How Perplexity shopping works for brands

Perplexity's shopping feature is easy to describe and hard to influence, because its own documentation is short and its search pulls from far more pages than most assistants show. This post separates what Perplexity states, dated as of October 2026, from what our October 2026 sample of its API shows. The two are not the same thing: the product experience lives in the app, and our numbers come from an API that returns text and a list of sources.

What does shopping in Perplexity include?

Perplexity launched its current shopping experience on November 25, 2025, describing it as free for all U.S. users. Its launch post lists conversational discovery that remembers past searches, product cards with specifications and reviews, and checkout through PayPal, with the retailer remaining the merchant of record. Its help center adds the mechanics of Instant Buy:

• A shopping question returns product cards inside the answer.
• Selecting Instant Buy saves the shopper's checkout details once, and the order and payment go to the merchant for fulfillment.
• Instant Buy works only with merchants compatible with Perplexity's agentic checkout, and the help page says the supported categories are those expected to be most popular with shoppers.
• Perplexity's help page says the service is available to users in the United States.

An earlier version existed. In November 2024 Perplexity launched Buy with Pro for U.S. Pro subscribers. The current help center and launch post we read do not mention it, so we do not know whether it still exists under that name.

How does Perplexity say it picks products?

The help center gives three statements worth keeping:

1. Ranking. Perplexity ranks product listings using criteria similar to those it uses to rank sources in its answers: authority and relevance. A more specific question lets it match products more precisely.
2. Personalization. It also uses a shopper's past searches, interactions and saved preferences to refine recommendations over time.
3. Merchant data. Its answer engine is more likely to recommend merchants that provide deeper product details such as availability, reviews, pricing and specifications.

On paid placement, Perplexity's help page says its algorithm determines all product listings based on relevance and ratings, and that advertisers cannot pay to list products in its related products section.

Nothing there gives weights. The third point is the one a merchant can act on: availability, reviews, price and specifications are all things your pages and feeds either state clearly or do not.

How can a merchant get products in front of Perplexity?

The last row is a statement about what those pages say, not about what is technically possible. If you sell on Shopify, do not assume that a setting in your admin reaches Perplexity.

Perplexity also documents a second bot, Perplexity-User, which supports user actions and which Perplexity says generally ignores robots.txt rules because a person requested the fetch. The crawler explainer compares it with the other AI bots. Perplexity's page also says that if a firewall protects your site you may need to allow its bots, and gives a Cloudflare rule that combines user agent and IP address conditions.

What does our Perplexity data show?

Our October 2026 samples include 787 answers from Perplexity's Agent API with the Sonar model (perplexity/sonar) and its web search tool turned on, across 160 categories in the United States, Mexico and Canada. For Perplexity, the source list we record is what its search retrieved, not what the answer used: the API does not say which pages the answer relies on, so we count the whole list. We report each engine on its own, and OpenAI's numbers are beside Perplexity's only for comparison.

The two columns measure different things, linked pages on one side and retrieved pages on the other, so the gap is partly a recording difference. Read each column on its own. The retailer list is hand-made, and the own-site rule is strict (a domain whose first label equals the brand's name), so it misses country sites and parent companies.

The most frequent domains in Perplexity's retrieved lists look like this:

By our reading, three of those ten are retailers (amazon.com, amazon.ca and target.com) and the other seven are publishers or review sites. A domain in a retrieved list does not show that it shaped the answer. Counts for individual domains are per engine and are also in the websites in AI shopping answers and the Wirecutter post.

On brands: excluding the alternative-to-a-brand question, a Perplexity answer recommended 2.5 brands on average (median 2) across 631 answers, and 1.9% recommended none. The shortlist is short. Whether your brand is on that shortlist is a separate question from whether your site is in the source list. And a brand recommended by one engine is often not recommended by the other. Engine overlap, defined as the brands recommended by both engines divided by the brands recommended by either, per category and averaged over 160 categories, was 47.0% excluding the alternative question. It is not answer-by-answer agreement. The OpenAI versus Perplexity post has the detail.

What can a brand take from this?

• Retail listings are in the picture. A retailer was in the retrieved list of 76.7% of Perplexity answers. If you sell through Amazon, Target or others, the listing's title, price, stock and images are among the pages the search can retrieve. We have not measured what they change.
• Editorial pages are in the picture too. Seven of the ten most frequent domains are publishers or review sites. That fits Perplexity's own wording about authority, and it does not prove that those pages decide which products are named.
• Perplexity says details help. Availability, reviews, price and specifications are the items it names. Put them in plain text on product pages, and keep them identical in feeds (see product feeds for AI shopping).
• A recommended brand's own site is often in the list. In 62.0% of the 787 Perplexity answers (all prompts), the retrieved list held the own site of at least one recommended brand, by our strict domain rule (a domain whose first label equals the brand's name). That makes crawl access and readable pages basic hygiene. It does not show that a given brand's site shaped the answer.
• Measure Perplexity apart from other engines. Different recording, different shortlists.

What can the API tell you about the app?

Less than it seems. Our requests carry no account, search history or memory, and the app can personalize product cards using exactly those. The app shows product cards and Instant Buy buttons that our samples do not record. Our October numbers come from Perplexity's Agent API with the Sonar model. If a tool you use describes itself as running on "Perplexity's Sonar API", ask which endpoint it calls: Perplexity's documentation says Sonar Chat Completions support ended on September 27, 2026.

Perplexity's documentation also says that whether an answer carries source markers depends on the prompt. Our prompts are plain shopper questions, which is another reason we treat the list as retrieved, not used. The API versus app post lays out the limits in full.

How can you check your products in Perplexity?

1. Ask the five question types. A best-overall question, a budget question with a ceiling, a premium question, a specific use, and a buying-advice question without brand names.
2. Open the product cards and the sources. Note which products appear, which merchants, whether an Instant Buy button shows, and which pages sit in the source list.
3. Compare with your data. Check price, stock and specifications against your pages and your feeds. Report mismatches to the merchant systems you control first, since Perplexity's support page says it cannot help with product-specific questions.
4. Repeat on different days and keep a log. Answers vary. The manual measurement method explains how many repeats are needed.
5. Count the sources per type. For each answer, mark whether the list holds a retailer, a review site and your own site, and divide by the number of answers.

What is still unknown?

• Weights. Perplexity names ranking criteria and gives no weights.
• Who is in the partner data. Instant Buy names platforms and a few retailers, and no complete list.
• The Merchant Program's current status. We have the 2024 announcement and nothing current.
• Markets. Perplexity's help pages we read describe the U.S. Our index also samples Mexico and Canada through the API, and says nothing about the app there.

Frequently asked questions

Does a page in Perplexity's source list mean the answer used it?

No. Perplexity's API returns the pages its search retrieved and does not mark which ones the answer relies on. Our counts are retrieval counts. A domain that appears often, such as a retailer or a review site, is retrieved often, and we cannot say how much it influenced the brands named.

Is shopping in Perplexity available outside the United States?

Perplexity's launch post and Instant Buy page describe U.S. availability. We found no page that describes the app's shopping features for other countries, so we do not state one.