Retail listings in AI shopping answers: Amazon, Target, Best Buy and Rufus

In the October 2026 edition of our sample, a retailer domain was in the source list of 76.7% of Perplexity answers and 17.6% of OpenAI answers, but the engines record sources differently, so the two numbers measure different things. The first half of this post gives both. The second half covers the retailers' own AI assistants, which we do not measure, and the listing work that follows from what is known.

How often do retailer pages show up in the sources?

The October 2026 edition of the index gathered 800 answers by calling OpenAI's API with its web search tool on (model gpt-6-luna), plus 787 through Perplexity's Agent API with the Sonar model (perplexity/sonar), spread over 160 categories in the US, Mexico and Canada. For each answer we recorded the source list. For OpenAI the list holds the pages its answer links; for Perplexity it holds the pages its search retrieved, which the answer may or may not use. We then checked each list against a hand-made list of 49 retailer domains.

The gap is partly a recording effect. In that edition, OpenAI's answers linked a mean of 2.2 pages from 2.0 distinct domains. Perplexity's retrieved lists held a mean of 27.3 URLs from 15.2 distinct domains. A longer list has more chances to contain a retailer, so the two percentages should not be read as one engine "using" retailers four times as much as the other. Read each against its own engine.

Which retailers, on each engine?

The tables use the 30 most frequent domains of each engine's list and keep only the ones on the retailer list. Each engine's count is on its own answers: for OpenAI, the 797 answers that link at least one source, and for Perplexity, all 787.

Perplexity, domains in retrieved source lists (787 answers)

OpenAI, domains in linked pages (797 answers with a link)

Four things to keep in mind when reading them.

• The lists span three markets. Amazon's Canadian and Mexican domains, Best Buy Canada and the Mexican retailer domains appear because the Canada and Mexico runs are inside each engine's total. We cannot split the engine lists by market here.
• Amazon does not appear in OpenAI's top 30. Neither amazon.com nor bestbuy.com is among the 30 most frequent domains OpenAI linked, the last of which appeared in 7 answers, so each was linked in at most 7. That is a statement about this sample's links, not about how OpenAI treats Amazon.
• The retailer list is hand-made. It counts REI and IKEA as retailers, though either could also be a brand an answer recommends. It holds only the domains we listed, so a store we missed is not counted, and a listed retailer counts wherever it appears, not only when it sells the recommended product.
• Presence is not influence. A domain in a source list does not show that it shaped the recommendation. Perplexity's list is what its search retrieved, not what its answer relied on.

What is Rufus, and what other retail assistants exist?

Retailers also run assistants inside their own apps and sites. Our two engines do not include them, so Dopbot does not measure any of them. These are the facts we could verify from the retailers' own pages as of October 2026. The Walmart and Target announcements date from 2025, and we did not check whether those products have changed since.

• Amazon. Rufus was Amazon's shopping assistant. Amazon says that on May 13, 2026 Rufus was renamed Alexa for Shopping. Amazon's page about Rufus says that Rufus generates answers from listing details, customer reviews and community Q&As. An Amazon seller guidance article says AI-powered search tools can use your listing content to recommend products. That article is dated April 29, 2026, which we read on October 28, 2026; it uses the name Rufus and does not mention Alexa for Shopping. Search for both names when you look for Amazon's guidance.
• Walmart. Walmart introduced Sparky, a shopping assistant, in June 2025. Its launch announcement says Sparky helps customers search for items and synthesizes reviews, and that customers could reach it through an "Ask Sparky" button in the Walmart app.
• Target. Target introduced a conversational Gift Finder in November 2025, where a shopper types details about the recipient or occasion.

As of October 2026 we did not find a Best Buy shopping assistant described on Best Buy's own pages, so we leave it out; that is not a statement that none exists. None of these pages say how much shoppers use the assistants or how they rank products, so we say nothing about that either.

What listing work follows from this?

No measurement here shows that a better listing leads to more recommendations. The steps below rest on two verified facts: Amazon says Rufus builds answers from listing details, and in the October 2026 edition Perplexity retrieved retailer pages in 76.7% of its answers. If an assistant reads a retailer page, errors and gaps on that page can travel into an answer.

1. Make the facts match everywhere. Compare model name, size, materials, compatibility, price and availability on each retailer page against your own site. A mismatch can leave an assistant with conflicting facts to choose between.
2. Write the description as a buying guide. Amazon's seller guidance suggests saying who the product is for and what problem it solves, rather than listing features alone. That is also the shape of several prompts in our October 2026 edition, such as the specific-use one.
3. Use original text. Amazon's guidance recommends original descriptions rather than copied manufacturer content.
4. Answer the questions shoppers ask. Amazon's page about Rufus names reviews and community Q&As as inputs. Respond to questions on the listing, within the retailer's rules.
5. Keep price and stock current. The price and stock errors post covers how a wrong fact gets into an answer and how to confirm a fix.

How can you check your own retail presence?

1. Pick four shopper questions for your product type, none naming a brand: best overall, a budget with a price ceiling, premium, and one specific use.
2. Ask each on each engine you can reach, in clean sessions, three times. On two engines that is 24 answers, so expect about an hour. Record the source list of every answer.
3. Mark every retailer domain, and note whether your own product page is the URL.
4. Open your product page on each retailer that appears and compare it with your own site, using the five steps above.
5. Repeat after any change and count answers, not impressions. The manual measurement method gives a spreadsheet layout.

Each category page in the index has a sources table with one column per engine. The inventory of non-retail sites is in the websites post, and how the figures are counted is in the methodology.

Frequently asked questions

Do I need to sell on Amazon to be recommended by an AI assistant?

Our data cannot say. We did not record where each recommended brand is sold. We can say that amazon.com was in the retrieved sources of 265 of 787 Perplexity answers and was not among OpenAI's 30 most frequent linked domains.

Which Amazon domains did Perplexity retrieve most often?

In the October 2026 edition, amazon.com was in the retrieved sources of 265 of 787 Perplexity answers, amazon.ca in 123 and amazon.com.mx in 91. The three domains come from the US, Canada and Mexico runs combined in the engine total, and these are counts of answers, not a ranking of retailers.