ChatGPT search lets ChatGPT look up current information on the web and answer with inline citations and a list of sources. OpenAI previewed it as SearchGPT in 2024, built it into ChatGPT later that year and has since opened it to free and logged-out users. ChatGPT decides on its own when a question needs a search, and users can also trigger one directly. For many buyers it is now a routine research tool, which makes it one of the main surfaces for AI visibility.
How ChatGPT search finds pages
OpenAI documents a dedicated crawler, OAI-SearchBot, used to surface websites in ChatGPT's search features, and says sites that block it may not appear in search answers. OpenAI has also said it works with third-party search providers. The exact mix of sources, ranking signals and index coverage is not published.
When ChatGPT fetches a page because a user asked it to, it may use a separate user agent, ChatGPT-User, which OpenAI treats differently from its crawlers.
- OAI-SearchBot: crawls to surface sites in search results
- ChatGPT-User: fetches pages for a specific user request
- GPTBot: collects data for model training, separate from search
What to check
Confirm OAI-SearchBot is allowed in robots.txt and not blocked by a CDN or firewall rule. Make sure key pages render their main content in HTML. Then test the prompts your buyers use, noting which pages ChatGPT cites and whether its description of your product is accurate.
ChatGPT links commonly carry a utm_source=chatgpt.com parameter, and visits also arrive with chatgpt.com as referrer in many cases, which helps separate this traffic in analytics. Some visits will still show up as direct.
Example
Cited for comparisons, missing for pricing
Tidewell, a fictional status page tool, finds ChatGPT cites its comparison pages often but describes its pricing from a third-party review that is two years old. Tidewell's own pricing loads through a script after page load. Rendering the plan table in the initial HTML and adding a plain-text summary gives ChatGPT's search a current source to use.
Common mistakes
Blocking GPTBot and assuming search is blocked too
Training and search crawlers are separate. Decide on each one deliberately.
Testing while logged in with memory on
Personal memory and history can change answers. Test in a clean session as well.
Questions
There is no submission form for ordinary sites. Allow OAI-SearchBot, keep pages crawlable and indexable, and make sure your content is linked from the wider web.
OpenAI has said it uses third-party search providers alongside its own crawling, and Bing has been widely reported as one of them. The details are not fully documented.
