On this page
What Are Amazon Sponsored Prompts?
Key Takeaway: Amazon Sponsored Prompts are Sponsored Products and Sponsored Brands placements that appear inside conversational shopping answers rather than on a search results page. They launched on 25 March 2026 and run at standard CPC. In May 2026 Rufus was folded into Alexa for Shopping, making conversational the default surface rather than a feature.
The mechanics are ordinary — you bid, you pay per click, it draws from your existing campaigns. What changes is the context. A shopper asking "what's a good beginner espresso machine under $300?" gets an answer, not a grid. Somewhere in that answer sits a sponsored placement.
That difference is larger than it sounds, and it changes three things: how many slots exist, what decides whether you fill one, and how confidently you can measure any of it.
The Shelf Just Got Much Smaller
A search results page shows dozens of products. A shopper scrolls, compares, opens tabs. Even a middling ranking gets seen.
A conversational answer recommends two or three things. There is no page two. The shopper asked a question and got a reply, and the reply either contains your product or it does not.
This concentrates outcomes sharply. Winning a Prompts placement is worth more per impression than a mid-page organic position, because it is one of very few options presented at the moment of decision. Losing it costs more too — there is no long tail of visibility to fall back on.
Dimension | Search results page | Conversational answer |
|---|---|---|
Products shown | Dozens, plus pagination | Typically two or three |
Shopper behaviour | Scroll, compare, open tabs | Read or hear one reply |
What wins the slot | Bid, relevance, imagery | Relevance and attribute match |
Value of losing | Still visible further down | Not present at all |
Main image matters | Heavily | Little or not at all |
Placement reporting | Established | Limited |
It also raises the relative importance of relevance over bid. Amazon has an obvious incentive to keep conversational answers useful; an answer stuffed with poorly-matched sponsored results damages the surface. Expect relevance thresholds to be doing more work here than on a results page.
Your Listing Is Now Read by a Model
This is the part with real operational consequences.
On a results page, a strong main image does an enormous amount of work. In a conversational answer, there may be no image at all, or a thumbnail at best. The shopper is reading — or hearing — a short description generated from your listing data.
So the things that decide whether you get recommended shift:
Structured attributes become load-bearing. Size, material, compatibility, wattage, capacity. A model answering "under $300 with a built-in grinder" filters on attributes. A missing field means you are not considered, regardless of bid.
Title clarity beats title stuffing. Keyword-crammed titles have always been ugly; in a conversational context they actively degrade the summary a model can produce from them.
Reviews become evidence, not decoration. A model summarising "is this good for beginners" draws on review content. Review substance starts mattering as much as star average.
Price and availability get read literally. No sale badge, no urgency framing. Just the number, compared against the constraint the shopper stated.
A+ content and imagery matter less here. Not worthless — they still convert once the shopper arrives — but they do not influence whether you get into the answer.
Key Takeaway: Conversational placements reward listing completeness over listing persuasion. The work that wins here — filling every attribute field, writing an honest title, keeping specs accurate — is unglamorous and has always been worth doing. It is now the difference between being considered and being invisible.
Why Is Measurement So Hard?
Here is where honesty is required rather than optimism.
You can see spend and you can see attributed sales. What is much harder to see is which placement produced them, because Prompts inventory draws from campaigns that also serve conventional placements. Unless reporting exposes placement-level breakdowns for conversational surfaces — and granularity here has been limited — you are looking at a blended number.
That creates a specific trap. Conversational placements plausibly convert well, because they reach shoppers at a moment of stated intent. If they do, blended campaign performance improves and you conclude your campaign strategy is working. If they underperform, blended performance dips and you go looking for a cause in the wrong place.
Either way you are attributing an effect to something you cannot isolate.
Honest limitation: We are deliberately not quoting CTR, conversion or share-of-placement figures for Prompts, because no credible public data exists yet and the placement is a few months old. Anyone presenting benchmarks for this surface is extrapolating. Treat early numbers — including your own — as directional until you have a full quarter and a clean baseline.
How Should You Approach It?
Cautious sequence: Baseline → Clean → Isolate → Watch → Judge.
Baseline before anything changes. Record 30 days of ACOS, conversion rate and total ACOS per campaign. Without it you cannot separate a Prompts effect from ordinary variance later.
Clean the listings that matter. Fill every attribute field on your top ASINs, fix titles that read as keyword strings, confirm specs are accurate. This is the highest-leverage work and it pays off across every surface, not just this one.
Isolate where you can. If placement controls or separate campaigns become available for conversational inventory, use them. A blended campaign teaches you nothing about a new placement.
