AI can navigate Wikipedia links but not as fast as humans
Explains how AI compares to people in moving through linked knowledge without hype or alarm.
- Publication
- arXiv
- Stage
- Preprint
- What we read
- Summary of the paper
- Authors
- Alexander Hickey, Joanna Knight, James Bishop
- Universities and research institutions
- Not yet supplied in verified metadata; the Brief does not guess.
What they did and found
In tests, AI tools tried to reach a target article by following links on Wikipedia and OpenAlex, up to 100 steps. Some AI runs matched human success rates, but most needed more steps and had more failures.
Why it matters
Users can choose between faster, more costly AI routes and cheaper, human-guided searches, depending on the task and the budget requirements.
The study compared simple text methods and larger language tools with people. Some AI found the target, but it was not consistently faster or cheaper.
For routine tasks, AI prompts can help; for complex or risky work, expect more checks and some human guidance to avoid errors and costs upfront. Should you rely more on AI?
What remains uncertain
Results vary by language and topic; findings may not apply to all domains or longer tasks in practice for some cases.
Original sources · 1
- Can Language Models Play the Wikipedia Game? ↗arXiv · 2026-10-09
Check the original paper for its authors, methods, version and access terms.
What is your take?
Ask a question, add useful context or share a different perspective. Keep the conversation respectful and grounded.