25%
Travellers who received outdated or inaccurate AI travel information
25% of travellers report having received outdated or inaccurate travel information from AI.
What goes wrong: stale answers, unverified claims, and the policy response taking shape.
A quarter of travellers report receiving outdated or inaccurate travel information from AI. That is not a model failure in the abstract; it is usually a content failure, where the facts a model could reach were wrong, old, or absent.
This is the operational core of the problem for travel brands. A model that cannot find current, structured information about your property will either omit you or describe you from whatever it last saw, and the second outcome is worse than the first.
On the policy side, intergovernmental work has moved from describing AI to setting adoption conditions: data protection, SME access, workforce transition and equitable distribution of benefits.
Booking.com · February–March 2026 · Europe
UN Tourism · 2025 · Global
Google · September 2025 · Global (24 countries)
Booking.com · April–May 2025 · Global (33 markets)
Amadeus · 2025 · Global
Cloudflare · June 2025 · Global
25%
25% of travellers report having received outdated or inaccurate travel information from AI.
42%
29%sometimes fact-check42% of consumers always fact-check AI outputs and a further 29% do so sometimes.
6
The OECD identifies six priority areas for responsible AI adoption in tourism across APEC economies.
70%
70% of tourism businesses already use AI technologies, principally chatbots, dynamic pricing and demand forecasting.
94%
60%properties with under 10 staff94% of properties with 250+ staff feel prepared for cybersecurity threats, against 60% of those with fewer than ten.
52%
52% of executives report their organisations have deployed AI agents in production, per a Google Cloud study from September 2025.
3 of 4
1 of 4state no such exceptionFor automatic crawling, yes. For a fetch a user triggered, three of the four major assistant makers publish that robots.txt may not apply.
~14%
312domains disallowing GPTBotAbout 14%, or 546 of the 3,816 top domains where a robots.txt file was found. GPTBot was both the most blocked and the most explicitly allowed.
25.9%
13.3%Anthropic and Common Crawl25.9% of tokens in the head of the C4 corpus were restricted to OpenAI's crawlers by April 2024, against 13.3% for Anthropic and Common Crawl.