Amazon A10 Algorithm: What Changed in 2026 and What Sellers Must Do Now

Search “Amazon A10 algorithm” and you will get a wall of contradictions. One article says it launched in 2020. Another says late 2025. A third dates the shift to 2022 and calls it “A9 to A10.” None of them cite Amazon, because Amazon has never used the name A10 for anything.

What changed is real. The name is not. And the gap between those two facts is exactly where most sellers waste their optimization budget this year.

What “Amazon A10 Algorithm” Actually Means (and Why Amazon Never Confirmed It)

Amazon’s product search runs on a system it built years ago, called A9. That name is real. Amazon has referenced it in its own documentation and in interviews with its search engineers. “A10” is not that. It is a label the seller community invented to describe how much ranking behavior has shifted since A9 was first documented, and it stuck because “A9, but different” is a mouthful.

Here is what actually happened, as best it can be pieced together from what Amazon has said publicly. Amazon layered an AI system called COSMO (Commonsense Knowledge Graph) on top of A9’s existing ranking mechanics. COSMO does not replace A9’s core signals like conversion rate or sales velocity. It sits above them, interpreting what a shopper is actually looking for before those signals ever get evaluated. A search for “shoes for a wedding” used to mean: find listings with those exact words. Now it means something closer to: find listings that fit what someone shopping for wedding shoes actually wants, even if the listing never uses the word “wedding.”

That distinction matters more than it sounds like it should. If you are optimizing a title to match search strings, you are optimizing for a version of Amazon search that has been partially replaced. If you are optimizing to make it obvious what your product is, who it is for, and what problem it solves, you are optimizing for the version that is actually live.

Most of the “A10 launch date” claims floating around blogs are unsourced. Some say 2020. Some say 2022 to 2023. Some say Q4 2025. If three sources disagree by five years on when something launched, the honest answer is that nobody actually knows, because Amazon never announced a launch. What we do know, because Amazon has confirmed it directly, is that COSMO exists and that it is deployed across search relevance and recommendations. That is the fact worth building a strategy around. The naming debate is not.

Part of why the confusion sticks around is that “A10” is a useful shorthand even though it is inaccurate. Agencies and tool vendors need a name to put in a headline, and “gradual, ongoing changes to A9’s ranking weights, layered with an AI relevance system” does not fit in a title. So the industry settled on a number that sounds official, borrowed the pattern from A9’s own naming, and never checked whether Amazon would confirm it. That is a marketing convenience, not evidence. It does not make the underlying changes any less real, but it should make you skeptical of anyone who states a specific “A10 rollout date” with confidence and no source attached.

A9 vs A10: What Actually Changed

Strip away the branding argument and there is a real, testable shift in what gets rewarded. The table below reflects the pattern most SEO practitioners and Amazon’s own advertising documentation describe, not a specific dated “version.”

Two things in that table deserve a second look. Ad spend still matters. It just does not buy rank the way it used to. A campaign that drives clicks without converting now reads to the algorithm as evidence your listing does not match the query, which can actively work against you. And external traffic is not a nice-to-have anymore. If a shopper finds you through Google or a creator’s post and buys, Amazon reads that as demand validation the same way it reads an organic click.

None of this means keywords stopped mattering. It means keyword density stopped being the lever. Backend search terms still index your listing. Titles still need your primary term. The difference is that stuffing no longer buys you anything past the first mention, and it actively hurts click-through rate on mobile, where a 200-character title just gets cut off mid-word.

The Ranking Factors That Matter in 2026

Amazon does not publish its ranking weights. Anyone who tells you “conversion rate is worth 34% of your score” is making that number up. What can be said with more confidence is the order of importance that shows up consistently across seller-tool data, Amazon’s own advertising guidance, and account-health thresholds Amazon does publish directly.

Conversion rate sits at the top. It is the cleanest signal Amazon has for whether a listing satisfies demand, because it directly reflects whether someone who saw the product bought it. A listing converting at 12% with modest traffic will generally outrank a listing converting at 5% with heavier traffic, because the algorithm is optimizing for revenue per impression, not raw volume.

Sales velocity still matters, but the window is short. Amazon reads recent performance, roughly the trailing 7 to 14 days, far more heavily than lifetime totals. A product that sold 10,000 units over three years but has gone quiet in the last two weeks will rank behind a six-month-old listing with strong recent sales. This punishes coasting and rewards launches that keep momentum going past week one.

Click-through rate tells Amazon whether your title, main image, and price are compelling enough to earn the click in the first place. A low CTR on a relevant listing usually means one of those three elements is weak, not that the product is wrong for the query. This is also where a lot of sellers misdiagnose the problem. A listing with a strong CTR but a weak conversion rate has a page problem, not a discovery problem, and the fix is different: better bullets, better A+ Content, clearer pricing logic, not a new main image or a different title.

