Why does AI get my company's location or contact details wrong?

An AI answer about your company comes from one of two places. Either the model answers from memory, or it searches the web and builds the answer from pages it retrieves. Wrong contact details can enter through both.

• Memory (parametric knowledge). The model learned facts up to a fixed cutoff. OpenAI's model docs list a February 16, 2026 cutoff for the GPT-5.6 family and August 31, 2025 for ChatGPT's Instant path. If you moved offices after that, a no-search answer still has the old address.
• Your own site. If the footer, contact page and an old press release show three different phone numbers, a retrieving engine has three sentences to choose from.
• Third-party pages. Engines often cite directories, review sites and old news coverage instead of your site. If those pages are stale, the answer is stale.

What this means: most location errors are copies of a real, outdated or conflicting fact. That is good news, because copies have a source you can find. If you are unsure whether an error is a copy or an invention, the guide on hallucination vs outdated answers shows how to tell.

Start by seeing the error in context. Ask each engine where your company is based and how to contact it, logged out and with memory off. The guide on finding out what AI says about your company covers how to run this cleanly.

Then read the answer for clues:

1. No citations, old address. The answer probably came from memory. Check whether your move predates the model's cutoff.
2. Citations to your own site, wrong detail. Open the cited page. Look for a second version of the fact elsewhere on your site, or for contact details that only load after JavaScript runs.
3. Citations to a directory or old article. The engine is repeating that page. That page is your fix target.

Retrieval is discretionary on every platform, so the same question can be answered from memory one time and from search the next. Run each question more than once before you decide.

The strongest evidence concerns access, not wording. Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. If your address sits in a map widget or a contact component that renders in the browser, those crawlers may never see it.

Access is also per bot. OpenAI's crawler page says OAI-SearchBot controls whether a site appears in ChatGPT search, while GPTBot governs training only, and each setting is independent. Blocking the wrong one can remove your contact page from ChatGPT search answers. For Google, the AI features page says a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode, with no extra requirements.

On content, the one positive signal is simple: a specific fact stated plainly on a fetchable page gives an engine a sentence it can quote. That is why one clear line such as "Headquarters: 120 Main Street, Austin, TX" beats a design that implies it.

What the evidence does not support: schema markup as a fix. Google says structured data isn't required for generative AI search, and Ahrefs' matched test of 1,885 pages found AI Overview citations fell 4.6% after adding schema. Existing markup does no harm. It just isn't the lever.

Work from the source you control outward.

1. Make one version true everywhere on your site

Pick the canonical address, phone number, email and headquarters city. Search your site for every older variant: footer, contact page, about page, legal pages, PDFs, old blog posts and press releases. Update or clearly date each one. Contradictions inside your own site are the cause you can remove completely.

2. Make that version fetchable

Put the details in the raw HTML, not only inside a JavaScript widget or an image. Check that your robots.txt allows the search bots (OAI-SearchBot, PerplexityBot, Claude-SearchBot) and that your CDN or firewall doesn't return a 403 to them. A robots.txt that allows a crawler means nothing if the firewall turns it away. The AI Crawler Checker requests a page as each AI crawler and shows which ones get blocked.

3. Correct the third-party sources engines cite

List every outside page the engines cited in your test. Update the directory listings you own, ask publishers to correct or annotate old articles, and close duplicate listings with the old address. Being named correctly on the sites engines already cite is associated with being described correctly. It is not a guaranteed switch.

4. Recheck on a schedule

Re-ask the same questions after the fixes have had time to be recrawled. The free checker on this site gives a snapshot of how engines answer up to three buyer questions, which is a quick way to see whether the cited sources changed.

• You moved recently. Memory-based answers will lag until the next model generation. Your controllable job is making the new address win whenever the engine searches.
• Multiple locations. State each location on its own page with its own phone number, and say which one is the headquarters. Ambiguity invites the engine to merge them.
• AI says you closed. A closed old location can read as a closed company. See why AI might say your company closed.
• Google surfaces specifically. AI Overviews draw on Google's index, so fix what Google indexes first. The page on wrong facts in AI Overviews covers that path.
• Invented details. If the engine produces a phone number that appears nowhere, treat it like other fabricated facts, covered in why AI makes up details like headcount.

You can't edit a model's memory, choose which pages it retrieves, or decide which source it cites. A critical survey of 45 GEO studies (arXiv, July 2026) found no technique with a stable, cross-platform effect on discoverability. What you control is whether the correct fact exists, is consistent, and can be fetched. Those fixes remove the reasons an engine gets it wrong. For the wider process across every brand fact, use the plan for fixing AI getting your brand wrong.

Can I ask OpenAI or Google to correct my address directly? There is no documented form that rewrites a model's answer about your company. Feedback buttons exist inside the apps, but nothing promises a change. The reliable route is fixing the sources the engine reads: your own pages and the third-party listings it cites. The page on [asking AI companies to correct brand info](/questions/ask-ai-company-correct-brand-info) covers what each vendor offers.

Will adding LocalBusiness schema fix a wrong address in AI answers? There is no evidence it will. Google says structured data isn't required for generative AI search, and Ahrefs' matched test found AI Overview citations fell 4.6% after pages added schema. Keeping existing markup is fine. The fix that matters is a plain, consistent address in the HTML that crawlers can fetch.

How long until AI shows my new phone number? Nobody publishes a timeline. Answers built from live search can change once the engine recrawls your corrected pages and the listings it cites. Answers from memory won't change until a newer model with a later knowledge cutoff ships. Recheck the same questions periodically rather than expecting a fixed date.

Why does ChatGPT show the right address but another engine shows the wrong one? Each engine uses its own index, crawlers and sources. ChatGPT search depends on OAI-SearchBot access, Google's AI features depend on Google's index, and Copilot grounds on Bing. One engine being right tells you nothing about the others, so test and fix each one separately.

Does blocking GPTBot cause wrong contact details in ChatGPT? Not in search answers. OpenAI says GPTBot governs training only, and OAI-SearchBot controls ChatGPT search, with each setting independent. Blocking OAI-SearchBot, or a firewall rejecting it, can keep your contact page out of ChatGPT search answers, leaving older third-party pages as the main source.