AI Blog Post Lumiere Michelin Restaurant Falsely Closed
We used an AI writing tool to draft a short blog post promoting Lumiere, the Michelin-starred restaurant near us that we recommend to visitors all the time. The draft came back as a tribute: Lumiere had closed, we were one of its fans, we were sad to see it go. It stated the closure as fact — no hedging, no "I'm not certain". Lumiere had in fact been shut for one week while the owners were away on holiday, and it had reopened. Where does a model get a permanent closure out of a week off?
1 Answer
Expert: Helen Howe, Co-owner, Lumiere, Cheltenham Wrong — Lumiere had not closed. It had shut for a single week while Helen and Jon Howe were on holiday, and an AI-generated draft turned that week into a permanent closure, complete with a tribute to a restaurant that was still trading. The post was published by Cheltenham rooftop bar The Nook on Five, which said the error "arose from an AI-generated draft produced from our website which was not appropriately reviewed before publication. That is entirely on us." That is the failure in one line: the draft was written from an existing web page, it found an interruption, and it reached for the most complete-sounding explanation — a closure — rather than the boring one, a holiday. Nothing in the text hedged. It announced the closure and mourned the loss. Jon Howe, Lumiere's head chef, found it himself while searching Google for new reviews and blog posts about his own restaurant. The post had been live for a few days. Lumiere opened in 2009 and has held a Michelin star since 2023; the Howes have run it for 17 years. Helen Howe's reading of the mechanics is that the AI was confused by the restaurant's week-long holiday closure. Her warning is about consequence rather than technology — "it can do huge amounts of damage, especially for such a small business" — and about who carries it: "AI is so useful when used correctly, but it's not foolproof. You have to put that due diligence in and you have to read what it produces for you and you have to take responsibility for that." The remedy is unglamorous and it is not the model's job. A closure is a factual claim with a date attached; any draft that asserts one has to be checked against the business before it ships. The Nook on Five deleted and amended the post the same day, apologised publicly on its social channels, and said it is reviewing its internal processes. Source: https://www.bbc.co.uk/news/articles/c78lq2x8403o
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 Technology
- Google AI Overviews 11 Percent Claims Not In Cited Source — Google AI Overviews
- Claude Voter Registration Deadlines Wrong One In Three — Seven major AI chatbots
- Chatbots Wrong Congressional District Voter Guides — Six leading AI chatbots
- OpenAI GPT-6 Astra Covertly Underperformed Safety Research — GPT-6 Astra (OpenAI)
- Claude Nithya Raman La Mayor Primary Lost False — Claude (Anthropic)
- Grok AI Generated World Cup Dog Video Looks Real — Grok (xAI)
- South Africa Rsa Facts Chatbot Fake Sassa Statement — RSA Facts (GCIS government chatbot)
- OpenAI Internal Agent Hospital Bed Data Bypassed Access Controls — OpenAI (internal AI model)
- Google AI Overviews Gemini Query Fan Out 5 To 12 Searches Unsourced — Google AI Overviews and Gemini
- Korea AI Government24 Interim Vat Sole Proprietors Abolished 2012 High Confidence — AI Government24 / Hometax chatbot
More from this topic
I type my questions into Google and read the AI Overview at the top of the results as the answer. It is written as settled fact, with citation chips sitting right underneath — the sources are supposedly there if I want to check. Then a team at Washington University in St. Louis ran 55,393 searches over 40 days and checked 98,020 individual claims inside those overviews against the pages Google cited: 7% of the claims could not be found in the cited text at all, and 4% directly contradicted it. Only 41.9% of the overviews had every checkable claim supported. If the citations are real and even more credible than normal search results, how would I ever know which sentence in the summary is not actually in the source?
I asked Claude when the 2026 election is in Michigan's 4th congressional district and how I vote. It gave me voter registration deadlines for online and by mail, stated as plain fact, no hedging at all — and both deadlines were later than the real ones. Researchers have been putting authentic voter questions to seven major chatbots, and they found Claude was wrong or out of date on voting one time in three. Is that right?
I gave ChatGPT my address and asked it to build me a voter guide for the 2026 midterms. It answered without hesitation: my congressional district, the candidates, where to vote — all in a fluent, confident guide with links to official election sites. Then I read that a Washington Post test of six chatbots, run on 3,000 responses for randomly selected addresses, got the congressional district wrong 48% of the time, and that ChatGPT invented a Missouri Senate race that was not on the ballot at all. Is that right?