Three things are true about your listings at the same time. The algorithm almost everyone names does not exist under that name in any Amazon document, the title limit most guides quote was replaced on July 27, 2026, and a share of your catalog may not be competing at all because Amazon never put it in the search index. Contradiction is not confusion. Amazon does rank listings, it does so on inputs you control, and the reason so much optimization spend produces nothing is that sellers buy the fix before they diagnose the problem. So which of the three failures is yours: not indexed, indexed and ineligible, or indexed, eligible and buried? Answer that wrong and the next quarter of work goes into the wrong field. The pages that should tell you the difference are still printing a 200 character title rule.
Marketplace SEO is not open-web SEO with different keywords. No backlinks, no domain authority, no crawl budget. Amazon owns the index and matches a query against fields inside your listing.
Reviewed October 2026. Amazon changed a title rule on July 27, 2026 and added an indexed field in August. Every platform claim below is dated, third-party estimates are labelled as third-party estimates, and Amazon-owned sources are named.
Indexed Is Not Ranked, and the Difference Costs Real Money
Google asks whether a page deserves to be in the results. Amazon asks two questions: is this listing allowed into the candidate set, and where does it sit inside it? Three failure states exist, in this order of severity.
Not indexed at all. The listing exists in your seller account but never entered Amazon's searchable index. The seller-side test: search your own ASIN. If it does not appear for itself, no keyword work matters yet.
Indexed but search-suppressed. The detail page resolves through a direct link, but Amazon removed the ASIN from search and browse because content is missing, non-compliant, or under review. Amazon names the reason in Fix Your Products or Manage All Inventory. A catalog problem, not a copy problem.
Indexed and eligible, but not ranking on a term. The listing appears for its own ASIN and for broad queries, but not for the buying term you care about. The only case where relevance and conversion work applies.
That order has commercial weight: a seller with the first or second problem who buys keyword optimization has bought the wrong service. The same discipline appears on a direct storefront, the spine of our guide to Shopify SEO services.
How Amazon Ranks a Listing in 2026: Two Input Sets, Two Jobs
Amazon matches a query against fields inside the listing, not against a page. The indexed fields are the title, bullets, description, backend search terms, category-specific structured attributes, and category placement itself. Each decides eligibility. None decides position.
Position comes from behavioural history on that listing:
- Click-through rate, driven by the main image, price, rating, review count, and how readable the title is in a mobile result.
- Conversion rate on the sessions the listing receives, which Amazon reports as Unit Session Percentage in Business Reports.
- Sales velocity: recent volume and its consistency, not lifetime units.
- Customer experience signals: returns, refunds, negative feedback, order defect rate.
Relevance gets you into the room and conversion decides whether you get the good chair. Perfect keyword coverage with a weak Unit Session Percentage means eligibility for a hundred terms and a rank on none of them.
The A10 Question: What Is Actually Documented
Two parts, because the field only tells you one of them.
What is verified. Amazon has never announced, documented, or named a search ranking algorithm A10. There is no Amazon help page, seller announcement, developer documentation, or research paper using the name. Two independent 2026 treatments make that argument directly (evolveamz.com, published 2026-04-24 and updated; sellershorts.com). The second names the origin: A9 traces back to A9.com, a former Amazon subsidiary that worked on search technology.
What is also true. The name is used as established fact across most of the field, including the guide ranking near the top for this query. Saying A10 does not exist while you look at ten pages that say it does would be arguing with your own screen. The narrower truth: the label is community shorthand, the behaviours are real, and the percentages sold alongside it are invented.
What is documented is a stack.
COSMO is real and peer-reviewed. "COSMO: A Large-Scale E-commerce Common Sense Knowledge Generation and Serving System at Amazon" (Yu, Liu, Maia et al.) ran in the SIGMOD-Companion 2024 proceedings, pages 148 to 160, DOI 10.1145/3626246.3653398. It describes a commonsense knowledge graph across 18 product categories, roughly 6.3 million nodes and 29 million knowledge edges, built by having a large language model hypothesise why each sampled purchase happened and keeping only hypotheses human annotators judged plausible. The paper reports around 60% improvement in search relevance and a 0.7% live sales uplift on 10% of US traffic.