External traffic quality, not volume, is the newer factor. A hundred clicks from an email list that convert at 15% help your ranking more than a thousand cheap social clicks that bounce. Amazon Attribution is the free tool that tracks which off-Amazon sources are actually paying off, and if you are running any external campaigns without it wired up, you are flying blind on the one metric that would tell you whether the spend is working. If PPC is where most of your budget currently goes, our breakdown of Amazon PPC strategy under the current ranking model covers how paid and organic signals interact in more depth than this section does.

Review quality now outweighs raw review count. A product with 200 reviews, 80 of them marked helpful, will generally beat a product with 500 reviews and almost no helpful votes. Recency matters too. A steady trickle of recent reviews signals an active, trusted listing better than a pile of old ones.

Return rate is tracked at the ASIN level and is one of the more punishing signals, because returns cost Amazon money directly on shipping and restocking. A listing converting at 12% with a 15% return rate performs worse in the algorithm’s eyes than one converting at 8% with a 3% return rate, even though the first one is selling more units.

Seller authority, meaning Order Defect Rate, late shipment rate, and account health standing, acts less like a single ranking factor and more like a multiplier across everything else. Strong account health does not directly push a listing up the page, but weak account health can suppress every ASIN in the catalog at once.

Engagement after the click is the factor most sellers never think to check. Time on page, scroll depth, and video views all signal whether a shopper who clicked actually stuck around, and a listing that gets clicks but sees people bounce in under five seconds tells the algorithm something is off even if the eventual conversion rate looks acceptable. Our guide to building A+ Content that actually gets read goes deeper on this, but the short version is that A+ Content and product video matter beyond just looking polished. They give the algorithm something to measure engagement against, not just a static page a shopper skims for two seconds and leaves.

Here is a rough benchmark table pulled from the ranges most commonly cited by Amazon-focused analytics tools and Amazon’s own published account-health targets.

Rufus to Alexa for Shopping: The Important Update

Amazon A10 algorithm: On May 13, 2026, Amazon shut down the standalone Rufus chatbot and folded its functionality into a new assistant called Alexa for Shopping, according to CNBC’s reporting on the change. Rufus’s recommendation logic and a shopper’s existing history carried over, but the standalone chat icon is gone. The entry point is now the search bar itself.

This is not a cosmetic rebrand. It is Amazon consolidating its answer to ChatGPT, Gemini, and Perplexity into a single conversational surface, and it changes what your listing needs to do. Under the old model, your product page was a container that had to match a typed query. Under the new one, it is also the source document an AI assistant reads to decide whether to recommend you at all, before the shopper ever sees a traditional results grid.

Two things follow from that. First, this AI layer runs alongside traditional ranking, not instead of it. A9’s core signals, conversion rate, sales velocity, CTR, still determine your position in a conventional search. Alexa for Shopping is a separate, additional surface where you can win or lose visibility based on different criteria: whether your bullets read like natural language, whether your Q&A section actually answers real buyer questions, and whether your reviews mention specific use cases instead of generic praise.

Second, most of the content published about “A10” this year was written before this consolidation, or ignores it. If you are getting your strategy from an article that never mentions Alexa for Shopping, you are working from a partial picture. Structured attributes, complete Q&A sections, and reviews that read like real answers to real questions are the groundwork for showing up in this surface. Keyword-stuffed copy is invisible to it.

How the 2026 FBA Fee Changes Interact With Ranking

Fee changes and ranking changes usually get covered as two separate stories. They are not separate this year. Amazon’s own 2026 fee update raised average FBA fees by roughly $0.08 per unit, a small headline number, but buried underneath it are structural changes that hit ranking directly.

The Aged Inventory Surcharge now begins at 181 days of stock remaining, roughly 90 days earlier than in 2025, while Returns Processing Fees have expanded across nearly every category instead of just apparel. Additionally, a 3.5% fuel surcharge started April 17, 2026, for standard FBA and extended to Multi-Channel Fulfillment and Buy with Prime on May 2.

The ranking connection is straightforward. Stockouts already damage momentum under the current model, because sales velocity resets and rankings do not recover the moment you restock. Now a stockout also triggers a surcharge earlier than it used to, before you are even out of stock. That means inventory planning is no longer purely a fulfillment cost decision. It is a ranking decision with a fee attached. Sellers who treat their reorder point as a pure cash-flow question, without accounting for the ranking hit that comes with running low, are underpricing the actual cost of a stockout by a wide margin.

Amazon A10 Algorithm Optimization Playbook: What Sellers Must Do Now

Write for intent, not keyword count

Your title, bullets, and A+ Content should state what the product is, who it is for, and what problem it solves, in that order, within the first 80 characters where possible. Titles over 150 characters generally lose click-through rate on mobile because shoppers cannot scan them, not because Amazon penalizes length directly. A title that reads like a sentence a person would actually say beats one built as a string of keywords separated by commas.

Fill in every discovery attribute in your product template, even the optional ones. Subject matter, target audience, intended use. These fields feed directly into COSMO’s intent matching, and leaving them blank means the algorithm has less to work with when deciding whether your product fits a situational search it cannot match on keywords alone.

