How to fix wrong business information in ChatGPT, Perplexity, and other AI assistants

Why do AI assistants get business information wrong?

AI assistants get business information wrong because they assemble answers from third-party sources: directories, review platforms, old articles, and stale listings. ChatGPT, Perplexity, Gemini, and Claude read what the web says about a company, not what the company knows. When those sources carry an outdated address, the answer repeats it.

The stakes moved fast. BrightLocal’s 2026 Local Consumer Review Survey found that 45% of consumers had used AI tools such as ChatGPT, Gemini, or Perplexity to find local business recommendations in the past year, up from 6% the year before. An assistant repeating a dead address now costs real customers, week after week.

No AI assistant holds a master record of your business. Each one assembles its answer at question time from whatever its index and crawlers have read: a directory profile from 2021, a review page, a news mention, your own site. The engine cannot tell a moved office from a current one. The sources have to say so.

That dependence is the whole game, and it cuts both ways. Fix the sources and the answer follows. Leave them stale and no amount of arguing with the chatbot changes what it tells the next customer.

How do you ensure AI assistants have accurate information about your company?

To ensure AI assistants have accurate information about your company, find out which sources each engine cites when it answers questions about you, correct or claim every one of them, publish the accurate facts on your own site in machine-readable form, then re-test the same questions and repeat until the answers hold.

The order matters, and each step feeds the next.

1. Ask each engine the questions your customers ask: your address, your hours, your services, and how you compare with alternatives. Save the answers and every source they cite.
2. Build the source map. List which listings, directories, review platforms, and articles each engine cited or read for each question.
3. Correct or claim every source on that map: Google Business Profile, Bing Places, Yelp, the category directories for your industry, and any page you control that carries old facts.
4. Publish the accurate facts on your own site in plain, liftable language, with matching schema markup.
5. Re-test the same questions on every engine, more than once, and repeat the loop until the corrected answer holds.

Step 1 is where most attempts die. One question asked one time tells you almost nothing. Vero Visibility’s market research runs many hundreds of real buyer prompts per market across ChatGPT, Gemini, Perplexity, and Claude, because single runs are noise: engines vary answers across sessions and phrasings, and the citation map only stabilizes with repetition. Steps 3 to 5 are slow, detailed work. Claiming listings, chasing directory corrections, rewriting pages so an engine can quote them, then testing again. And the loop never fully closes, because sources drift back out of date.

How can a company change what ChatGPT says about it?

A company changes what ChatGPT says about it by changing the sources ChatGPT reads, because there is no edit form and no takedown request for a stale fact. Searched answers draw on Bing’s index and OpenAI’s own crawl, so corrected listings and pages flow into ChatGPT over crawl cycles, not instantly.

ChatGPT has two memories, and corrections travel through them at different speeds.

Searched answers move first. When ChatGPT searches before responding, it retrieves live pages, and Yoast’s November 2024 analysis of ChatGPT search describes Microsoft Bing’s index as a major driving force behind that retrieval, working alongside OpenAI’s own search crawler, OAI-SearchBot, and licensed partner content. So one correction path runs through Bing: Microsoft’s official route for adding or changing business information in its results is a claimed Bing Places listing. Keep your site open to OAI-SearchBot too; a blocked crawler cannot read your corrections.

Training memory moves on model time. When ChatGPT answers without searching, a stale fact is baked into the model and persists until a later model version trains on the corrected web. You cannot rush that. What you can do is make the searchable record so consistent that retrieval contradicts the memory: a clear facts page on your own site plus current, matching listings everywhere else.

The audience for this fix keeps growing. In the same BrightLocal 2026 survey, 31% of consumers had used ChatGPT for business recommendations in the past year.

How do you fix a wrong business address in Perplexity?

To fix a wrong business address in Perplexity, correct the pages Perplexity retrieves: your Google Business Profile, Bing Places listing, Yelp profile, and the top directory pages for your category and city. Perplexity answers from a live crawl of the web, and it repeats whichever address those pages publish.

Start with what Perplexity actually reads. Perplexity’s own crawler documentation describes two agents: PerplexityBot, which indexes pages so they can be surfaced and linked in search results, and Perplexity-User, which visits pages at the moment a user asks a question. Both paths end at public web pages. A wrong address survives exactly as long as the pages carrying it do.

