75 characters including spaces applies in all categories except media. The 200 characters were introduced in January 2025 and applied until July 2026. Longer titles can gradually be replaced by an AI suggestion. Brand owners get 14 days to review beforehand. Careful: Amazon's own public SEO page still lists the old 200-character limit to this day. What counts is the announcement of June 10, 2026, in Seller Central.
Amazon published the announcement in the German Seller Central forum as well. It therefore also applies to amazon.de. Onlinehändler News by Händlerbund reports the same for the European marketplaces.
Amazon writes that the offers remain active during the process. The title gets replaced, not the offer. That is no all-clear: the announcement says nothing about sanctions, and several trade sources additionally name the risk that overlong titles are served less. For you, the process is above all quieter than a block. Brand owners see the suggestion 14 days in advance. If you miss that message, you will only notice the new title through the ranking loss.
Cutting off costs you exactly the part that sets you apart. The front of the old title holds the brand and the generic product type. The back typically holds the features that distinguish you from the competition, and the niche keywords. If you cut mechanically at character 75, you keep the interchangeable part and lose the buying argument. Rewrite the title instead of truncating it.
That is a mix-up with the backend search terms. There, counting is done in bytes, which is why accented characters take two bytes each. For the title, Amazon phrases the limit as 75 characters including spaces. An accented character counts as one character there, exactly like a plain one. Anyone who writes ASCII workarounds into the title out of old backend habit gives away space and gains nothing.
Neither fits. Amazon's public seller blog says to separate words with spaces only and leaves lowercase up to you. Amazon describes the same field as an overflow for terms that fit nowhere else on the product page and advises against repeating information already in the title, brand or bullets. What effect a repetition would have on ranking, Amazon does not say. The argument lies elsewhere: every byte you spend on a repetition is missing for a synonym that appears nowhere yet.
That is unconfirmed. Händlerbund reported an extension in 2024; even the comments under the article doubted it. Amazon itself publicly names three different numbers: 250 bytes in the US seller forum, 249 bytes in the European forum and 200 bytes in an answer on flat-file error 97779. Amazon contradicts itself here. What exactly happens when you exceed the limit, none of these sources say. If you stay clearly below, you never have to settle the question.
There is no evidence of a lever effect. The field comes from the old Platinum Merchants program and still appears in some category templates. Amazon's attribute table has two separate columns, "Indexed by search?" and "Used by browse?", and can only be viewed with a seller account. That the field has no function at all is claimed, in agreement, by several service-provider blogs. There is no Amazon source for it. Practically this means: prioritize it last, behind the required and filter fields of your category.
Free ranking space they are not. Amazon's attribute table does not list target_audience_keywords and thesaurus_attribute_keywords as indexed by search. But it has a second column, "Used by browse?". Not search-indexed therefore does not mean without effect. Whether and when Amazon will remove the fields, no publicly verifiable source says. The table itself can only be viewed with a seller account. Your time is better spent on the required and filter fields of your category.
Amazon considers spelling mistakes unnecessary. Amazon's guidance says to include spelling variants, and that spelling mistakes are not necessary. Spelling variants are welcome, deliberate typos are not. They only cost you bytes that are then missing for a synonym.
Amazon writes: use either singular or plural; there is no need to enter both forms. Amazon recognizes the word forms itself. The second form costs you bytes that are then missing for a real synonym.
Amazon's requirements for bullet points name more than 10 and fewer than 255 characters per bullet, less in some categories. 150 characters are reported there. Where the 500 characters come from is open: none of the checked sources names that value, neither as a guideline nor as a field length. Violations of the requirements can lead to Amazon changing or removing a bullet. Which value applies to you is in the style guide of your category.
Amazon's search glossary says the opposite: bullet points are not always indexed by Amazon search, but always displayed in full on the detail page. Displayed and indexed are two different things. Your most important terms therefore belong at the front.
Amazon announced through the official forum account News_Amazon that after June 8, 2021, no HTML tags are supported on detail pages. Lines with an HTML tag are removed or displayed without formatting. The announcement is readable without a login. That bold tags still work on a few old listings changes nothing. Write the description as plain flowing text in short paragraphs.
