Amazon Optimisation
Prompt advertising is here. What Amazon’s Sponsored Product Prompts really change
With Sponsored Products prompts and Sponsored Brands prompts, Amazon has moved from “ads in Rufus” to ads made out of Rufus.
15 November 2025
8 min read
Amazon has just done something important at UnBoxed 2025.
Up to now, Rufus – Amazon’s AI shopping assistant – was a discovery layer that sat on top of search. It answered questions, compared products and quietly slotted Sponsored Products into the conversation in ways brands could not really see or control.
With Sponsored Products prompts and Sponsored Brands prompts, Amazon has moved from “ads in Rufus” to ads made out of Rufus.
This is the moment prompt advertising on Amazon really arrives.
What are Sponsored Product Prompts?
In Amazon’s own words, Sponsored Products prompts and Sponsored Brands prompts are a free beta that layers AI powered prompts onto your existing campaigns. The idea is simple. Amazon’s systems detect “meaningful decision moments” and inject short conversational prompts that help shoppers ask better questions. Those prompts then connect straight into your Sponsored Products or Sponsored Brands ads.
Under the hood, Rufus is doing the lifting. It has been trained on Amazon’s catalogue and information from across the web so it can answer natural language product questions, give buying tips and pull in relevant items.
So instead of someone typing “running sunglasses” and seeing a grid of ads, Rufus can lead with something like:
> “What are the best lightweight sunglasses for marathon training?”
And that question can now be effectively sponsored.
The flow becomes:
1. Rufus suggests or shapes the question
2. Rufus calls relevant Sponsored Products or Sponsored Brands
3. The shopper sees a friendly “answer” that is part help, part ad
Amazon positions this as making campaigns “more conversational” and helping shoppers at key decision moments.
From bidding on intent to manufacturing it
We have been here before in search.
Autocomplete has been nudging queries for years. Merchandising has always influenced what shoppers see. But prompt advertising tightens the loop.
* The assistant proposes the intent
* The platform controls the answers
* The brand pays to live inside that answer
It is still performance media, but the origin of the demand has shifted. You are no longer only bidding on what the shopper decided to ask. You are increasingly bidding on what the system encouraged them to ask.
That is a big change for retail media and for how we think about “intent”.
Amazon is very open about the upside. They say customers who engage with Rufus are 60 percent more likely to complete a purchase than those who do not, using a seven day attribution window to capture delayed conversions.
If your assistant can lift conversion by that much, it is only natural to start selling more structured access to that layer.
This is bigger than Amazon
The pattern is not unique to Amazon.
* Google is now showing ads in and around AI Overviews and has begun testing ads inside the AI generated snapshot itself for eligible Search, Shopping and Performance Max campaigns.
* Independent publishers in Europe are already challenging AI Overviews on antitrust grounds, arguing that AI summaries sit on top of their content and divert clicks while Google also monetises the space with ads.
Across platforms, you see the same move.
Helpful AI answers up front.
Paid placements blended into that help.
Very little visibility for everyone else.
Amazon’s Sponsored Product Prompts are simply the most explicit example on the commerce side so far.
What this means for brands
For brands and agencies, a few things change and a few things do not.
1. Treat prompts as a new shelf, not a gimmick
Prompts are a new surface on the AI shelf. They will not replace classic Sponsored Products any time soon, but they will sit on top of them and in some cases pull attention away from the traditional grid.
This is worth testing, not ignoring.
* Start with tightly defined use cases. One category, one hero product, one or two audiences.
* Watch not just ACOS but also incremental visibility in Rufus conversations. Right now Amazon groups Rufus ads under “Other” placements, but the effect will show up in blended performance.
2. Move from keywords to question clusters
Prompt advertising lives in full sentences, not in “exact match” terms.
Rufus thrives on questions like:
* “What should I look for in a vitamin D supplement?”
* “Which work boots are best for concrete floors all day?”
If you want to win that game you need:
* Product detail pages that read like answers to those questions
* A semantic understanding of the question clusters your buyers actually use
* Ad structures that mirror real life problems, not just product types
The brands that map their category questions and build creative around them will be the ones the model can confidently surface.
3. Design for AI shopping assistants, not just human skim readers
Rufus is trained on product catalogue data and information from across the web. That means your retail readiness and your off Amazon presence both feed the model.
Practically, that looks like:
* Clean titles, bullets and A plus that explain who the product is for and when to use it
* Consistent claims and benefits across Amazon, your site and other major surfaces
* Structured data and reviews that echo key use cases rather than random keyword stuffing
AI shopping assistants are incredibly literal. If your copy does not spell out the use case, they will often ignore you.
The LMO7 view: AI shelf space, now with prompts
At LMO7 we talk a lot about AI shelf space.
For years, retail media has been about buying visibility against demand that already exists. Prompt advertising is a step toward buying visibility inside the very questions that define that demand.
In practical terms, for a challenger brand this means three jobs:
1. Audit your AI presence
Where do you already show up in Rufus, ChatGPT, Gemini and other assistants when someone asks natural language questions in your category
2. Write for prompts, not just pages
Refresh your Amazon content and your wider web content so it can be pulled into AI answers. Think problem first, product second.
3. Test the new prompt shelves with discipline
Use Sponsored Product Prompts as a learning lab. Start small, compare against baseline performance and feed what you learn back into your wider AI search strategy.
Prompt advertising will not fix a weak offer or a bad product. What it will do is reward brands that do the thinking up front on how people actually shop in language.
Amazon has just given us a clearer view of where retail media is heading. The next question is simple.
