Topic · ChatGPT SEO 2026
The exact playbook: crawler access for OpenAI's three bots, JSON-LD schema, FAQ structure, llms.txt, brand-mention strategy. Plus a free citation test so you can see what ChatGPT currently says about your business.
Why this matters
Hundreds of millions of buyers use ChatGPT to vet vendors before calling. ChatGPT Search and browsing pulls live content from sites it can reach. If your business doesn't read cleanly to OpenAI's crawlers — or isn't allowed in at all — you are invisible to those buyers. The fixes are concrete and unglamorous.
From a client
Verified"We weren't cited at all. Two weeks after the fixes, we were the first source ChatGPT named for our category."
1 · Crawler access
| Crawler | Used for | Recommended action |
|---|---|---|
GPTBot | Training-data collection for future GPT models | Allow (unless you have a specific reason not to be in the training data) |
ChatGPT-User | On-demand fetches when ChatGPT browses the web for a user's query | Allow (required for citation in browsing answers) |
OAI-SearchBot | Index for ChatGPT Search | Allow (required for ChatGPT Search citation) |
Add explicit User-agent: GPTBot / Allow: / blocks in your robots.txt. Many sites unintentionally block these bots with overly broad rules.
The 9-step playbook
1
GPTBot, ChatGPT-User, OAI-SearchBot — explicit allow in robots.txt.
2
Organization, LocalBusiness, Service, Product, FAQPage, Article — whichever fit each page.
3
Real buyer questions, definitive answers. Both visible HTML and FAQPage schema.
4
"X vs Y" tables with concrete figures get quoted verbatim.
5
Plain-text summary at /llms.txt describing what the site is and listing key pages.
6
Real <h1>, <article>, <table>. Avoid layouts that require JS to read.
7
datePublished, dateModified, named author, named publisher. ChatGPT prefers recent and attributed.
8
Industry publications, established directories, podcast appearances with transcripts. ChatGPT's "real entity" sense is partly trained on third-party breadth.
9
Specific numbers, dated claims, definitive statements. "$50/hr" beats "affordable." "5 business days" beats "fast."
FAQ
Allow GPTBot, ChatGPT-User, OAI-SearchBot in robots.txt. Add complete JSON-LD. Structure pages as Q&A blocks. Publish an llms.txt. Build third-party brand mentions on trusted sites. Use precise, citable factual phrasing.
Three: GPTBot (training collection), ChatGPT-User (in-conversation web fetches), and OAI-SearchBot (ChatGPT Search index). All three should be explicitly allowed.
Yes — via live browsing (ChatGPT-User, OAI-SearchBot) and via prior training (GPTBot). Both paths require explicit crawler access.
Search/browsing reflects changes in days to weeks. Training-rooted citations update with model releases — months.
FAQ blocks, comparison tables with numbers, clearly attributed statements, dated articles, complete JSON-LD. Marketing fluff gets paraphrased or skipped.
Yes — possibly more so than for SEO. ChatGPT's "real entity" recognition is partly trained on third-party mentions across credible sites.
See what ChatGPT says about you
We ask ChatGPT (GPT-5 and GPT-4o) 20 buyer-intent questions about your category and report verbatim where you appear — or don't. Plus a prioritized fix list.