Literal keyword matching is therefore no longer the whole relevance story, because what your listing means is part of how the system connects a query to a product. A search for shoes for pregnant women has to reach slip-resistant shoes without the word appearing in the query. Rufus is Amazon's AI shopping assistant, a conversational layer rather than a ranking algorithm. Amazon's Q4 2025 earnings call, reported second-hand, puts it at 300 million customers and 38% of shopping sessions by Black Friday 2025. Separately, Amazon's Q2 2026 call on 2026-07-30 reports a shopping assistant used by over 350 million customers in the last 12 months, phrased around Alexa for Shopping. Two different products in Amazon's own wording, so do not merge them.
Third-party sources report roughly 7 to 14 days for listing edits to surface in the semantic layer, so treat that as an expectation and do not re-edit on day three.
The Fix-First Order for a Live Listing
Most guides start with keywords. On a live listing that is usually the third fix.
- Check whether the ASIN is indexed. Search your own ASIN. If it does not appear for itself, fix that before spending anything on copy.
- Check for a suppression alert. Fix Your Products and Manage All Inventory name the missing field or policy when Amazon has pulled a listing from browse.
- Fix the catalog, not the copy. Category, mandatory attributes, variation relationships. Missing mandatory attributes are a named trigger for suppression.
- Make the term exist somewhere indexed, once. Title if primary, bullets if supporting, backend if it is a synonym that does not fit visible copy.
- Only then treat it as a conversion problem. If the term is indexed and the listing is still buried, the missing input is behavioural.
That last step is the uncomfortable one. Buyers of SEO services for small business arrive assuming a text problem is causing a ranking problem, and frequently it is not.
Title, Bullets and Backend Search Terms: What Each Field Does
Check this against your own Seller Central screen, because the live guides disagree with the platform.
Title: 75 characters, enforced 2026-07-27
Titles in all categories except media must be 75 characters or less including spaces, effective July 27, 2026, in Amazon's own Seller Central wording. The new companion field, Item Highlights, allows up to 125 characters, is indexed for search, and sits below the title, reported visible from August 10, 2026 on desktop and mobile.
Break the limit and Amazon rewrites the title with an AI-generated recommendation, gradually and on its own schedule, with no seller approval step. Brand Registered sellers reportedly get a 14 day review window in Review Listings Changes. Sellers without Brand Registry do not (third-party reporting).
Now the correction. Live results still tell sellers Amazon allows 200 characters. One 2025 guide states 200 directly, another still lists it as operative, and the guide ranking near the top says "under 80 characters when possible, never over 150". None reflects the enforced 75 character limit. Two substance rules survived: the first segment of the title carries the weight, and best seller or top rated language is disallowed.
Bullets: five slots, and the ceiling is not the target
Amazon allows five bullet points, with a per-bullet ceiling in the high hundreds of characters. The 1,000 character figure quoted across the tooling category is a third-party number, not an Amazon-published one. The practical target is far lower: the field's recommendation of 150 to 250 characters is sound, because a mobile shopper sees only the first two or three (third-party recommendation). Bullets are indexed, and they are where objections get answered. Writing them as keyword containers and wondering why the Unit Session Percentage is flat is the most common self-inflicted failure here.
Backend search terms: less than 250 bytes
Amazon's own wording for this field is less than 250 bytes, with a warning message shown beside it on the Keywords tab when you exceed the limit. Bytes, not characters. Accented characters and non-Latin scripts consume more than one byte each, so a seller pasting Spanish or German keywords runs out of room sooner than one using plain ASCII.
The field is repeated back wrong constantly. The guide ranking near the top states 249 bytes. Another live 2026 guide states 500 characters, nearly double the real budget. A third states 250 bytes correctly. Three pages, three limits, one field. Put synonyms, spelling variants and long-tail terms there that do not appear in your visible copy, and skip singular and plural, because Amazon treats them as the same term.
