UK Universities AI Answer Study 2026: what ChatGPT, Perplexity and Google AI Overviews tell applicants

A fact-checked snapshot of what three AI search tools told UK university applicants in October 2026. On 20 and 21 October 2026, 21 applicant-style questions about UK universities (entry grades, fees, league-table positions, graduate outcomes, contextual offers and "which is better" comparisons) were put to ChatGPT, Perplexity and Google AI Overviews: 63 answers in total. Every specific, checkable claim in those answers (198 claims) was checked against the university's own pages or official data (DfE, HESA, Discover Uni, UCAS), and given one of five outcomes: Correct, Wrong, Outdated, Overstated or Unverifiable. Headline results: 125 of 198 claims (63%) were correct, and 48 (24%) were wrong, out of date or overstated. ChatGPT was correct on 72% of its claims, Perplexity on 61% and Google AI Overviews on 55%. Accuracy depended more on the source an answer drew on than on the platform: answers drawn from a university's current official page were right 79% of the time, against 7% for answers drawn from out-of-date pages on the university's own website. The dataset also includes an audit of the rankings pages of all 31 universities in the study set (the 24 Russell Group members plus Bath, Dundee, Lancaster, Loughborough, St Andrews, Strathclyde and Surrey), recording which league-table editions each page showed, with every row backed by a dated PDF or screenshot. Pilot runs using open "which universities should I apply to?" questions are published separately (pilot_claims.csv, pilot_runs.csv) and are not included in the headline figures. They show ChatGPT tailoring university shortlists to the user's approximate location, signed in and signed out. Full write-up, report cards for each university and a checklist for university web teams: https://aivisibilitygap.com/uk-universities-ai-answers.html