A+ only covers the description visually. An Amazon employee confirms in the publicly readable seller forum that the field remains in Seller Central. Whether Amazon treats the description as a required field depends on category and template and cannot be verified publicly. Whether the covered text keeps being indexed, likewise. Both are open, and that is exactly why the math is simple: keeping it filled costs you nothing, and you keep a fallback in case A+ is rejected or removed again.
That cannot be substantiated. No public Amazon document names the A+ body text as a ranking factor. The most-cited counter-source contains no controlled test and no A/B data and itself concedes that there is no Amazon help article on this. Treat A+ as an instrument for the presentation of the detail page, not as a keyword field. And never move a ranking keyword out of the title or bullets into the A+ text.
Amazon itself writes that A+ content is free for eligible sellers. A single, widely recirculated trade report additionally says Premium A+ was rolled out broadly at the end of May 2026. There is no Amazon announcement about it, and no source says whether the change has reached every marketplace. Check in the A+ Content Manager whether the premium modules can be selected. That is the only reliable test here.
Brand Registry has an explicit reseller role for authorized third-party sellers. It grants access to A+ content, Brand Stores and Sponsored Brands — just not Brand Analytics. The prerequisite is that the brand owner assigns you the role. It costs them nothing.
Through Amazon's IP Accelerator you can register in Brand Registry immediately after filing the trademark application. For Germany, Amazon names negotiated capped fees of 360 euros (DPMA) or 540 euros (EUIPO) for the filing through a partner law firm, and charges no fee of its own. The official filing fees come on top.
Since February 22, 2026, Amazon's Content Quality Analysis (beta) evaluates published A+ pages weekly. It checks readability, completeness of information, visual presentation and conversion effectiveness; the result reads "Needs Improvement" or "Meets Standards". Approval at publication is no longer the end of the review.
Amazon Posts was discontinued. Since June 2025 no new posts can be created; in July 2025 the program ended. Amazon points to Brand Stores and Amazon Live as replacements. If a guide still recommends Posts in 2026, it is not being maintained. Check everything else in it as well.
On Amazon, the file name is a mapping key made of product identifier, variant code and extension. It is read by the bulk upload's routing mechanism, not by search. Extra characters can break the mapping. With a single upload through the web interface, the file name is arbitrary anyway. This is a Google Images rule transplanted onto a platform that works differently.
Three thresholds are being mixed up here. Amazon's own seller page names 500 to 10,000 pixels on the longest side and recommends images with more than 1,000 pixels per side for the zoom function. The 1,600 pixels are third-party sources relaying Amazon's optimum recommendation. An image with 1,200 px does not get rejected. With 2,000 x 2,000 px you sidestep the argument entirely.
Amazon requires pure white with RGB 255, 255, 255 and names no tolerance. The most common unnoticed causes are soft, semi-transparent cutout edges, and missing or wrong color profiles on export that shift the white.
Amazon does not prohibit AI editing. What matters is that the image correctly depicts the product actually delivered and follows the main-image rules. The only new thing since the end of July 2026 is the labelling duty for photorealistic AI people via the metadata keyword contains-synthetic-performer.
That sweeping claim is wrong in two directions. Amazon's own duty concerns exclusively images and videos with photorealistic, fully AI-generated people. And art. 50 of the EU AI Act creates no general labelling duty. For sellers, the visible duty only kicks in for deepfakes — depictions that resemble real persons, objects or places and could falsely appear genuine. AI-generated product copy in commercial advertising is generally not covered.
The rule depends on the category and is partly the reverse. By Amazon's clothing style guide, the main image for women's and men's apparel should show the product on a human model, because fit cannot be judged otherwise. Children's and baby clothing is photographed flat. For clothing accessories, no part of a mannequin may be visible. One note: the underlying style guide is from spring 2021. Check the current category style guide in Seller Central before the shoot.