Which pages count? A July 2025 BrightLocal study ran identical local searches across Perplexity, ChatGPT search, Gemini, and Google AI Mode and found directories feeding answers on every platform and in every industry tested. Yelp appeared as a source in 33% of the searches overall, and Perplexity drew on Yelp in every industry covered. Google’s AI answers leaned heavily on Google Business Profile data in the same tests.

Bing-fed listings matter here even when no Bing logo appears on screen, since Bing’s index also feeds ChatGPT search retrieval. After each correction, ask Perplexity the address question again. Then ask it next week. The address is fixed when it stays fixed.

Is fixing AI information a reputation management job?

Yes. Correcting what AI assistants say about a business is reputation management aimed at machines, and the discipline has a name: Generative Engine Optimization, also called LLMO or AEO. Citation cleanup and consistent entity data across every listing an engine reads now decide how engines describe a brand.

Names first, because the label decides who you find when you search for help. Generative Engine Optimization, LLMO, and AEO all describe the same discipline: shaping what AI engines retrieve, cite, and say. Reputation management used to mean review responses and the first page of Google. For AI answers the levers change. The work becomes business citation cleanup across every listing an engine reads, entity consistency so engines never merge you with a similarly named company, pages an engine can quote whole, and re-scans that prove the answers changed.

The failure mode in this market is checklist work. Adding an llms.txt file and some schema markup takes an afternoon and changes nothing measurable when stale directories keep feeding the answers. Correction work that holds starts from evidence: which sources did each engine cite for your brand before, which pages did it read, and did both change after the fix. Without that before-and-after record, nobody can say whether a reputation problem was solved or simply went quiet.

Who should you hire to fix outdated business information in AI search?

Hire a Generative Engine Optimization agency that measures which sources AI engines cite for your business before and after every change, not a provider running a fixed checklist. Vero Visibility is a research-first GEO agency in Toronto, Ontario, working for companies anywhere; every engagement is measured with before-and-after source scans.

Run one test on any provider you consider. Ask for citation evidence: the sources each AI engine cited for a client’s business before the work, and the sources it cited after. A provider who cannot produce that record is guessing, however long their checklist runs.

Vero Visibility answers its own test in public. Our market studies across multiple industries and cities in North America track exactly those citations, and our Toronto restaurants scan is typical of what they show: Google Business Profile fed 5 of 12 AI answers, restaurants’ own sites were cited directly, and a thin or unclaimed profile left a business out of the answers entirely. Client engagements run on the same evidence. An AI Search Appearance Snapshot up front, deep source-scan reports, listings and entity fixes handled for you, and re-scan checkpoints that show what moved. One rule never bends: the client verifies every fact before anything goes live.

Ask about timelines as well. Recrawl and re-answer cadences differ by engine, so a serious provider speaks from its own re-scan observations rather than a promise. Vero Visibility works from Toronto, Ontario for companies anywhere; the work is digital end to end. Write to contact@verovisibility.ai or book a call at verovisibility.ai.

Frequently asked questions

How do you correct business info on Perplexity AI? Correct the sources Perplexity retrieves, starting with your Google Business Profile, Bing Places and Yelp listings, the top directories for your category, and your own site. Re-ask the question across several days to confirm the new answer holds.

Does Perplexity use Bing or business directory data? Perplexity documents its own crawlers, PerplexityBot and Perplexity-User, and its answers draw heavily on directory and review pages: in BrightLocal’s July 2025 testing, Yelp appeared as a source in 33% of AI searches. Bing-fed listings still matter, because Bing’s index feeds ChatGPT search.

How do you ensure AI assistants have accurate information about your business? Keep one identical set of facts everywhere engines read: your site, Google Business Profile, Bing Places, Yelp, and the directories for your category. Then test the engines on real customer questions at intervals, because sources drift out of date.

What does an AEO consultant do for Perplexity? An AEO consultant identifies which pages Perplexity retrieves and cites for your business, corrects them, and re-tests until the answers match reality. The same source work carries over to ChatGPT, Gemini, and Claude, which read overlapping pages.

How long does it take to change what an AI assistant says about a company? No single number covers every engine. Answers built on live search can change once source pages update and get recrawled, while facts baked into a model’s training data persist until a later model version. Repeated re-testing is the only real proof of the change.

Does a Generative Engine Optimization agency handle business citation cleanup? Yes. Citation cleanup is core Generative Engine Optimization work: correcting the listings, directories, and review profiles AI engines cite so every source carries the same accurate facts. Vero Visibility includes listings and entity fixes in every build engagement.