Short version: A9 is Amazon's real search and ranking engine, built by a real Amazon subsidiary — it ranks on two pillars: relevance (do you match the query) and performance (do you convert once shown). "A10" is not an official Amazon name — it's seller-community shorthand for how A9's performance weighting has shifted toward conversion rate, external traffic, and account health over time. Amazon has never published the exact formula; everything below is drawn from consistent, cross-source observation, not an Amazon disclosure.

"The Amazon algorithm" gets talked about like it's one mysterious black box, and a lot of seller-blog content adds to the confusion by presenting "A10" as a confirmed successor to A9 with its own official rulebook. It isn't. This is the grounded version: what A9 actually is, what's genuinely known about how it ranks, and where the "A10" shorthand is useful versus where it overstates what Amazon has actually said.

What A9 actually is

A9 isn't a marketing label — it's named after a real company. A9.com, Inc. was founded in 2003 as an Amazon subsidiary, originally built as a standalone web search engine before that public-facing product was shut down in 2008. Its search technology continued to power Amazon's own product search, and "A9" became the informal name for that engine.

Functionally, A9 does two jobs for every search query: it builds the candidate pool of listings eligible to appear (indexing), then it ranks that pool in the order shoppers see it. Both jobs run on every single search, and both are still active today — nothing has replaced this layer.

The two pillars A9/A10 ranks on

Across years of seller testing and industry analysis (Amazon has never published an official, weighted list), the ranking logic consistently breaks into two categories:

  • Relevance — does your listing match the query at all? Driven by keywords in your title, bullets, backend search terms, and category/attribute mapping. This determines whether you're in the pool in the first place.
  • Performance — among relevant listings, which ones does Amazon trust to convert? Driven by click-through rate, conversion rate, sales velocity, price competitiveness, review count and rating, inventory availability, and fulfillment method (FBA vs. FBM).

A listing has to clear the relevance bar before performance signals matter at all — the best conversion rate in the world doesn't help a listing that was never indexed for the query.

Is "A10" real? What the shorthand actually means

No, not as an official, separate system. Amazon has never confirmed an algorithm called "A10" in any public statement, patent, or Seller Central announcement. What's real is that sellers and industry researchers have consistently observed A9's performance weighting shift over time — reportedly leaning more on conversion rate, traffic from outside Amazon, review quality, and seller account health, and less purely on raw 30-day sales velocity than in A9's earlier years.

Use "A10" as shorthand for that shift in emphasis, not as a system with documented rules you can reverse-engineer. Chasing specific "A10 hacks" from SEO blogs is chasing a moving, unconfirmed target — the safer bet is optimizing the two pillars above, since both hold regardless of what the underlying weighting is called this year.

Where COSMO and Alexa for Shopping fit in

A9/A10 doesn't operate alone anymore — it's the foundation layer in a three-layer stack. COSMO, an AI commonsense-knowledge graph Amazon published as a research system at SIGMOD 2024, sits on top of A9/A10 and re-ranks the candidate pool by matching shopper intent, not just literal keyword overlap. Alexa for Shopping (the name Amazon gave Rufus on May 13, 2026, when it brought Rufus together with Alexa+) sits above that, pulling direct answers from your listing's attributes, reviews, and Q&A. Neither layer replaced A9/A10 — they both depend on it to have a candidate pool to work with. For the full three-layer walkthrough, see how Amazon search works in 2026, or the field-by-field playbook in our complete Amazon SEO guide.

Example — same product, two different A9/A10 outcomes

Listing A: title stuffed with keywords, no backend search terms filled in, 6% conversion rate, FBM with slow shipping. Result: indexed for a broad set of terms but ranks low on most of them — weak performance signals drag it down even where it's relevant.

Listing B: same product, title with primary keyword + brand, backend terms covering synonyms and misspellings, 22% conversion rate, FBA with Prime shipping. Result: indexed for the same core terms, but performance signals push it toward the top of the relevant pool.

