How to get your products recommended by ChatGPT

How does ChatGPT find and choose products?

ChatGPT builds product recommendations from three pipelines, and you can influence all of them. This is the discovery layer of agentic commerce: before an agent can help a shopper compare, answer questions, or check out once they decide, a model decides which products it has read about, trusts, and can cite.

1. Crawling. OAI-SearchBot indexes your pages for ChatGPT search. Per OpenAI's bot documentation, sites that opt out of OAI-SearchBot are not shown in ChatGPT search answers.
2. Merchant feeds. Through the Agentic Commerce Protocol, you submit a structured product feed with required fields like item_id, price, and availability, refreshed daily. The feed spec accepts Google Shopping-compatible files.
3. Third-party corroboration. Reviews, editorial roundups, and community threads the model reads. Products that appear in both the feed and independent coverage do better than either alone.

Placement cannot be bought. Product results in ChatGPT are organic and unsponsored, ranked on relevance. When several merchants sell the same item, OpenAI weighs availability, price, quality signals, and whether you are the maker or primary seller.

Is ChatGPT shopping traffic worth chasing?

Two years of data settle this. In March 2025, AI referrals to US retail sites converted 38% worse than other traffic. A year later the channel converted better than non-AI traffic, with revenue per visit 37% higher, per Adobe. Volume grew just as fast:

The traffic converts because the model already did the legwork. By the time a shopper clicks through from an AI shopping agent, the reviews are read, the price is checked, and you are on the shortlist. The decision is still theirs; your job is to be the answer they land on.

What changed in 2026: merchants keep their own checkout

On September 29, 2025, OpenAI and Stripe launched Instant Checkout: US ChatGPT users could buy from US Etsy sellers inside the chat, with over a million Shopify merchants announced as coming, including Glossier, Vuori, Spanx, and SKIMS. The Agentic Commerce Protocol behind it is an open standard whose shared payment token design keeps raw card credentials away from ChatGPT.

In March 2026, OpenAI changed course: the first version of Instant Checkout lacked the flexibility it wanted, so merchants can now use their own checkout while OpenAI concentrates on product discovery. Walmart shipped a dedicated ChatGPT app with its own payments. Target, Sephora, Nordstrom, Best Buy, Home Depot, and Wayfair integrated ACP for discovery, and Shopify's Catalog now syndicates merchant data into ChatGPT automatically.

The takeaway: do not wait for checkout mechanics to settle. Every version of this experience starts with the model choosing which products to put in front of a shopper who still makes the call. Discoverability is the durable asset.

What makes a product citable by ChatGPT?

Citable means the model can read your product data, verify it against other sources, and quote it with confidence. Four things drive that:

• A complete, fresh feed. Required fields make products display correctly; recommended attributes like rich media, reviews, and performance signals (star_rating, popularity_score, return_rate) improve ranking and trust.
• Machine-readable product pages. Server-rendered content, schema.org Product markup with price, availability, and ratings, clean canonical URLs. JavaScript-only storefronts are the common failure mode; an agent-readable storefront fixes this at the root.
• Accuracy. Price or availability mismatches between feed and page kill eligibility. Being the brand or primary seller helps, and return policy data is required for checkout eligibility.
• Independent proof. Reviews and best-of editorial mentions are the off-site half of citability. The model cross-references what it can verify across sources.

One quiet detail: the is_eligible_search flag controls whether a product can surface in ChatGPT search at all. A feed can exist and still opt products out, and checkout eligibility requires search eligibility first.

The merchant checklist: make your products citable

Everything above compresses into eight steps. Most stores can clear the first five in a week.

• Allow OAI-SearchBot in robots.txt, and never block ChatGPT-User if you want agents opening your pages. GPTBot is a separate, training-data decision.
• Submit a product feed through OpenAI's merchant path (API or SFTP): a sample feed for validation, then daily snapshots. Shopify and Etsy merchants get platform-level syndication.
• Fill every required feed field, then the high-leverage recommended ones: variants, size systems, reviews, Q&A, return policy.
• Add schema.org Product markup and keep page copy specific: materials, dimensions, use cases, comparisons. Write the answer you want the model to quote.
• Keep price and availability synced between feed and page. Stale data is the fastest way to be dropped from a recommendation set.
• Earn third-party mentions: reviews on retail and community sites, category roundups, and forum presence where it is authentic.
• Verify crawler access in your server logs (look for OAI-SearchBot hits) and audit what AI engines can actually see of your store.
• ChatGPT search has historically drawn on Bing's index, so basic Bing Webmaster Tools hygiene remains cheap insurance.

How do you know if ChatGPT is sending you traffic?

OpenAI appends utm_source=chatgpt.com to links users click out of ChatGPT search results. In GA4, build a custom channel group or exploration matching session sources containing chatgpt, and read it in the session-scoped Traffic Acquisition reports.

Expect undercounting. Some visits arrive with stripped referrers and land in Direct or Unassigned, and pages fetched by ChatGPT-User on a shopper's behalf never fire client-side analytics at all. Server logs are the ground truth: OAI-SearchBot and ChatGPT-User hits tell you the machines are reading you before referral traffic shows up.

Benchmark quality, not just volume. Adobe's conversion and revenue-per-visit numbers above are your yardstick: if AI-referred sessions convert worse than the rest of your traffic, the model is recommending you for the wrong queries, or your pages lose shoppers the agent already qualified.