Amazon & eBay Listing Optimisation: SEO Tips That Still Work in 2026
Amazon has never officially confirmed an algorithm called “A10.” The system most sellers refer to is still A9, substantially rebuilt around an AI layer called COSMO — a distinction that matters because a large share of “A10 optimisation” content is published by tool vendors with a commercial incentive to make ranking sound like a moving target only their software can track. eBay’s Cassini algorithm, by contrast, is real, named, and has been eBay’s core search system since 2013 — but even there, the specific factor weightings sellers cite (roughly 40-50% relevance, 30-40% seller performance, 20-30% listing quality) are industry-consensus estimates, not numbers eBay itself publishes. This guide separates what’s actually verifiable from what’s repeated marketing content on both platforms, so the time you spend optimising goes toward what genuinely moves visibility rather than chasing an unconfirmed moving target.
Marketplace sellers get flooded with “algorithm update” content because uncertainty is good for tool and service sales — a seller convinced that rankings depend on a mysterious, constantly-shifting formula is a more receptive buyer than one who understands the underlying mechanics are actually fairly stable and well-documented by the platforms themselves. This guide combines Amazon and eBay listing optimisation into one resource because the underlying discipline is the same on both: complete, accurate structured data; content that matches real buyer search intent; and performance signals (conversion, reviews, seller metrics) that reward genuine quality over keyword tricks.
Both sections below follow the same structure deliberately: what’s confirmed by the platform itself, what’s well-corroborated industry consensus, and what’s marketing content dressed up as insider knowledge. Knowing which category a specific piece of advice falls into is itself a useful skill for any seller navigating the enormous volume of “SEO tips” content both platforms attract — and it’s a skill that transfers directly to evaluating any future platform change or “new algorithm” claim that surfaces after this guide is published, not just the specific claims addressed here.
Amazon: what’s real, what’s myth
The A9/A10 confusion, resolved
“A9” is the seller-community name for Amazon’s product search and ranking system — Amazon has publicly discussed Amazon SEO in terms of keyword research and listing optimisation, but has never published the full ranking formula under any name. “A10” entered seller vocabulary more recently to describe apparent shifts in ranking behaviour, but multiple neutral sources — and a direct discussion thread on Amazon’s own Seller Central forum — confirm Amazon has not officially released a system by that name. The more accurate technical description: Amazon’s underlying search system (still generally called A9) has been substantially enhanced with an AI layer called COSMO (Amazon’s “common sense” contextual reasoning engine), which changes how listings get matched to search queries — inferring intent (“shoes for a wedding” implying formal dress shoes) rather than only matching literal keywords — without this being a wholesale replacement system deserving its own new name. The practical difference this makes for a seller optimising a listing is genuinely significant, which is exactly why the naming confusion matters beyond pedantry: a seller who believes “A10” is a confirmed, separate, more sophisticated system may reasonably conclude that older Amazon SEO fundamentals (relevant keywords, complete attributes, strong conversion signals) have been superseded by some new, undocumented ranking logic — when the more accurate picture is that those fundamentals still apply, now interpreted through a more contextually intelligent matching layer rather than replaced by a different one. Chasing a moving target that doesn’t actually exist as described diverts effort away from the fundamentals that remain the real, confirmed drivers of visibility.Why the “A10” framing persists anyway. Several Amazon seller tool vendors publish detailed “A10 ranking factor” guides, some citing “200+ factors” or specific percentage weightings with no traceable source. This level of specificity sounds authoritative, but Amazon has never published a ranking formula of this kind under any name — treat any source claiming precise, numbered A10 ranking-factor weightings with real scepticism, and prioritise sources (including this one) that clearly distinguish confirmed Amazon guidance from seller-community speculation.