Protect conversion rate before you scale traffic

If your conversion rate is under 5%, more traffic will not fix the underlying problem. It will just spend more ad budget failing to convert the same way. Fix the main image, the price positioning relative to competitors, and the first two bullets before increasing spend. Pricing meaningfully above the category median is one of the fastest ways to suppress conversion rate, independent of anything else on the listing, though the exact impact varies enough by category that we would rather point you to check your own price-to-conversion pattern in Seller Central than hand you a made-up percentage.

Set up Amazon Attribution before you run external traffic

If you are sending clicks from Google, social, or an email list without tagging them through Amazon Attribution, you have no way to know which channels are actually helping your rank and which are just burning budget. Set it up inside Seller Central, tag every external link you run, and review conversion rate by channel every month. Channels that convert well are worth scaling. Channels that mostly bounce are worth cutting, even if the raw click volume looks impressive on a dashboard.

Build reviews around helpfulness, not just volume

Use Amazon’s Request a Review button after every order rather than third-party review services. Encourage detailed feedback by following up, through compliant, non-spammy channels, with customers who had a good experience, since a review that mentions a specific use case is worth more to your ranking than five generic five-star reviews with no detail.

Treat account health as a ranking input, not just a compliance checkbox

Keep Order Defect Rate under 1% and late shipment rate under 4%. These numbers do not just protect your selling privileges. Sellers with consistently strong account health metrics see better visibility across their entire catalog, not just the listings directly tied to a defect.

Audit your listing on a phone, not just a monitor

Amazon evaluates mobile performance as its own signal, separate from your aggregate desktop and mobile blended metrics. A listing can be quietly underperforming on mobile specifically while your overall numbers still look fine, and you would never catch it without checking directly.

Build for Alexa for Shopping specifically

Write your Q&A section with actual buyer questions and complete, honest answers, seeded proactively rather than left empty. Make sure your reviews and A+ Content use language that matches how a shopper would ask a question out loud, not how a keyword tool would phrase it. This is separate work from traditional listing SEO, and skipping it means losing visibility on a growing surface most competitors have not adapted to yet.

Common Mistakes Sellers Make Under A10

The same handful of mistakes show up on almost every underperforming listing.

Keyword stuffing the title still happens constantly, even though it now actively hurts mobile click-through rate instead of helping ranking. A 200-character title packed with synonyms was smart advice five years ago. It is a liability now.

Chasing PPC volume without watching conversion is another common one. A campaign that drives a lot of clicks at a low conversion rate is not neutral. It can read as a relevance signal against you, since the algorithm interprets low-converting traffic as evidence the listing does not match what people searching that term actually want.

Ignoring return rate until it becomes a problem is a mistake that compounds quietly. Most sellers watch conversion rate closely and check return rate rarely, but a listing with a high return rate can lose rank even while conversion looks healthy on paper, because Amazon reads returns as a customer dissatisfaction signal independent of the sale itself.

Letting inventory run down without accounting for the ranking cost is the newest version of an old mistake. Under the 2026 surcharge timeline, a stockout now shows up on your fee statement earlier than it used to, on top of the ranking hit it always caused.

And leaving discovery attributes blank remains surprisingly common, given how little effort it takes to fill them in. Sellers spend hours optimizing bullet copy and skip the backend attribute fields entirely, even though those fields feed directly into how COSMO categorizes and surfaces the listing for situational searches.

Only checking listings on desktop is a smaller mistake with an outsized effect. A title or main image that tests fine on a monitor can fall apart at mobile thumbnail size, and since most Amazon sessions happen on a phone, that gap shows up in your CTR long before anyone thinks to check where the drop is coming from.

Amazon A10 Algorithm: 90-Day Action Plan 

Weeks 1 to 2. Audit account health metrics (ODR, late shipment rate, feedback score). Pull your current conversion rate, CTR, and return rate for every ASIN doing meaningful volume. Identify which discovery attributes are missing across your catalog.

Weeks 3 to 4. Rewrite titles and bullets for the listings with the weakest CTR or conversion rate first, not your whole catalog at once. Fill in every discovery attribute. Add or refresh main images and at least one infographic per listing.

Weeks 5 to 8. Set up Amazon Attribution if it is not already running. Launch or expand one external traffic channel (a blog, an email list, one creator partnership) and tag it properly from day one. Seed your Q&A sections with real buyer questions and full answers.

Weeks 9 to 12. Review inventory forecasts against the new 181-day low-inventory surcharge window. Request reviews consistently through Amazon’s compliant tools. Recheck conversion rate, CTR, and return rate against your week 1 to 2 baseline, and adjust whichever listing is lagging furthest behind the rest of the catalog.

If you are looking at a catalog full of listings built for the keyword-density model and want a second set of eyes on what is actually holding your conversion rate back, our listing optimization team can walk through a full audit of titles, images, and A+ Content against what’s actually driving rank under the Amazon A10 algorithm today.