Watch total ACOS, not just advertising ACOS. Conversational surfaces may shift organic sales to paid rather than adding incremental volume. Advertising ACOS can improve while total ACOS worsens — and only Seller Central has the denominator.
Judge on a quarter. New placements go through their own learning period, and the surface itself is still changing. Reading two weeks of data will mostly measure noise.
Step four is the one most sellers will skip and most need. It requires joining advertising data to Seller Central order data, which are separate systems with separate credentials — the structural problem that makes Amazon the hardest platform to measure honestly.
Read this next → Best MCP Servers for Amazon Sellers
Why advertising data alone cannot tell you whether an ASIN is profitable, which servers reach Seller Central as well as Amazon Ads, and the seven questions to ask before committing to one.
Where This Is Heading
Sponsored Prompts is not an isolated feature. It is Amazon's answer to the same question Google and OpenAI are answering with their commerce protocols: when a shopper delegates the search to a model, where do the ads go?
Amazon's answer is the most commercially direct of the three. Rather than opening the surface to third-party agents — it sued Perplexity to keep Comet off the site — it built the conversational layer itself, kept it inside the marketplace, and put ad inventory in it. Same shopper, same intent, same ad business, new interface.
For sellers that is arguably the least disruptive outcome available. The placement is new but the mechanics are familiar: your campaigns, your bids, your CPC. The thing to prepare is not a new ad strategy but listing data clean enough for a model to read.
What Should You Take from This?
Sponsored Prompts launched 25 March 2026, runs at standard CPC, and draws from existing campaigns. Rufus became Alexa for Shopping in May, making conversational the default surface.
Conversational answers offer two or three slots rather than a page of results, which concentrates both the upside and the cost of losing.
Listing completeness now decides consideration. Attributes, honest titles and review substance outrank imagery and A+ content for getting into an answer.
Placement-level measurement is limited, so baseline first and judge on a quarter. Watch total ACOS, not just advertising ACOS.
See Ads and Orders Together
Brandlio writes to Amazon Ads and reads Seller Central orders, sales and inventory, so total ACOS and per-ASIN profitability are one prompt in Claude or ChatGPT rather than two exports and a spreadsheet. See pricing or connect your first account.
Frequently Asked Questions
What are Amazon Sponsored Prompts?
Sponsored Prompts are Sponsored Products and Sponsored Brands placements that appear inside conversational shopping answers rather than on a search results page. They launched on 25 March 2026 and run at standard cost-per-click, drawing from your existing campaigns rather than requiring a separate campaign type.
What happened to Rufus?
Rufus, Amazon's shopping assistant, was folded into Alexa for Shopping in May 2026, which added shopping guidance, price history and automatic cart and routine purchasing. The practical effect is that conversational shopping became a default surface rather than a separate feature, which is what makes Prompts placements significant.
Do I need a new campaign type for Sponsored Prompts?
No. Prompts inventory draws from existing Sponsored Products and Sponsored Brands campaigns at standard CPC. That is convenient for setup but creates a measurement problem, since the same campaigns serve both conventional and conversational placements and reporting granularity for the latter has been limited.
How do I optimise a listing for conversational placements?
Prioritise completeness over persuasion. Fill every structured attribute field, since a model answering a constrained question filters on attributes and a missing field means you are not considered. Write titles that read clearly rather than as keyword strings, keep specs accurate, and treat review substance as evidence a model can summarise.
Do images still matter for Sponsored Prompts?
Less for getting into the answer, and still fully for converting once the shopper arrives. A conversational answer may show a thumbnail or no image at all, so imagery and A+ content have little influence on whether you are recommended — but they remain important on the product page itself.
Can I measure Sponsored Prompts performance separately?
Only partially. Because Prompts inventory serves from campaigns that also run conventional placements, and placement-level reporting for conversational surfaces has been limited, you are largely looking at blended figures. Baseline your campaigns before making changes, and treat early placement-specific conclusions with caution.
Why should I watch total ACOS rather than advertising ACOS?
Because conversational placements may shift sales you would have made organically into paid, rather than adding incremental volume. Advertising ACOS divides spend by attributed sales and can improve in that scenario, while total ACOS — spend divided by all sales including organic — worsens. Only Seller Central holds the total.
How does this relate to agentic commerce generally?
It is Amazon's version of the same shift Google and OpenAI are addressing with their commerce protocols: what happens to advertising when a shopper delegates search to a model. Amazon's approach is to build the conversational layer itself, keep it inside the marketplace, and place ad inventory in it — rather than opening the surface to third-party shopping agents.