When the assistant is the shelf, are you present in the conversation that starts the sale – or are you still only bidding on what is left over at the end of it?
Main source: Sponsored Products prompts and Sponsored Brands
Up to now, Rufus – Amazon’s AI shopping assistant – was a discovery layer that sat on top of search. It answered questions, compared products and quietly slotted Sponsored Products into the conversation in ways brands could not really see or control.
With Sponsored Products prompts and Sponsored Brands prompts, Amazon has moved from “ads in Rufus” to ads made out of Rufus.
This is the moment prompt advertising on Amazon really arrives.
What are Sponsored Product Prompts?
In Amazon’s own words, Sponsored Products prompts and Sponsored Brands prompts are a free beta that layers AI powered prompts onto your existing campaigns. The idea is simple. Amazon’s systems detect “meaningful decision moments” and inject short conversational prompts that help shoppers ask better questions. Those prompts then connect straight into your Sponsored Products or Sponsored Brands ads.
Under the hood, Rufus is doing the lifting. It has been trained on Amazon’s catalogue and information from across the web so it can answer natural language product questions, give buying tips and pull in relevant items.
So instead of someone typing “running sunglasses” and seeing a grid of ads, Rufus can lead with something like:
> “What are the best lightweight sunglasses for marathon training?”
And that question can now be effectively sponsored.
The flow becomes:
1. Rufus suggests or shapes the question
2. Rufus calls relevant Sponsored Products or Sponsored Brands
3. The shopper sees a friendly “answer” that is part help, part ad
Amazon positions this as making campaigns “more conversational” and helping shoppers at key decision moments.
From bidding on intent to manufacturing it
We have been here before in search.
Autocomplete has been nudging queries for years. Merchandising has always influenced what shoppers see. But prompt advertising tightens the loop.
* The assistant proposes the intent
* The platform controls the answers
* The brand pays to live inside that answer
It is still performance media, but the origin of the demand has shifted. You are no longer only bidding on what the shopper decided to ask. You are increasingly bidding on what the system encouraged them to ask.
That is a big change for retail media and for how we think about “intent”.
Amazon is very open about the upside. They say customers who engage with Rufus are 60 percent more likely to complete a purchase than those who do not, using a seven day attribution window to capture delayed conversions.
If your assistant can lift conversion by that much, it is only natural to start selling more structured access to that layer.
This is bigger than Amazon
The pattern is not unique to Amazon.
* Google is now showing ads in and around AI Overviews and has begun testing ads inside the AI generated snapshot itself for eligible Search, Shopping and Performance Max campaigns.
* Independent publishers in Europe are already challenging AI Overviews on antitrust grounds, arguing that AI summaries sit on top of their content and divert clicks while Google also monetises the space with ads.
Across platforms, you see the same move.
Helpful AI answers up front.
Paid placements blended into that help.
Very little visibility for everyone else.
Amazon’s Sponsored Product Prompts are simply the most explicit example on the commerce side so far.
What this means for brands
For brands and agencies, a few things change and a few things do not.
1. Treat prompts as a new shelf, not a gimmick
Prompts are a new surface on the AI shelf. They will not replace classic Sponsored Products any time soon, but they will sit on top of them and in some cases pull attention away from the traditional grid.
This is worth testing, not ignoring.
* Start with tightly defined use cases. One category, one hero product, one or two audiences.
* Watch not just ACOS but also incremental visibility in Rufus conversations. Right now Amazon groups Rufus ads under “Other” placements, but the effect will show up in blended performance.
2. Move from keywords to question clusters
Prompt advertising lives in full sentences, not in “exact match” terms.
Rufus thrives on questions like:
* “What should I look for in a vitamin D supplement?”
* “Which work boots are best for concrete floors all day?”
If you want to win that game you need:
* Product detail pages that read like answers to those questions
* A semantic understanding of the question clusters your buyers actually use
* Ad structures that mirror real life problems, not just product types
The brands that map their category questions and build creative around them will be the ones the model can confidently surface.
3. Design for AI shopping assistants, not just human skim readers
Rufus is trained on product catalogue data and information from across the web. That means your retail readiness and your off Amazon presence both feed the model.
Practically, that looks like:
* Clean titles, bullets and A plus that explain who the product is for and when to use it
* Consistent claims and benefits across Amazon, your site and other major surfaces
* Structured data and reviews that echo key use cases rather than random keyword stuffing
AI shopping assistants are incredibly literal. If your copy does not spell out the use case, they will often ignore you.
The LMO7 view: AI shelf space, now with prompts
At LMO7 we talk a lot about AI shelf space.
For years, retail media has been about buying visibility against demand that already exists. Prompt advertising is a step toward buying visibility inside the very questions that define that demand.
In practical terms, for a challenger brand this means three jobs:
1. Audit your AI presence
Where do you already show up in Rufus, ChatGPT, Gemini and other assistants when someone asks natural language questions in your category
2. Write for prompts, not just pages
Refresh your Amazon content and your wider web content so it can be pulled into AI answers. Think problem first, product second.
3. Test the new prompt shelves with discipline
Use Sponsored Product Prompts as a learning lab. Start small, compare against baseline performance and feed what you learn back into your wider AI search strategy.
Prompt advertising will not fix a weak offer or a bad product. What it will do is reward brands that do the thinking up front on how people actually shop in language.
Amazon has just given us a clearer view of where retail media is heading. The next question is simple.
When the assistant is the shelf, are you present in the conversation that starts the sale – or are you still only bidding on what is left over at the end of it?
Main source: Sponsored Products prompts and Sponsored Brands