Attributes: how the listing tells the system what it is
Category-specific attributes such as material, size, colour and intended use belong in the relevance set, and the completeness checks in Seller Central look at them. They matter more since COSMO, because a semantic layer needs structured meaning.
Images and A+ Content: The Conversion Half of the Ranking Equation
Everything here affects rank through conversion, not relevance, which is why a photo swap will not lift your position in week one.
Amazon's own seller-facing photography standards: pure white background (RGB 255, 255, 255), product filling 85 to 100% of the frame, minimum 1,000 pixels on the longest side for the zoom function with 1,600+ pixels optimal, JPEG preferred. The main image is your fastest click-through lever, because it is what a shopper reacts to before reading a word. If impressions are healthy and clicks are not, start here rather than in the copy. The rest of the gallery should cover 6 to 9 images (third-party consensus).
A+ content is the objection-answering layer, available to Brand Registered sellers on a Professional plan, below the bullets and replacing the standard description. Here the disclosure discipline matters, because the field inflates the number. Amazon's own published guidance states up to about 8%, as a maximum and not a promise. The 5 to 10% Basic and 15 to 20% Premium figures on agency and tool pages carry no attribution. If a page quotes 20%, ask which of those it came from.
One decision rule almost nobody states: the percentage is the same for every listing and the money is not. A high-traffic ASIN at 30,000 sessions a month converting at 12% gains hundreds of orders from a one-time build, while a 300 session long-tail SKU gains a rounding error from the same work. Roll A+ out top-revenue-first, the opposite of how most catalog teams sequence it.
One more consequence of the AI layer: Amazon's assistant and the semantic layer read the whole page, bullets, description, A+ modules, Q&A and reviews included. Q&A has been visually demoted while still being read, so treat it as indexable content rather than a support appendix.
Diagnosing a Listing That Will Not Rank
Three failures, three different first moves. Find yours before you spend.
Case 1: indexed but not ranking on your term
Symptom: the listing appears for its own ASIN and for broad category queries, but not for the buying term you care about.
First moves, in order. Search the exact term to see whether it is indexed at all. If not, put it in the field where it belongs and nowhere else. If it is indexed and the listing is still buried, this is a conversion problem, so feed the term targeted traffic and fix the page blockers: main image, price positioning, review count, first title segment. Then wait.
Case 2: ranking but not converting
Symptom: sessions arrive and the Unit Session Percentage stays flat or falls.
First moves, in order. Price and offer first, the fastest conversion lever and the one a content agency least wants to touch. Main image and gallery second. A+ third, around real objections rather than brand aesthetics. Bullets fourth, rewritten as answers. Reviews and Q&A fifth.
One warning on benchmarks. Published Amazon conversion averages contradict each other badly: 5 to 10%, 3 to 10%, and 10 to 15% all appear in 2026 sources quoting different samples. Compare only against the same category at the same price band.
Case 3: both failing
Symptom: the listing does not rank, and the sessions it does get do not convert.
First moves, in order. This is usually a catalog problem. Read the suppression alert in Fix Your Products first, because it names the field or policy at fault. Confirm the category and every mandatory attribute. If the listing is active, in stock, correctly categorised and not suppressed and still will not index, the standard escalation is a full attribute update through a flat file to force a re-index, plus a case through Brand Registry support. That last step is community advice, not Amazon policy.
Keyword work is the third fix, not the first, and on a suppressed listing it is no fix at all.
Amazon SEO Tools and Services: What to Buy, and What the Bands Really Are
The tooling category does four jobs: keyword discovery, rank tracking, listing audit, and PPC reporting. Most roundups bundle them into one grid.
Entry pricing is misreported too. A live 2026 roundup lists Helium 10 at $29 per month. Two independent pricing trackers report that Helium 10 discontinued its Starter plan during 2026 and now sells Platinum at $129 per month month-to-month or $99 billed annually, Diamond at $359 or $279 annual, and Enterprise at $1,499 per month. The $29 figure matches the discontinued plan, so check the vendor's own page before budgeting from a roundup.
