How to rank your products in ChatGPT

What "ranking" in ChatGPT means — citation, not position

For a shopping query, ChatGPT cites one to three products, often with inline source links, as of mid-2026. The metric is close to binary: cited or not, with secondary measures for consistency across query variations and which competitors appear alongside. A brand that holds a citation for "best heavyweight tee under $50" across a dozen phrasings is winning; one that appears for none is invisible regardless of its Google position. The general framework behind this — surfaced, cited, recommended — is in the product AI visibility pillar; this guide is the ChatGPT-specific sequence.

Step 1 — Confirm ChatGPT is crawling your PDPs

ChatGPT's crawlers identify as GPTBot (the general crawler) and ChatGPT-User (on-demand browsing from inside a chat), with OAI-SearchBot serving search. Verify all can reach your PDPs: robots.txt does not block them, the development-store password is off the production domain, and Cloudflare or similar is not firewalling OpenAI's published ranges — documented at platform.openai.com/docs/bots. Merchants block these by accident when tightening bot protection more often than by policy. Zero crawler hits four weeks after publishing a product means something is blocking.

Step 2 — Fill out Product JSON-LD completely

The fields that carry weight: name, description, sku, gtin13, mpn, brand, a full offers block (price, currency, availability, priceValidUntil), the image array, and additionalProperty for material, color, and size — aggregateRating only when real. A complete block tells ChatGPT the PDP is a legitimate product entity worth citing. The field-by-field walkthrough with a copyable example is schema for AI search; the schema generator drafts a block from any URL. Re-audit monthly — theme updates break emission silently.

ChatGPT cites pages it can resolve to entities. Incomplete schema is an entity you never finished introducing.

Step 3 — Publish and maintain an llms.txt

A well-formed /llms.txt gives ChatGPT's crawlers a curated map: homepage, top collections, flagship PDPs, policy pages, each with a one-line description. Crawler logs show GPTBot fetching the file where published, though no engine documents the behavior formally. Shopify hosting requires a workaround — the full walkthrough is llms.txt for Shopify, and the generator produces a draft in under a minute.

Step 4 — Seed trusted review-source coverage

ChatGPT's shopping citations lean on third-party sources it already trusts. Self-published content rarely gets cited alone; it gets cited alongside review media, Reddit threads, and teardown videos that corroborate the claims. The practical cadence: one category review-site pitch per quarter, one detailed teardown video per flagship SKU, a few disclosed community answers per month. eCommerce Insights audits in early 2026 found brands with ten or more third-party review-media mentions per flagship SKU cited in ChatGPT at roughly three times the rate of brands with none (illustrative sample; causation imperfect — strong brands tend to have both). Slow, compounding work.

Step 5 — Clarify your brand entity on Wikidata

Engines resolve brands to known entities, and Wikidata is the easiest public knowledge base to influence: founding date, founders, product categories, canonical URL, press references. A one-time hour that pays across every engine, not just ChatGPT. A brand that exists only on its own domain is harder to cite with confidence.

Step 6 — Rewrite PDP intros for passage-level citability

ChatGPT quotes short, factual passages. "Meet the tee you'll live in" gets ignored; "The Signature Heavyweight Tee is an 8oz organic-cotton unisex tee, boxy fit, in bone, slate, olive, and black, sizes XS–XXL, $48, made in Portland with GOTS-certified cotton" gets quoted. Rewrite the first 40–80 words of every PDP this way and let marketing prose follow — the citable passage and the brand voice can coexist on one page. Full per-field patterns in optimize content for AI search; for the rewrite-at-scale workflow with human-approved diffs, see PDP optimization.

Step 7 — Monitor by prompt, weekly

Build a battery of 20–50 intent queries per product line — phrases real shoppers type, qualifiers included. Run them weekly. Track which SKUs are cited, which are missing, which competitors appear, and the deltas. The free ChatGPT product visibility checker runs a single product on demand; ChatGPT tracking in eCommerce Insights automates the full battery and alerts on lost citations.

The honest timeline

ChatGPT Shopping behavior shifts with OpenAI releases; treat every count here as observed mid-2026 behavior, re-measured weekly rather than assumed.