What Amazon’s own guidance and consistent seller-reported patterns actually support
Setting aside the unconfirmed “A10” framing, the factors below represent the intersection of what Amazon’s own seller guidance discusses and what independent, consistently-repeated seller-reported analysis corroborates — the most defensible current picture available without relying on a single unverifiable source’s specific percentage weightings.| Factor | What it means practically |
|---|---|
| Keyword relevance | Title, bullet points, and backend search terms need to match real buyer search language — but COSMO’s contextual layer means exact keyword matching matters less than accurately describing what the product is and who it’s for |
| Conversion rate | The percentage of viewers who purchase; consistently cited as one of the strongest signals across independent seller analysis, since it directly reflects whether a listing satisfies the searches it’s shown for |
| Click-through rate | Whether shoppers click your listing when it appears in search results — driven heavily by the main image, title, and price relative to competitors |
| Reviews and ratings | Both a trust/conversion factor and a reported ranking input; listings with 100+ reviews are commonly cited as converting multiple times better than those with under 20 |
| Inventory availability | Stockouts interrupt sales velocity data across Amazon’s rolling performance windows, and a restock does not instantly restore prior ranking position — momentum has to rebuild |
| Account health | Order defect rate, late shipment rate, and policy compliance affect whether a listing is suppressed or demoted regardless of how well-optimised the content itself is |
The July 2026 title length change
Amazon’s own Seller Central announcement confirms that, effective July 27, 2026, product titles in all categories except media must be 75 characters or fewer, down from a previous 200-character ceiling — with Amazon’s own AI now auto-rewriting non-compliant titles rather than simply rejecting them. A new 125-character “Item Highlights” field holds details that no longer fit in the title and is separately searchable. Any Amazon listing optimisation advice still recommending long, keyword-stuffed titles is working from an outdated ruleset; front-load the single most important 75 characters and move secondary detail into Item Highlights and bullet points instead.Backend search terms and attribute completeness
Backend search terms remain a legitimate, low-risk place to capture alternate spellings, synonyms, and regional word variants that don’t fit naturally in customer-facing copy — but they supplement relevance signal, they don’t substitute for accurate, complete front-facing attribute data. Amazon’s discovery-related fields (subject matter, intended use, target audience) increasingly feed COSMO’s contextual matching directly, meaning a completely and accurately filled-out product template does more for AI-driven relevance matching in 2026 than backend keyword stuffing ever did. A practical test worth applying to any backend search term before adding it: would a real shopper plausibly type this exact phrase into Amazon’s search bar? Terms that fail this test (competitor brand names, irrelevant but high-volume keywords, repeated variations of the same word) risk a policy violation for the practice known as keyword stuffing, without the relevance benefit sellers hope to gain from including them.A+ Content and enhanced brand content
Enrolled brand owners can add A+ Content (formerly Enhanced Brand Content) to product listings — richer formatted modules combining lifestyle imagery, comparison charts, and expanded feature descriptions below the standard bullet points. Amazon’s own guidance and consistent seller-reported data both support A+ Content improving conversion rate, which is itself a reported ranking input — but it’s worth being precise about the mechanism: A+ Content most likely helps ranking indirectly, by improving the conversion signal Amazon’s system already weighs, rather than functioning as a separate, directly-weighted ranking factor in its own right. Treat it as a conversion-rate investment with a plausible ranking side effect, not a direct SEO lever to “optimise” independently of that conversion impact.External traffic: a genuine but overstated signal
Multiple sources report that Amazon’s system reads external traffic (visitors arriving from Google, social media, or influencer links) that converts into a purchase as a positive demand signal, and several describe this as an increasingly weighted factor in 2026 specifically. This claim deserves the same scepticism applied throughout this guide: Amazon has not published a confirmed mechanical relationship between external traffic volume and organic ranking. What’s more defensible is the underlying logic — a listing that converts well regardless of traffic source is, by Amazon’s own confirmed emphasis on conversion rate, going to perform better in ranking terms than one that doesn’t, whether that traffic originated on or off Amazon. Treat “drive external traffic” as sound advice for the genuine reason (diversified demand, brand-building, potential conversion-rate benefit) rather than because of an unconfirmed direct ranking bonus for the traffic source itself.eBay: what Cassini actually rewards
A brief, verifiable history
Cassini is eBay’s search ranking algorithm, introduced in 2013 — reportedly built under Hugh Williams, then eBay’s VP of Experience, Search and Platforms — specifically to replace the earlier Voyager system, which ranked listings mainly by recency and price. That older model rewarded sellers for constantly ending and relisting items just to appear fresh, regardless of whether the listing was actually likely to satisfy a buyer. Cassini’s founding principle, consistently described the same way across independent sources, was to rank by predicted likelihood of a completed, satisfying transaction instead — a listing with strong sales history and complete data could outrank a newer, cheaper listing under this model, a genuine philosophical shift from Voyager. Understanding this history matters for a practical reason: sellers who learned eBay optimisation under the older Voyager-era logic sometimes carry forward habits (constantly ending and relisting items to appear “fresh,” prioritising the lowest price above all else) that actively work against Cassini’s actual priorities more than a decade after Voyager was retired. Cassini rewards a stable, well-documented, consistently-performing listing over one that’s frequently relisted from scratch, since relisting resets accumulated sales and engagement history rather than preserving it — the opposite of what worked under the system it replaced.The three-factor weighting (industry consensus, not officially published)
| Factor category | Estimated weight | What it covers |
|---|---|---|
| Relevance | ~40-50% | Title keyword match, item specifics, correct category assignment |
| Seller performance | ~30-40% | Defect rate, late shipment rate, feedback score, Top Rated Seller status |
| Listing quality | ~20-30% | Image quality, description completeness, pricing competitiveness, return policy generosity |
This weighting breakdown is consistently cited across multiple independent eBay SEO analyses, but eBay does not officially publish these exact percentages — treat as well-corroborated industry consensus rather than a confirmed eBay-published formula.