On the service side, the published bands disagree by a factor of four, which is the single most useful thing to know about buying amazon seo services.
Source | Reported band |
|---|---|
WebFX (Amazon SEO pricing page) | $99 to $2,000 per product; $450 to $1,800 per month for progressive optimization |
SalesDuo (Amazon SEO pricing, 2026) | $50 to $150 per ASIN; $200 to $1,500 one-time per listing; $2,000 to $10,000 per month retainer; $500 to $5,000 per audit; $500 to $3,000 per month freelancer |
SupplyKick (Amazon agency pricing, updated 2026-09-14) | $1,500 to $25,000+ per month depending on model and brand size; tools $200 to $2,000 per month |
Nobody is lying. The ranges describe different scopes. Per-product pricing is a one-time listing rebuild, while a monthly retainer buys ongoing rank monitoring and testing. Only the retainer can respond when Amazon changes a field rule mid-year, which it did on July 27, 2026. That is the question to put to an amazon seo consultant: what happens to my titles when the rules change again?
One trap from SupplyKick's own analysis: tool access locked inside an agency's proprietary dashboard, gone at contract end. Ask for the data in your own account first.
A single marketplace listing needs a marketplace tool plus a freelancer. A storefront plus a direct site, paid media and reporting needs more than marketplace tooling, and that crossover is what our keyword research layer covers. The boundary is blunt: Allable does not pull Amazon search volume, does not grade an ASIN, and cannot clear a suppression alert. If you also run a direct storefront, the diagnosis-first habit is what our keyword research for AEO guide applies to generative search.
What to Do Before Your Next Listing Edit
Check your ASIN in the search bar, read the suppression alerts, then confirm the field limits you are building to are the ones Amazon actually enforces. The weightings nobody can verify matter far less than the fields you can fix this week.
Frequently Asked Questions
What is Amazon SEO?
Making a listing eligible for a query and then competitive inside the eligible set. It covers six indexed fields (title, bullets, description, backend search terms, structured attributes, category) and the behavioural signals that decide position: click-through rate, Unit Session Percentage, recent sales velocity, customer experience metrics. No backlinks or domain authority are involved.
How does the Amazon A10 algorithm work?
Amazon has never published or named an algorithm called A10, so no documented ranking factors exist under that label. The behaviours sellers attribute to it are real: relevance gates eligibility and conversion decides position among eligible listings. The documented stack is A9 as a community label, COSMO as the peer-reviewed knowledge graph (SIGMOD-Companion 2024, DOI 10.1145/3626246.3653398) and Rufus as the conversational assistant. Any weighting sold under the name A10 is unsourced.
How do I rank higher on Amazon?
In this order. Confirm the ASIN is indexed by searching it. Clear any suppression alert in Fix Your Products. Fix the category and mandatory attributes. Put your target term into one indexed field, once. Only then work on conversion: price and offer, then the main image, then A+ content, then bullets. Keyword work cannot rescue a listing Amazon has removed from search.
Does Amazon SEO still work in 2026?
Yes, and the old playbook is now a liability. The title limit dropped to 75 characters including spaces on July 27, 2026, with Item Highlights indexing up to 125 characters from August 10, 2026, so titles built to a 200 character rule get rewritten by Amazon's AI without a seller approval step. Backend search terms remain a byte limit of less than 250 bytes.
How much do Amazon SEO services cost?
Published bands run from about $50 per ASIN to $25,000 or more per month, a four-times spread that reflects scope rather than dishonesty. One-time listing work sits at $99 to $2,000 per product or $200 to $1,500 per listing. Retainers run from roughly $450 per month for progressive optimization to $2,000 to $10,000 per month for a full agency. Per-product pricing rebuilds a listing once; a retainer is what can respond when Amazon changes a field rule mid-year.
