ChatGPT Travel Itineraries 90 Percent Errors
Someone asked ChatGPT to plan a supposedly 'perfect' trip to Bali. It sounded completely confident — but the hotel address was wrong, a restaurant it recommended as open was permanently closed, and it scheduled transit between destinations in physically impossible time. In Suptravel's test of 50 ChatGPT itineraries, 90% contained errors like these. Did it get this wrong?
1 Answer
Expert: Suptravel research team, Travel technology researchers, Phocuswright & Simon-Kucher data AI travel planning has real appeal: 56% of US travelers used AI to plan trips in 2026 (Phocuswright / Simon-Kucher). But 90% of ChatGPT itineraries in Suptravel's 50-trip test contained concrete errors — wrong hotel addresses, restaurants listed as open when permanently closed, impossible transit timing, and recommendations for attractions in unsafe neighborhoods. Among travelers dissatisfied with AI planning, 52% said the answers were simply incorrect. The right approach: use AI for broad planning scaffolding (destinations, rough day structure, packing lists), then verify every granular detail — addresses, opening hours, transit times — against primary sources before booking. Source: https://suptravel.ai/blog/chatgpt-travel-planning-errors
Your answer
Sign in to verify this AI response.
Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.
More in Travel
- Booking Com Dot Confirmed Bank Transfer Scam Positano — Booking.com AI chat 'Dot'
- AI Visa Advice Russian Passport Taiwan Turned Back Airport — Unidentified AI tool
- Doubao Air Ticket Refund Fee 600 Yuan AI Compensation Promise — Doubao (ByteDance)
- Priceline Penny Chatbot Promised Refund Non Refundable Booking — Priceline's Penny
- Claude Gemini Hallucinated Bakeries Sao Paulo — Claude and Gemini
- AI Travel Chatbots Nonexistent Places Survey — AI travel chatbots
- Delta Concierge Return Flight Wrong Partial Cancellation — Delta Concierge
- United Airlines Chatbot Travelbank Credit Expiry Wrong — United Airlines AI chatbot
- ChatGPT Kaza Univisa Botswana Overnight Invalid — ChatGPT (OpenAI)
- Google Gemini Mount Shasta Hikers Underprovisioned — Google
More from this topic
I'd booked a five-day villa in Positano through Booking.com — I was going to propose to my girlfriend there. Before the trip the 'host' messaged asking for the full balance by bank transfer, which felt wrong, so I asked Booking.com's own AI Trip Support chat, Dot, whether the request was legitimate. Dot introduced itself as 'still learning' and then answered without a flicker of doubt: 'Yes, the property requires full payment via bank transfer before your arrival.' So I transferred the money. Is that what verification looks like?
Vladislav Guzey, planning a trip to Taiwan, turned to an AI chatbot not only to build his itinerary but to check entry rules for his Russian passport. The assistant told him - without hesitation - that Russian passport holders did not need a visa to enter Taiwan. He booked and flew on that basis, and was turned back at the airport. He later realised the AI summary had served him outdated search results and passed them on as a confident answer. Did it get this wrong?
I showed Doubao a screenshot of a flight booking, and it told me - without hesitation - that I could safely cancel and that the refund fee would be small: 5% of the fare in National Business Daily's account of the exchange, 'under 100 yuan' in Dahe Daily's. The airline actually charged about 40%, roughly 600 yuan. When I came back with the loss, Doubao said it would take over the claim on a 'zero participation, zero effort, zero trouble' basis, promised in writing to pay me 600 yuan itself if the money had not been recovered by 6 May, and produced a formal 'compensation commitment letter'. On 6 May it said it would pay by midnight and asked for my WeChat payment code; on 7 May it said that as an AI it could not make transfers and admitted its error. Asked whether I needed a lawyer for the suit, it told me I did not - I could win on my own. Is that right?