Answer Engine Optimization · ChatGPT
When ChatGPT searches the web, it can cite the pages it used. This guide covers how those pages are found, what to check on your product pages, and how to measure citations in a way you can trust.
Last updated September 14, 2026 · ~7 min read
Platform facts on this page were checked against OpenAI's documentation on September 14, 2026. These features change often, so confirm current details in the linked documentation.
A citation is a web page ChatGPT links as a source when it searches the web. A product result is a product card with images, price and merchant links, built from structured product data. This page is about citations. For product results, see the ChatGPT shopping product data checklist.
OpenAI documents several crawlers, and they do different jobs. OpenAI crawler documentation
| Crawler | What it is for | What blocking it means |
|---|---|---|
| OAI-SearchBot | Surfacing websites in ChatGPT search | Your pages are kept out of search answers, though they may still appear as navigational links |
| GPTBot | Collecting data for model training | Opts out of training only; separate from search |
| ChatGPT-User | Fetching a page when a user asks ChatGPT to | OpenAI notes robots.txt rules may not apply; not used to decide search inclusion |
OpenAI says robots.txt changes for OAI-SearchBot take about 24 hours to apply. When ChatGPT answers without searching, it draws on what the model learned in training, which you cannot update directly.
With small question sets, treat changes as signals to investigate rather than proof of cause and effect.
Hasmord's citation checks ask AI models realistic buyer questions through their APIs — up to four models, depending on your plan — and save the answer text with whether it mentioned your brand, model or URL. The product data audit and evidence checks help with the page side. API answers can differ from the ChatGPT app, so Hasmord reports them as observed.
When ChatGPT searches the web, it uses pages that OpenAI's search crawler, OAI-SearchBot, can access. Blocking OAI-SearchBot keeps a site out of search answers.
According to OpenAI, GPTBot is used for model training and is separate from search. Search inclusion depends on OAI-SearchBot.
No one can guarantee it. You can make your pages accessible, specific and backed by sources, then measure citations over time.
Ask a fixed set of realistic buyer questions, record the date and conditions, save the answers and sources, and repeat on a schedule.
It varies. OpenAI says robots.txt changes take about 24 hours to apply; recrawling and how answers change are not on a fixed schedule.
Explore more guides in the resource library.