Item specifics: the highest-leverage, most consistently corroborated factor
Every independent source reviewed for this guide agrees on this point without exception: item specifics (brand, size, colour, material, model number, and dozens of category-specific attributes) are weighted heavily because they power eBay’s filtered search directly. A buyer filtering by “Size: Medium” or “Brand: Levi’s” will never see a listing missing that specific, regardless of how well-optimised its title is — the listing is mechanically excluded from that filtered result set, not merely ranked lower within it. Fill every available specific eBay offers for your category, not just the fields marked required, since Recommended and even Optional fields are exactly the attributes buyers most commonly filter by in practice. This mechanical exclusion effect is worth dwelling on because it’s genuinely different from most other ranking factors covered in this guide: a weak title or a mediocre image lowers a listing’s relative position within a result set it still appears in, while a missing item specific removes the listing from an entire filtered result set altogether. A seller who spends hours polishing title copy while leaving half the available item specific fields blank has optimised for the wrong bottleneck — the specifics gate whether the listing is even in the competition, before title quality ever gets a chance to matter.Title structure and the 80-character limit
eBay allows up to 80 characters in a title, and Cassini indexes the title as its primary keyword-matching source — the description is read far less for matching purposes. Use close to the full 80 characters (wasted space is wasted ranking opportunity), lead with brand or model where applicable since early title tokens carry slightly more matching weight, and avoid manipulation flags like excessive capitalisation, “L@@K,” or emoji, which eBay’s listing policy treats as spam signals that can actively demote a listing rather than merely fail to help it.What doesn’t affect organic ranking, despite common assumption
eBay’s Promoted Listings (paid placement) and Cassini’s organic “Best Match” ranking run on genuinely separate tracks, confirmed consistently across independent sources: paid promotion buys placement above the organic feed for a targeted query, but stopping a promotion doesn’t change your underlying organic position, because promoted placement was never factored into the organic score in the first place. Sellers sometimes assume investing in Promoted Listings will lift their organic ranking as a side effect — it doesn’t, though the increased visibility can indirectly generate the sales history and engagement data that does feed the organic algorithm over time. eBay’s Promoted Listings program itself has evolved into multiple formats (Priority and Standard placement tiers as of 2026), but the separation from organic Cassini ranking is consistently described as unchanged across that evolution.The new-listing boost window
Multiple independent sources report new eBay listings receive a temporary visibility boost lasting roughly 7-10 days after publication, intended to give Cassini a chance to gather initial engagement data on a listing with no history yet. What happens during that window matters more than the boost itself: a listing that earns a sale, watches, or strong click-through during this initial period tends to hold a stable position afterward, while one that generates no engagement loses the temporary advantage and has to compete on the same footing as established listings with real sales history. This is the practical argument for front-loading effort — complete item specifics, a strong title, competitive pricing — into a listing’s first week rather than treating it as something to refine later once it’s “seen how it performs,” since the boost window is largely spent by the time a delayed optimisation pass happens.What’s genuinely different between the two platforms
| Dimension | Amazon | eBay |
|---|---|---|
| Title limit | 75 characters (as of July 2026) | 80 characters |
| Primary structured data | Backend search terms + discovery attributes (subject, intended use, audience) | Item specifics (brand, size, colour, material, and category-specific fields) |