The study exists, and it is serious. But it does not measure Amazon data: the Spiegel Research Center at Northwestern University examined the data of several online retailers in 2017 together with PowerReviews. The comparison was also against products with zero reviews. The jump from 0 to 5 is trivially large, and the study itself stresses that the added benefit drops off quickly after that.
The opposite is documented. Purchase likelihood peaks between 4.0 and 4.7 stars and falls again towards 5.0, because flawless ratings are perceived as manipulated. The number comes from a 2017 study not based on Amazon data. The direction is more reliable than the exact thresholds.
Such percentages circulate in SEO blogs without a traceable source, sometimes passed off as "Amazon's internal data" without any Amazon publication existing. Only the indirect path is traceable: reviews act on your conversion rate, and the conversion rate is considered one of the factors behind your placement. How strongly, Amazon does not document.
That confuses the US price (200 US dollars) with the German one. For amazon.de, Amazon's announcement in the German seller forum of October 19, 2023, names a tiered scale: 0 euros for up to 2 units, 70 euros for up to 10 units, and 170 euros for up to 30 units per parent ASIN. The figure is from 2023 and sits in a forum post, not on a help page. Check the conditions yourself in the Vine section before enrolling.
Incentives of any kind violate Amazon's review policy: coupons, free products, discounts, warranty extensions, insert cards. The consequence can be permanent account deactivation. In Germany there is more: annex no. 23c to section 3(3) of the unfair competition act classifies commissioning fake reviews as per se unfair, without balancing and without a materiality threshold.
The information must sit directly with the product. Händlerbund phrases it unmistakably: a link to this information is not sufficient. Art. 19 GPSR requires the information clearly and visibly in the offer itself — before the buying decision, not one click away.
Not with private labels. Under art. 3 no. 8 GPSR, whoever markets a product under their own name or trademark counts as the manufacturer, regardless of who produced it. If you import from Asia under your own brand, you are the manufacturer — not the factory.
You are liable for the content of your offer. For health claims, the prohibition principle of the health claims regulation applies on top: only what is approved is allowed. That a supplier handles it differently is no defense in a cease-and-desist case.
The CE mark is a manufacturer's self-declaration regarding the harmonized EU rules. It replaces neither the GPSR information duties in the offer nor an efficacy claim.
Not without naming the reference. Advertising with test seals is only permissible if source and reference are stated in an easily findable way. If the reference is not easily recognizable, it is treated as if it were missing entirely. The seal itself is usually protected as a trademark and needs a license.
Amazon checks catalog and platform rules: title format, images, compliance fields. A few legal requirements Amazon enforces technically along the way — the GPSR information, the EPR numbers and the unit price. The largest part of German and EU competition and labelling law, nobody there checks for you: textile labelling, health claims, test-seal references and environmental claims. Competitors and competition associations look at those. That an offer is online mainly says something about Amazon's rulebook.
The reference. If you take only one thing from this page, take this table.
| Field or rule | Limit | Evidence |
|---|---|---|
| Product title (all categories except media) | 75 characters including spacesApplies since July 27, 2026. Before that, 200 characters applied, introduced in January 2025. The exact day of the old policy is not documented. Characters are counted, not bytes: accented characters count once. That the limit applies per parent and child ASIN and to bundles rests on only two trade sources and carries PRACTICE there. | Amazon |
| Item highlights | up to 125 charactersAmazon names 125 characters as a ceiling, not a target. Comma-separated attribute phrases make sense: material, size, set quantity, use case. That the text appears under the title in search and on the detail page is not in Amazon's announcement, only at ZonGuru. The weighting versus the title is not documented. | Amazon |
| Word repetition in the title | at most 2x per wordPrepositions, articles and conjunctions are exempt. Whether plural and other word forms count, the cited policy does not say. At 75 characters, the limit is practically enforced anyway. | Amazon |
