ChatGPT shifts from live fetches to its own index, Landwehr reports
A sustained drop in ChatGPT-User bot visits coincides with more OAI-SearchBot activity, pointing to retrieval from an internal cache.
Malte Landwehr posted on LinkedIn on 8 November 2026 that ChatGPT has changed how it gathers source content. Instead of fetching the same URLs live on every query, sometimes more than 100 times a day for one page, it now pulls from an index or cache.
He observed a substantial and sustained drop in visits from the ChatGPT-User bot. Over the same period, the OAI-SearchBot has become more active. Landwehr's reading is that content is being served from OpenAI's own index, which he has previously called Labrador on the Monthly Landwehr podcast.
He addresses two alternative explanations. Fewer sources per answer would also reduce ChatGPT-User traffic, but he says that is not what is happening. Content could come from Bing, Exa or Parallel, but that would not explain why OAI-SearchBot crawl activity is rising at the same time.
What to check in your own logs
This is one practitioner's observation across the sites he monitors, not a disclosure from OpenAI. Before treating it as settled, separate ChatGPT-User and OAI-SearchBot hits in your server logs and compare the last few months. If you see the same pattern, a page's standing with ChatGPT now depends on when OAI-SearchBot last crawled it, not on whether a live fetch happens at answer time.
That changes what stale content costs you. A page OpenAI indexed six months ago may be answering today's questions with six-month-old facts. The response is the ordinary one: keep the pages that matter current, and make sure OAI-SearchBot is not blocked in robots.txt or at the edge.