| New listing treatment | Sales velocity builds from zero; no confirmed fixed “boost” window | New listings get a reported 7-10 day visibility boost, with position after that determined by engagement earned during the window |
| Paid-organic relationship | Advertising performance and external traffic are reported to correlate with organic gains over time, though Amazon does not confirm a direct mechanical link | Promoted Listings and organic Best Match are confirmed to run on separate tracks with no direct organic lift from paid spend |
| Algorithm naming clarity | Low — “A9” and unconfirmed “A10” both circulate, with genuine ambiguity about what’s officially real | Higher — Cassini is a named, dated, consistently-referenced system since 2013, even without official published weightings |
Managing listings across both platforms without duplicating work
Sellers active on both Amazon and eBay face a structural data problem before they face an SEO problem: the same underlying product needs to be represented as an Amazon listing (with backend search terms and discovery attributes) and an eBay listing (with a full set of item specifics) simultaneously, using each platform’s own field structure and character limits. Manually maintaining two independently-optimised listings per product is exactly the kind of process that degrades over time as a catalog grows — a title update made on Amazon after the July 2026 character-limit change doesn’t automatically propagate to the eBay listing, and an item specific completed on eBay doesn’t automatically populate Amazon’s discovery attributes, even though the underlying product fact (the material, the colour, the dimension) is identical on both. The more durable approach treats a single, governed product record as the source of truth, with channel-specific feeds generated from it — the same principle covered in our broader catalog management guidance, applied here specifically to the Amazon/eBay pairing. A canonical attribute (say, “colour: navy”) gets translated into Amazon’s backend search term format and eBay’s item specific field simultaneously from one update, rather than requiring a seller to remember and separately execute the same change twice across two different admin interfaces with two different field structures and two different character limits. This same centralised approach also resolves a subtler consistency risk: a product described slightly differently across two marketplaces (a colour called “navy” on one and “dark blue” on the other, for instance) creates confusion for a buyer comparing your listings across platforms and, more importantly, wastes an opportunity to accumulate a single, consistent brand and product identity that customer reviews, repeat buyers, and even off-platform search engines can recognise as the same underlying item. Consistency isn’t just an efficiency gain — it’s a trust and brand-recognition asset in its own right.Common mistakes on each platform
Amazon-specific mistakes
- Still writing 200-character titles. Post-July 2026, this triggers Amazon’s automatic AI rewrite rather than displaying the seller’s intended title — worth proactively updating rather than waiting for the rewrite to happen.
- Treating backend search terms as a substitute for accurate front-facing attributes. COSMO’s contextual matching increasingly relies on genuine attribute completeness, not keyword stuffing in a hidden field.
- Ignoring inventory continuity. A short stockout can cost more in rebuilding ranking momentum afterward than the lost sales during the stockout itself.
- Chasing “A10 ranking factor” checklists from vendor content instead of the platform’s own confirmed guidance and consistently-corroborated seller-reported patterns.
eBay-specific mistakes
- Leaving optional item specifics blank. Every source reviewed for this guide agrees this is the single highest-leverage, most commonly skipped action.
- Using gimmick title language (“L@@K,” excessive capitalisation, emoji) that eBay’s policy flags as manipulation, actively demoting rather than merely failing to help a listing.
- Assuming Promoted Listings spend improves organic rank directly. It doesn’t — the two systems are confirmed separate, though paid visibility can indirectly generate organic-feeding engagement data.
- Wasting the new-listing boost window by publishing an incomplete listing and planning to “optimise it later” once it’s live.
