How to get your Shopify products recommended by ChatGPT and Perplexity

OpenAI's help page on shopping says ChatGPT picks product results itself. OpenAI says those results are not ads. Its partnerships do not influence them either.

Open a product in ChatGPT and it lists the stores that sell it. That list comes from product and merchant data. The data arrives from third-party providers or from merchants directly.

OpenAI names the factors that rank those stores. They are availability, price, quality, and whether the store makes or mainly sells the item.

Products reach ChatGPT by two routes. OAI-SearchBot crawls public pages. A product feed sends the catalogue straight to OpenAI. OpenAI's merchant page says a feed is optional if ChatGPT already crawls your site. A feed gives you more say in how products appear.

For Shopify stores, OpenAI says product data already reaches ChatGPT through Shopify Catalog. It says individual merchants need no extra work. Shopify's help centre adds conditions. The store must sell to US customers. Products must be eligible for Shopify Catalog. You must accept the Agentic Storefronts supplemental terms.

Shopify opened this to millions of merchants on 24 March 2026. That week OpenAI stepped back from Instant Checkout. Shoppers now finish the purchase on your own checkout. As of October 2026, OpenAI's merchant page says shopping is live for US users. More countries are planned.

How Perplexity picks the products it shows

Perplexity launched its shopping features on 18 November 2024. It also opened a free Merchant Program. Merchants in it share product data with Perplexity.

Its help centre says merchants with richer details get recommended more often. It names availability, reviews, pricing and specifications. It says an algorithm decides every listing, on relevance and ratings. Perplexity says listings are not sponsored.

Perplexity's crawler is PerplexityBot. Its docs say the bot surfaces and links websites in Perplexity search results. Perplexity says it is not used to train foundation models. It asks sites to allow the bot in robots.txt and to allow its published IP ranges.

Instant Buy lets US shoppers check out inside Perplexity, through PayPal. The retailer stays the merchant of record. It works only with merchants that support Perplexity's agentic checkout.

For what these answers look like in practice, read our 12 conversation teardowns. They show four assistants choosing a product turn by turn.

The checklist: eight things to check

Let the crawlers in

OpenAI's crawler docs are direct about this. Sites that opt out of OAI-SearchBot are not shown in ChatGPT search answers. They can still appear as plain navigation links. A robots.txt change takes about 24 hours to reach OpenAI's systems.

OAI-SearchBot is separate from GPTBot, OpenAI's training crawler. You can allow search and block training. OpenAI's advertiser guidance adds a detail for feeds. Let OAI-SearchBot fetch your product image URLs too.

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

On Shopify, you edit robots.txt through a theme template called robots.txt.liquid. Shopify warns that a wrong edit can lose all your search traffic. Save a copy of the current file before you change it. Our AI visibility check reads your robots.txt for you.

Mark up the product

Structured data is code that states facts about a page. For products, schema.org defines Product, Offer, AggregateRating and Review.

Google's docs say Product markup can make a page eligible for merchant listings. Those need an Offer. Product snippets need one of review, aggregateRating or offers.

Google also says its own AI features need no special markup. The markup still puts price, stock and rating in one labelled place. A crawler can read them without guessing from your layout.

Keep the markup in step with the page. The rating in the code should match the rating a shopper sees. Test a page with our review schema validator. Our guide to AggregateRating schema covers the rules.

Write titles and descriptions that answer questions

Shoppers ask assistants in full sentences. "A linen shirt under $80 that doesn't crease much." "Running shoes for wide feet."

Perplexity says fuller specifications help a product get recommended. So put the facts people ask about in plain text. Cover material, weight, fit, size range, care and what it is for.

"Effortless summer essential" tells a crawler nothing it can match. "100% linen, relaxed fit, XS to 3XL, machine washable" answers four questions. Our PDP audit scores a product page for AI search.

Make reviews and customer photos readable

Perplexity names reviews as a detail that raises a recommendation. It ranks listings on relevance and ratings. OpenAI says shopping research reviews quality sources across the internet.

Reviews only count if a crawler can read them. Some review widgets load after the page, through JavaScript. Open the page source and search for a line from a review. If it is missing, a crawler may miss it too.

Customer photos show fit, colour and scale in real light. Tag each one to the product it shows. A tagged photo ties that evidence to the right item.

Answer the questions on the product page

Shoppers ask an assistant what they would ask a person in a shop. Does it run small? Is it see-through? Can it go in the dryer?

Answer those on the product page, in text. Take the questions from your support inbox and your reviews. Keep each answer to a sentence or two.

Get mentioned on the pages assistants read

OpenAI says shopping research searches across the internet and reviews quality sources. Perplexity says it ranks products much as it ranks sources, on authority and relevance.

So pages about you count as well as your own. That means review sites, gift guides, comparison posts and forum threads. Ask your category question in each assistant. Write down which sites it cites. Those are the pages to be on.

Send samples to the people who write those lists. Ask happy customers to review you on the sites the assistant already cites. Our AEO brand report checks how citable your brand is.

Publish an llms.txt, last

llms.txt is a text file at the root of a site. It lists the pages an AI should read first. Jeremy Howard of Answer.AI proposed it in September 2024.

It is still a proposal. Google says its search ignores llms.txt and similar AI files. We found no statement from OpenAI or Perplexity that their assistants read it.

It is cheap to publish and does no harm. Do it after the seven fixes above. Our llms.txt guide explains the format.

Where Idukki fits

We make Idukki. It collects reviews and customer photos and videos, and tags them to products. On Shopify it adds Product, Review and AggregateRating JSON-LD to product pages. Each store also gets an agent feed. It serves products, UGC and reviews as JSON, JSON-LD and an llms.txt. Idukki does not send product feeds to OpenAI or Perplexity for you.

How to test whether you show up

Answers change between runs and between people. OpenAI says ChatGPT considers the shopper's context, including memory and custom instructions. So look for patterns across several runs.

When an agent tries to buy, other things break. We let one buy from our test store in We let Perplexity buy from our store. For try-on in ChatGPT, read ChatGPT can try on your products now.