Incident guide

ChatGPT says my business is closed. It isn't.

You usually find out from a customer: "I asked ChatGPT and it said you'd shut down." It happened to a bookshop in 2026 — the AI said "permanently closed" with a wrong address, and the owner only learned about it weeks later from a customer. Wrong "closed" answers cost real orders every day they stand. Here is the correction path that actually works, in order.

First: confirm what the AI actually says

AI answers are not stable. Ask the same question ten times and you can get different answers — research on answer variance shows a third of the variation comes from resampling alone. Before you act, ask the engine the same question several ways ("is X still in business", "can I order from X", "is X open") several times each. If "closed" shows up repeatedly, you have a real problem, not a one-off hallucination. Screenshot everything with timestamps — you'll want a record.

The fixes that work, in order of leverage

  1. 1.Say you're open, high on your own site

    Put a direct statement — "[Brand] is open and shipping orders" — in the first 100 words of your homepage or an FAQ page, and update the page's modified date. Engines quote what they can read near the top of a page. This is the single fastest lever for engines that browse the web.

  2. 2.Fix your Google Business Profile

    A stale "permanently closed" flag on Google Business Profile is the most common source of closed-business answers. Verify the profile, set the status to open, and confirm hours and address.

  3. 3.Fix the sources the engine cites

    Perplexity and browsing ChatGPT show the pages they read. If those citations point at a directory, an old news story, or a stale listing, that page is the problem — contact the owner and get it corrected. Third-party corrections typically take days to two weeks to be re-crawled.

  4. 4.Make every listing agree

    Yelp, Bing Places, Apple Maps, industry directories, Wikidata if you have an entry: every source that says something different keeps the wrong answer alive.

  5. 5.Re-check weekly, and keep records

    Corrections propagate on the engine's schedule, not yours. Re-ask the questions weekly. Keep timestamped captures of every wrong answer — if the misrepresentation caused measurable harm, that record is what a lawyer will ask for.

Honest expectations, engine by engine

Perplexity is the most fixable: it cites the exact pages it read, and fixes on your own site are usually re-crawled within one to three days. ChatGPT with browsing picks up source fixes in days to weeks. Google AI Overviews depend on Google's index and typically take weeks. But Gemini, Claude, and ChatGPT answering without browsing speak from training data — there is no mechanism to correct them today; they update at the next retrain, months out. For those engines, the honest play is to fix every source now (so the retrain learns the truth) and document the wrong answers while they last.

One more thing: the thumbs-down and "report" buttons are feedback channels, not overrides. Use them, but don't wait on them — Google has ignored formal legal demands until a court intervened.

Frequently asked questions

How long until ChatGPT stops saying my business is closed?

If ChatGPT is browsing the web, source corrections usually show up in days to weeks. If it's answering from training data (no sources shown), the answer can't be corrected directly and typically persists until the next model update — fix your sources now so the update learns the truth.

Does clicking 'report' or thumbs-down on the answer fix it?

No. Those buttons are feedback, not overrides. There is no documented case of feedback alone reliably correcting a wrong business-status answer. Fix the sources the engine reads instead.

Can I sue over a wrong AI answer about my business?

It has been tried. A Minnesota contractor sued Google claiming $110M+ in damages after an AI Overview falsely said the state had sued them, and a 2026 German court ruled AI Overview statements are attributable to Google. But the first US AI defamation case (Walters v. OpenAI) was dismissed — these cases are hard to win. Whatever you decide, timestamped records of every wrong answer are the prerequisite.

Will an llms.txt file fix this?

No. Google's own documentation confirms llms.txt has no effect on Search or AI Overviews, and major AI crawlers largely don't read it. Spend the effort on your homepage, Google Business Profile, and the cited sources instead.

How do I find out if AI says this about my store before customers do?

Continuous monitoring. PositionBird's Accuracy Alerts asks the major AI engines about your business status every week, many times per question to beat answer randomness, and alerts you only when a wrong claim persists — with a hash-stamped record of every answer.

Find out what AI says about your store — before customers do

Run a free one-shot accuracy check right now, or let PositionBird monitor your business facts weekly across ChatGPT, Perplexity, Gemini, and Claude — with statistical debouncing so you're only alerted to real patterns, and a hash-stamped evidence record of every answer.

Run a free accuracy scan→Get the Shopify app ↗

Wondering how common this is? Read the research: how often AI gets small businesses wrong →