A practical listing optimisation checklist for both platforms
The overlapping discipline between the two platforms is larger than the differences — both reward the same underlying behaviour even though the specific mechanics and terminology differ:- Front-load titles with the single most important distinguishing detail — brand and core product type first on both platforms, within each platform’s current character limit
- Complete every available structured data field — backend search terms and discovery attributes on Amazon, every item specific (not just required ones) on eBay
- Write for the shopper first, since both platforms’ current systems (COSMO’s contextual matching, Cassini’s relevance-plus-quality scoring) reward content that clearly and accurately describes what the product is and who it’s for over content stuffed with disconnected keywords
- Treat reviews and seller performance metrics as ranking inputs, not just trust signals — order defect rate, late shipment rate, and review volume/quality feed directly into both platforms’ visibility systems
- Avoid manipulation tactics explicitly flagged by either platform’s policies — keyword stuffing, fake engagement, and gimmick punctuation in titles carry real suppression risk, not just an ineffectiveness cost
- Monitor inventory continuity on Amazon specifically, since stockouts interrupt the rolling sales-velocity data multiple sources report as central to ranking recovery time after restocking
A launch-week priority order
For a genuinely new listing on either platform, sequence the work in this order rather than treating every element as equally urgent: structured data completeness first (item specifics on eBay, discovery attributes and backend terms on Amazon), since these determine whether a listing is even eligible to appear for relevant searches at all; title and primary image second, since these determine click-through once a listing does appear; and pricing and policy competitiveness (shipping, returns) third, since these affect both click-through and conversion once a shopper has already clicked. Treating any of these as optional at launch — planning to “add item specifics later” once the listing is live — wastes exactly the new-listing boost window covered above, on both platforms, since eBay’s boost and Amazon’s initial sales-velocity accumulation both depend on the listing being fully competitive from day one rather than mid-optimisation.Image requirements: a shared but under-discussed lever
Both platforms weight image quality heavily in click-through rate, and both have specific technical requirements worth verifying directly rather than assuming: Amazon requires a pure white background for the main product image, at least 1000px on the longest side to enable the zoom function shoppers use to inspect detail, and prohibits watermarks or promotional text overlays on the primary image. eBay is somewhat more permissive on background but still penalises low-resolution or heavily-cropped images in its listing quality scoring, and multiple sources note that listings with 5 or more images consistently outperform those with only the minimum, since additional angles and context (scale, in-use shots) reduce the pre-purchase uncertainty that drives returns and hesitation alike. Image optimisation sits at an unusual intersection for both platforms: it’s simultaneously a compliance requirement (non-compliant images risk suppression on both platforms), a click-through driver (the primary lever multiple sources cite for whether a shopper clicks a listing in search results at all), and a conversion driver (additional images reducing purchase hesitation). Treating it as a single “upload a photo” checklist item undersells how much leverage sits in this one area compared to its perceived importance relative to title and keyword work, which tends to get disproportionately more seller attention.Measuring whether your optimisation work is actually paying off
Both platforms provide the data needed to check this directly, though under different names: Amazon’s Business Reports (session count, unit session percentage, and conversion rate by listing) and eBay’s Seller Hub Performance dashboard (impressions, click-through rate, and sell-through rate by listing) let you track the same underlying story on each platform — is a listing being shown, is it being clicked when shown, and is it converting when clicked. Track these three stages separately rather than watching only total sales, because a listing failing at the impressions stage (a discoverability problem, usually structured data or title relevance) needs a completely different fix than one failing at the conversion stage (usually price, images, or reviews) even though both ultimately show up as “not enough sales” if you only watch the bottom-line number. Run this measurement on a defined cadence rather than only when something feels wrong: a weekly check on new or recently-optimised listings during their first month, tapering to a monthly review for established, stable listings. This mirrors the same discipline covered in our broader guidance on ecommerce SEO governance generally — optimisation is an ongoing practice with a measurement loop built in, not a one-time task completed and then left alone indefinitely.When professional listing management makes sense
A seller with a handful of listings can reasonably manage both platforms’ optimisation manually using the checklist above. The calculus changes as catalog size grows: beyond a few dozen SKUs across two platforms, the manual burden of keeping titles compliant with Amazon’s current character limit, item specifics complete on eBay, and both platforms’ listings synchronised with the same underlying product truth becomes a genuine operational cost, not just an occasional task. This is precisely the point at which a governed, centralised product data approach — one source of truth generating both platforms’ feeds automatically — starts paying for itself in time saved and errors avoided, rather than being an unnecessary layer of process for a small catalog. A useful self-test for whether this threshold has been crossed: can you currently answer, without checking multiple systems, exactly which listings on either platform have incomplete item specifics or a title exceeding the current character limit? If that answer requires a manual audit rather than an immediate, confident response, the catalog has likely outgrown ad hoc, listing-by-listing management, regardless of the specific SKU count — the right trigger is process visibility, not an arbitrary size threshold. A seller who can answer this question confidently for ten listings but not for two hundred has already found their own practical scaling limit empirically, without needing a generic rule of thumb to tell them where it sits.Managing listings across Amazon and eBay at scale?
We handle catalog data, structured attributes, and listing optimisation across both marketplaces from one governed source of truth — no more manually managing item specifics and backend search terms in two disconnected systems. Talk to a team that’s done this for 17+ years.
Talk to the Team →