| Banned special characters in the title | ! $ ? _ { } ^ ¬ ¦Exception: part of the registered brand name. Banned are the negation sign ¬ and the broken bar ¦. Both are often confused with the vertical bar |. The vertical bar and the hyphen are not on this list. That emojis are also not allowed is a common reading and does not appear in the cited policy. | Amazon |
| Review period before automatic title rewriting | 14 days for brand ownersAmazon names the period explicitly for brand owners in Brand Registry. The quote itself is genuine and readable without a login. Whether the period also applies without Registry, Amazon does not say. The sources read this silence differently. Interpreting silence is not evidence. Without Registry, rely on no period. | disputed |
| Bullet points, number | at least 3, at most 5 (sellers); vendors up to 10 in some categoriesAmazon's requirements of August 15, 2024, name at least three bullets. Whoever provides fewer risks warnings, automatic interventions in the listing, or removal of the bullets. The sources relay Amazon's requirements; they are not Amazon pages. | Practice |
| Bullet points, length per bullet | 10 to 255 charactersThe sources' wording is more than 10 and fewer than 255 characters. At the edges that is a difference. In some categories the limit is lower; 150 characters are reported. Amazon's public seller blog recommends a one-to-two-word label plus an explanation of at most 100 characters. Binding is the style guide of your category. | Practice |
| Bullet points, total length | reported: at most 1,000 characters or 1,000 bytes across all bulletsThe value arithmetically conflicts with 5 times 255 characters, and the sources contradict each other on the unit. The main source can only be viewed with a seller account. The chapters do not use this value. Practical consequence: your most important terms in bullets 1 to 3. | disputed |
| Backend search terms (generic_keywords) | contested: 250, 249 or 200 bytesAmazon publicly names three different numbers. 250 bytes in the US seller forum, 249 bytes in the European seller forum, 200 bytes in an answer on flat-file error 97779. Amazon contradicts itself here. Bytes are counted, not characters: accented characters take 2 bytes each. What happens when you exceed the limit, none of these sources say. Whoever stays clearly below never has to settle the question. | Practice |
| Backend search terms, separators and form | spaces as the only separator; capitalization irrelevant; leave out articles and prepositionsAmazon phrases lowercase as permission, not obligation. Punctuation is unnecessary and only costs bytes. The list of don'ts (no brand names, not even your own, no ASINs, no judgments, no time-limited claims) sits in the German help text, publicly available only through a forum quote. It therefore carries only PRACTICE. | Amazon |
| Number of backend search term fields | 1 (formerly several separate lines)Amazon merged the formerly separate search term lines into a single field. What happened to the contents of long-grown catalogs in the migration is not publicly documented. Sellers report in the Amazon forum that they had to shorten and re-cut their old lines themselves. So export old listings once and check the byte length. | Practice |
| Product description | no publicly documented character count; the limit sits in your category's templateWhether the description is a required field depends on category and template and is not publicly documented. What is documented: the field remains in Seller Central and can keep being maintained, even with A+ content on top. The circulating practical value of 2,000 characters has no verified source. | Practice |
| HTML in the product description | not permittedamazon.de announced the shutdown after June 8, 2021. Lines with an HTML tag are removed or displayed without formatting. That bold tags still work on a few old listings changes nothing about the announcement. | Amazon |
| Main image, background | pure white, RGB 255, 255, 255Amazon names no tolerance. Most common causes of deviations: soft, semi-transparent cutout edges and missing or wrong color profiles on export. Export in sRGB. | Amazon |
| Main image, product share | at least 85% of the frameThe product must not be cropped by the frame. No text, no logo, no watermark, no props and no accessories — only the product being sold. On the other gallery images, text is normal. | Amazon |
| Image size | 500 to 10,000 pixels on the longest sideFor the zoom function Amazon recommends more than 1,000 pixels per side, not on the longest side. Practical recommendation: 2,000 x 2,000 px, square, JPEG in sRGB. That puts you on the safe side of every circulating figure. | Amazon |
| Image file formats | JPEG, TIFF, PNG, non-animated GIFAnimated GIFs are excluded. The chapter recommendation is JPEG. That PNG or TIFF bring no advantage over JPEG is not in the Amazon source. | Amazon |
