Check this page with an assistantOpens a chat asking it to summarise this article and name the evidence behind each claim.
Claude opens with the prompt on your clipboard: Anthropic does not support prefilled prompts on the web, and we would rather copy it than ship a button that drops it.
A Search Engine Optimising for Revenue
Google is trying to answer a question well. Amazon is trying to maximise revenue per search. That single difference in objective explains almost every divergence between the two disciplines, and it is why habits carried over from web SEO point slightly wrong here.
The mechanical consequence is a feedback loop web search does not have. If your listing ranks and people do not buy, Amazon has evidence it made a poor choice and your position degrades. If they do buy, it has evidence the opposite way. Ranking and selling are not sequential steps; they are the same loop.
That reframes things sellers usually treat as commercial decisions. Price is a ranking input. Your main image is a ranking input. Availability is a ranking input, because an out-of-stock listing converts at zero. None of that is true on Google, where a page can rank indefinitely while converting badly.
What Actually Moves Position
The reported factor list runs to hundreds of signals. The ones worth acting on for most sellers are few, and they cluster around whether people buy and whether they are happy afterwards.
| Factor | Why it counts | What sellers get wrong |
|---|---|---|
| Conversion rate | Direct evidence the result was right | Chasing traffic that browses rather than buys |
| Sales velocity | Demand signal relative to competitors | Expecting a new listing to rank before it has any |
| Reviews and ratings | Satisfaction signal, and a conversion driver | Treating them as reputation rather than ranking |
| Returns and complaints | Negative satisfaction signal | Overselling in the listing, which raises returns |
| Availability | An out-of-stock listing converts at zero | Letting stock lapse and losing accumulated position |
| Indexed terms | Determines which searches you can appear in | Repeating terms instead of covering them once |
| Converting external traffic | Demonstrates demand and produces sales | Sending untargeted traffic that depresses conversion |
The returns row is the one most sellers never connect to search. A listing that oversells produces sales and then produces returns, and the second signal cancels the first. Accuracy in a listing is not a compliance matter here; it is protection for the metric your ranking rests on.
Writing Listings for Rufus and for People
Amazon's conversational assistant reads your listing to answer questions like whether a product suits a particular situation. It matches meaning rather than strings, so a bullet answering "will this fit a small kitchen" is usable and a bullet repeating your keyword three times is not.
This is the same shift that happened in web search, arriving on a marketplace. The practical translation is to write bullets as answers rather than as specification lists. Who it is for, what problem it solves, what it does not do, which situations it suits. The specifications still belong in the listing, in the fields designed for them.
It also creates a discovery layer before the store. People increasingly ask an assistant which product to buy and arrive at Amazon already holding a name, which means being described accurately on sources those assistants read matters alongside your listing. That is the same third-party dynamic covered in earning third-party citations, and it does not resolve inside Amazon at all.
Keyword Coverage Without Stuffing
You still need to be retrievable for the terms people search. What has stopped working is repetition: covering a term once in an indexed field is what makes you eligible, and saying it again does not increase relevance while it does consume space that could answer a question.
Work from how buyers describe the problem rather than from how the category describes the product. Search term reports from your own advertising are the best source available, because they show the actual language of people who converted on your listing rather than a tool's estimate. That is the marketplace equivalent of the approach in long-tail keywords.
Then place terms where they are indexed, once each, in natural sentences, and use the remaining characters for persuasion. A listing that covers its terms and then explains why someone should buy beats one that covers its terms four times, on both the retrieval path and the conversion loop that decides whether you keep the position.
What Transfers From Web SEO, and What Does Not
Some instincts carry across cleanly. Others actively mislead, usually because they assume ranking is the goal rather than an intermediate step toward a sale that then feeds back.
Transfers well: understanding real search language, writing for the specific buyer rather than the generic one, and treating images as functional rather than decorative. The photography discipline in visual search applies directly, since a main image is doing conversion work in a fraction of a second.
Does not transfer: content length as a proxy for quality, link building, which does not exist here, and the idea that a strong position is durable. On Amazon a position is a lease renewed by every search that ends in a purchase, and losing stock for a fortnight can cost what took months to build. That fragility is the main thing to plan operations around.
Questions Sellers Ask About Amazon SEO
- How is Amazon search different from Google?
Amazon is optimising for revenue per search, not for answering a question. Its ranking systems weight whether people buy what they are shown, so conversion rate, sales velocity, and customer satisfaction signals carry weight that has no equivalent in web search. A listing that ranks and does not sell loses position; a page that ranks and does not convert on Google generally does not.
- What is the single biggest Amazon ranking factor?
Conversion rate, by most accounts. Amazon treats the proportion of people who buy after seeing a listing as direct evidence that the listing is the right result. That makes price, images, reviews, and availability ranking inputs rather than merely commercial choices, which is the part sellers coming from web SEO find hardest to accept.
- Does traffic from outside Amazon help my ranking?
External traffic that converts is generally understood to help, because it demonstrates demand and produces sales. The important qualifier is converting: sending poorly targeted traffic that browses and leaves can depress the conversion rate that the ranking depends on, making it worse than sending nothing.
- What is Rufus and how should it change my listings?
Rufus is Amazon's conversational shopping assistant, which reads listings to answer questions like whether a product suits a specific situation. It matches meaning rather than keyword strings, so bullet points written to answer real buyer questions serve it, while bullets that are a wall of specifications and repeated keywords do not.
- Do keywords still matter in Amazon listings?
Yes, for the classic retrieval path, which has not gone away. What has stopped working is stuffing: repeating a term does not increase relevance, and the space is better used answering a question. Cover the terms once, in natural language, in the fields that are indexed, and spend the remaining characters on why someone should buy.
