Google and OpenAI AI search controls side by side
One table of what each vendor documents, with the source for each row. Where a vendor does not document something, the cell says so. Read on 2026-10-15.
Last updated 2026-10-15
The table
| Question | OpenAI | |
|---|---|---|
| Allow crawling for AI search | robots.txt for Googlebot; page must be indexed and eligible to show with a snippet. | Allow OAI-SearchBot in robots.txt and allow its published IP ranges. |
| Keep a site out of AI search answers | Exclude in the Search generative AI control, or noindex to leave Search entirely. | Opt out of OAI-SearchBot. Such sites can still appear as navigational links. |
| Limit use for AI training | Google-Extended. The generative AI control does not affect training. | Disallow GPTBot. |
| User-started fetches | Not described in the pages we read. | ChatGPT-User; robots.txt rules may not apply. |
| Measure AI exposure | Generative AI performance report (impressions only), plus the Web search type in Performance. | No site-owner report described in the pages we read. |
| Time for a change to apply | Search Console control: about 1-2 days after it goes live, some content longer. | About 24 hours after a robots.txt update, for search results. |
| Guarantee of inclusion | Indexing and serving are not guaranteed. | No way to guarantee top placement. |
Sources by row: Google rows use the three Google pages, OpenAI rows use the crawler page and the ChatGPT search help page. Links are on the Google and OpenAI pages of this site.
What the table does not show
- Which sites are chosen for any answer.
- Any effect of these settings on citations or traffic. Neither vendor documents one, and we have no measurement.
- Other AI products that use their own crawlers.
Last read 2026-10-15.