| Images, number | 1 image required; Amazon recommends at least 6A video recommendation is not on the cited Amazon page. How many images the product page actually serves, Amazon does not document publicly. Count on your own live listing instead of relying on a circulating number. | Amazon |
| AI labelling for images and videos (Amazon) | keyword contains-synthetic-performer in the XMP field dc:subjectAnnounced July 23, 2026; the scope is Amazon's stores worldwide. Applies to images and videos in the listing and to A+ content. Only for photorealistic, fully AI-generated people. Not for real people (even after AI editing), not for characters from films, series and games, not for recognizably artificial depictions, and not for images without people. The rule is only days old — check the current state before uploading. | Amazon |
| Mandatory GPSR information in the offer | required since December 13, 2024; the information must sit in the offer itselfA link is not enough. Required are four blocks: manufacturer contact; for manufacturers without an EU establishment, additionally the responsible person in the EU; product identification with an image; and warnings in the language of the destination country. By its own notice, Amazon deactivates new offers without this data within 24 hours. That period carries only PRACTICE: the original notice sits behind the login; only the retelling of June 4, 2025, is public. Whether Amazon still applies the period unchanged is not documented. | Law |
| Unit pricing (German PAngV) | 1 kg, 1 l, 1 m, 1 m², 1 m³; for net quantities up to 250 g or 250 ml also 100 g / 100 mlThe unit-price obligation applies regardless of category, and you are the one liable. Amazon can only calculate the unit price if net content and unit exist as structured data. In 2021 Amazon announced removing consumables without this data from search. Whether and in which categories Amazon still does that is not publicly documented. | Law |
| AI transparency duties (art. 50 EU AI Act) | applies from August 2, 2026; fines up to 15 million EUR or 3% of worldwide annual revenueNo general labelling duty for AI content. You only have to visibly label deepfakes — depictions that resemble real persons, objects, places or events and could falsely appear genuine. What matters is the impression of authenticity, not the use of AI. AI-generated product copy in commercial advertising is generally not covered. The notice must be recognizable at first perception; terms and conditions or footnotes are not enough. For SMEs, the lower amount applies. | Law |
| Greenwashing ban in the German UWG (EmpCo implementation) | rules apply from September 27, 2026The German UWG amendment is passed and promulgated. Becoming impermissible: general environmental claims without demonstrably excellent environmental performance, claims based solely on offsetting, self-made seals without independent certification, and future targets without a verifiable and monitored implementation plan. The bans sit in the UWG blacklist and apply without case-by-case balancing. | Law |
| Textile labelling in the offer | fiber composition in percent, descending, only names from Annex I; fines up to 10,000 EURThe information must be in the online offer before the contract is concluded. A label only in the delivery is not enough. A brand name like Lycra or the US term spandex does not replace the official Annex I name. At most it may stand directly next to it. Correct is elastane. | Law |
| WEEE registration (electrical and electronic devices) | registration number requiredRegistrations are country-specific. A German number is not valid for France or Italy. For packaging, the LUCID number from the German packaging register comes on top. Since July 1, 2022, a marketplace operator may not let unregistered producers sell; Amazon blocks accordingly. On the size of possible fines this page deliberately states no number, see below. | Law |
| Batteries: online information duties | since August 18, 2025Required are statements on waste prevention, on the separate collection and treatment of used batteries, safety notes, and information on the environmental impact of hazardous substances. New: you have to show the costs of extended producer responsibility separately to the end user. Under the old battery directive, exactly that was prohibited. So replace old boilerplate completely, do not just add to it. | Law |
| Fake reviews | per se unfair, without balancing and without a materiality thresholdAnnex no. 23c of the German UWG covers, by its wording, the submission or commissioning of fake consumer reviews. Whether a genuine review prompted with a coupon falls under it, the norm does not say. Incentives of any kind are banned by Amazon's review policy anyway, up to account deactivation. For review requests, use only Amazon's Request a Review feature. | Law |
