ChatGPT search is OpenAI’s web-connected answering mode in ChatGPT, which retrieves current web content, composes an answer and links the sources it used. This is a platform reference page: it records how the surface works, how content becomes eligible to appear in it, and the controls publishers have, drawn from OpenAI’s own documentation. Facts on this page are verified against that documentation and re-checked at the review date shown below.
In one sentence
ChatGPT search is the version of ChatGPT that looks things up on the live web and cites what it used, which makes it an answer surface a site can be present in or absent from.
How ChatGPT search retrieves and cites
OpenAI operates three distinct crawlers with independent controls, and conflating them causes most publisher mistakes. OAI-SearchBot crawls and indexes content to power search features; allowing it is the eligibility control for appearing in ChatGPT search results. GPTBot collects content that may be used for model training; blocking it is a training decision and does not remove a site from search. ChatGPT-User fetches pages when a user directly asks about them. Answers in search mode present linked citations to the pages used. Outbound clicks carry a utm_source=chatgpt.com parameter, so referral traffic is measurable in standard analytics.
What determines inclusion
OpenAI states there is no way to guarantee placement, and that inclusion requires allowing OAI-SearchBot in robots.txt and ensuring the site’s host or CDN does not block OpenAI’s published IP ranges, which are listed in machine-readable JSON files. Changes to robots.txt take up to about a day to be reflected in search eligibility. Beyond eligibility, the working levers are the same ones that govern citation probability generally: indexable, server-rendered pages carrying self-contained answers.
Publisher controls, in brief
Allow OAI-SearchBot to be eligible for search citations. Allow or block GPTBot separately as a training choice. Check CDN and firewall rules against OpenAI’s published IP lists, since a WAF block above robots.txt silently removes eligibility. Monitor server logs for all three user agents to confirm actual crawl behaviour rather than assumed behaviour.
Why this surface matters
ChatGPT is the largest assistant audience asking buying questions in natural language, and search mode is where those answers connect to live pages and send measurable traffic. For most brands it is the single highest-value engine to measure first, and the one where the gap between being mentioned from training data and being cited from the live index is most visible across repeated runs.
Related concepts
References
- OpenAI Help Center, ChatGPT Search: help.openai.com
- OpenAI Help Center, Publishers and Developers FAQ: help.openai.com
Author: Harpal Singh · Last verified: 7 August 2026