ChatGPT search

ChatGPT search is ChatGPT answering with fresh web results. ChatGPT decides when to search, or you select search. Answers carry inline citations and a Sources panel.

Results come from search partners, Bing among them, and from OpenAI’s own index crawled by OAI-SearchBot. As of 2026, the crawler does not execute JavaScript. Shopping answers with product cards arrived in 2025.

Three tokens control access in robots.txt: OAI-SearchBot for search indexing, GPTBot for model training, and ChatGPT-User for on-demand fetches when a user opens your link from ChatGPT.

Your page can be cited even if you do not rank first on Google. A clear answer can earn a citation and send users to your page from the Sources panel.

It rewards crawlable HTML. If your product docs or /pricing render with JavaScript only, OAI-SearchBot will miss key facts.

Product pages can appear as cards. Make price, availability and variant details visible in HTML. Do not hide core content behind tabs that need JavaScript to render.

1. 1

   Decide your policy

   Allow OAI-SearchBot if you want to appear in ChatGPT search. Allow ChatGPT-User so users can fetch your page when they click a citation. Set GPTBot by your training policy.

2. 2

   Set robots.txt

   Place user-agent rules at /robots.txt. Test that they load for both HTTP and HTTPS.

3. 3

   Serve crawlable HTML

   Render key text server side for docs, pricing and FAQs. Keep titles, headings and key facts in the markup.

4. 4

   Mark up sources clearly

   Use precise page titles like “Pricing, YourProduct”. Write concise intros. Put facts near the top, for example “Starts at $X” at the start of /pricing.

5. 5

   Keep canonical and sitemaps clean

   Avoid duplicates. Link important pages in your nav. Submit XML sitemaps so partner engines like Bing can find them.

# Allow ChatGPT search and user fetches, block training
User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: GPTBot
Disallow: /

Track crawling first. Watch server logs for OAI-SearchBot hits on /pricing, /guides/getting-started and similar. If it cannot fetch, you will not be cited.

Track on-demand fetches. Log the ChatGPT-User agent. Spikes here often match new citations after you update a page title or a key fact.

Watch referral traffic patterns. Users open your site from the Sources panel. As of 2026, referrers vary, so use on-page analytics events on key actions to see lift after citations.

Use partner signals. Since Bing is a partner, monitor Bing Webmaster Tools for indexing and coverage on pages you want ChatGPT to cite.

• Blocking OAI-SearchBot by copying a GPTBot ban without thinking about search
• Shipping content that only appears after client-side rendering or user action
• Thin product pages with no plain-text specs, price or availability
• Using vague titles like “Solutions” that cite poorly
• Letting canonical or hreflang point away from the page you want cited