Landwehr (PeecAI): ChatGPT Query Fanout changed in November 2025

Key Takeaways:

• Landwehr and Wells analyzed PeecAI data on 20M+ Query Fanouts (QFOs):

  • Fanout queries used to have an average length of 6 words.

  • Mid November they started to become longer.

  • They peaked at 16 words in early December.

  • Since Christmas, they are stable at 12.

  • Global change: observed across multiple languages and countries 10+ countries in North America, Europe, and Asia

  • The structure of these new fanout queries suggests they are optimized for vector search - not traditional string matching.

• Interesting theory of Landwehr on why the change:

  For shorter queries, Google is doing a lot magic with query expansion, normalization, stemming, synonym matching, etc.

  That is good for new users. But bad for people who are looking for very specific documents. Like an LLM-based answer engine performing a search grounding.

• GEO implication:

  • QFO analysis for your specific industry and query is key to understanding citations and optimisation potential

  • QFO analysis got a lot harder as synthetic queris vary and rankings might differ more for longer more specific queries

For the prompt "best european car manufacturer" these were the last 10 fanout queries

Sources:

• https://www.linkedin.com/posts/landwehr_i-recently-wrote-about-how-chatgpt-fanout-activity-7421889498805891072-Ap9Q/