How to optimize for A9/A10 specifically

  1. Win relevance first. Primary keyword in the title (within the 75-character limit), secondary terms in bullets and Item Highlights, and every synonym or misspelling you can fit into backend search terms.
  2. Then work performance. Price competitively against your true comparison set, keep inventory in stock (an out-of-stock listing loses ranking history), and use FBA or fast, reliable FBM shipping where possible.
  3. Protect conversion rate. Main image quality, complete attributes, and accurate bullets reduce returns and lift CTR and CVR together — both are performance signals A9/A10 tracks.
  4. Earn and monitor reviews. Review count and rating are consistently observed performance factors; using Amazon's Request a Review button the right way and a prompt customer service response help both metrics.
  5. Protect account health. Order defect rate, late shipment rate, and policy compliance are seller-level trust signals that seller-community consensus ties to A10-era ranking, on top of the listing-level factors above.

Common myths about A9/A10

  • "A10 is Amazon's official new algorithm." Not confirmed by Amazon — it's community shorthand for a weighting shift.
  • "Keyword stuffing still helps." It doesn't. Amazon indexes each keyword once regardless of repetition, and stuffing can hurt the clear, natural language COSMO rewards on top of A9/A10.
  • "Sales velocity is still the only thing that matters." It's one performance signal among several — conversion rate and account health carry real weight in current observation.
  • "COSMO replaced A9/A10." It didn't. COSMO re-ranks a pool that A9/A10 builds first; a listing invisible to A9/A10 is invisible to COSMO too.

A9/A10 optimization checklist

  • ☐ Primary keyword + brand in the title, within the 75-character limit
  • ☐ Secondary keywords in bullets and Item Highlights
  • ☐ Backend search terms filled with synonyms, variants, and misspellings not used elsewhere
  • ☐ Price checked against your real comparison set, not just your category average
  • ☐ Inventory monitored to avoid stockouts that reset ranking momentum
  • ☐ FBA or fast, reliable FBM shipping in place
  • ☐ Review count and rating actively monitored, with reviews requested through Amazon's Request a Review button
  • ☐ Account health metrics (ODR, late shipment rate) checked regularly

Amazon's own dashboards grade you on Amazon's terms

Amazon's free seller tools tell you whether your listing meets Amazon's own catalog rules. They don't tell you whether the copy itself is any good. OptimalCentral's AI audit judges your listing section by section against Amazon's listing rules and what shoppers need to see, in under a minute, with a 14-day free trial, and on the Brand plan, with your account connected, its optimizing agent acts on what it finds under the permissions you set.

Start free trial →

For the field-by-field build process, see how to optimize an Amazon listing, or start with Amazon keyword research to build your relevance foundation. Already have a listing live? Run it through the Amazon listing audit checklist, or see how the paid tools stack up in our roundup of the best Amazon SEO tools.

Frequently asked questions

What is the Amazon A9 algorithm?

A9 is the search and ranking engine that decides which listings appear for a given query and in what order. It was built by A9.com, Inc., an Amazon subsidiary founded in 2003, and it ranks products on two pillars: relevance (does your listing match the query's keywords and category) and performance (how well the listing converts and sells once shown).

Is Amazon A10 a real, separate algorithm?

Amazon has never officially confirmed an algorithm called A10. It's a label the seller-tool industry coined to describe how A9's ranking behavior has evolved over time — reportedly weighting conversion rate, external traffic, and seller/account health more heavily than pure sales velocity. Treat it as a useful mental model for a shift in emphasis, not a documented, separate system with its own rules.

What are the most important A9/A10 ranking factors?

Widely observed factors split into relevance (title, bullets, backend search terms, category and attributes matching the query) and performance (click-through rate, conversion rate, sales velocity, price competitiveness, review count and rating, inventory availability, and fulfillment method). Amazon has not published an official, weighted list — these are drawn from consistent seller and industry observation, not an Amazon disclosure.

Does COSMO replace A9/A10 ranking?

No. COSMO is an AI commonsense-knowledge layer that sits on top of A9/A10, matching shopper intent to products even when the exact keywords don't overlap. A9/A10 still builds the initial candidate pool from your indexed keywords and ranks by relevance and performance; COSMO re-ranks and expands that pool afterward. Neither replaced the other.

How is A9 different now than when it launched?

A9 launched focused heavily on keyword matching and raw sales velocity. Over time — the shift the community calls "A10" — Amazon has reportedly leaned more on conversion rate, external traffic sources, review quality, and seller account health as tie-breakers among relevant listings. The underlying engine and its two-pillar structure (relevance, then performance) hasn't changed; the weighting within performance has.