| Amazon Vine, fees on amazon.de | 0 EUR (up to 2 units), 70 EUR (up to 10), 170 EUR (up to 30) per parent ASINThe often-quoted 200 is the US price in US dollars. The evidence is an Amazon announcement in the seller forum of October 19, 2023, not a help page. The Vine help page sits behind the login. The fee only applies from 3 enrolled units, and only if the first review arrives within 90 days of enrollment. An ASIN can only be enrolled once; the tier cannot be changed later. Check the conditions yourself in the Vine section before enrolling. | Practice |
| Optimal star average | 4.0 to 4.7Towards 5.0, purchase likelihood falls again, because a flawless rating appears implausible to many buyers. Measured in 2017 on the data of several online retailers, not on Amazon data. The direction transfers; the exact thresholds do not. | Study |
| A+ content and Brand Registry | no extra fee; professional account plus Brand Representative or Reseller role for an enrolled brandAmazon writes that A+ content is free for eligible sellers. The professional account still costs money. The reseller role grants access to A+ content, Brand Stores and Sponsored Brands, not Brand Analytics. Assigning it costs the brand owner nothing. The generic-listings route Amazon additionally names is contested; an Amazon employee contradicts it in the forum. Do not plan with it. | Amazon |
| A/B test (Manage Your Experiments) | recommended runtime 8 to 10 weeksTestable are images, titles, bullets, descriptions and A+ content including brand story. Prerequisites: professional account and Brand Representative role for an enrolled brand. With the to-significance setting Amazon also ends the test earlier. Sensible on your highest-traffic ASINs. | Amazon |
| Revenue effect of A+ content (Amazon's own figure) | Basic up to 8%, Premium up to 20%Up to is a ceiling, not an average. Amazon names neither sample nor period nor category. Both values stood verbatim on Amazon's own publicly readable seller page on July 30, 2026. The chapters do not use the number. | Amazon |
| Revenue effect of a detail-page video (Amazon figure) | up to +9.7%90 days, US data 2022, methodology not disclosed. The source is a thread in the Seller Central forum, not an Amazon help page, hence PRACTICE instead of AMAZON. The also-circulating values 3.6x and 5x compare video watchers with non-watchers and mostly measure self-selection. | Practice |
| Removal of the featured-offer eligibility check | removal of eligibility requirements from July 2026, gradually worldwide through the end of 2026According to Amazon, the selection criteria stay unchanged: competitive price, delivery speed and performance. Only the upstream yes/no filter is dropped. Being considered does not guarantee the featured offer. The evidence is an announcement in the Seller Central forum, not a help page, hence PRACTICE instead of AMAZON. | Practice |
| Target audience, other attributes, platinum keywords | target audience and other attributes are not listed in Amazon's attribute table as indexed by searchAffected are target_audience_keywords and thesaurus_attribute_keywords. Not indexed by search does not mean without effect: the table has a second column, used by browse. For the platinum keywords there is no Amazon source, only agreeing service-provider blogs. This row therefore makes no statement about their function, only the advice to prioritize all three fields last. The attribute table is only reachable after login. | Practice |
| COSMO: improvement of relevance classification | +60% macro-F1 versus the best baseline (frozen encoders)A research benchmark for relevance classification — no statement about revenue or ranking. With fine-tuned encoders: +28% macro-F1 and +22% micro-F1. Both values stood verbatim in Amazon's own public blog post on July 30, 2026. COSMO derives its relationships from customer behavior, not from your listing copy. The practical advice in the chapter therefore rests on search behavior and carries only PRACTICE there. | Amazon |
These numbers circulate in guides, forums and agency posts. We checked them and did not adopt them: either the source could not be found, it said something different from what was claimed, or it was so old that it no longer proves anything today. When one of them crosses your path, you now know what is behind it. This section is not a confession. It is the reason you can trust the rest of the page.
Working through the points above takes about an evening per product. The Listing Check takes the first part off your hands: enter an ASIN, and within a minute you see which of the rules your listing violates. The Amazon listing generator takes care of the rest: images, copy and A+ content from one photo.
The check is free and needs no sign-up. For creating, you get 140 welcome credits — enough for a complete listing plus A+ content. Prefer learning step by step? The Listing track of the Listimo Academy covers all of this in 15 lessons — free.