Who can help fix wrong information AI gives about our company?

“Who can help when ChatGPT shows outdated prices or products for our company?”

A customer asks about a product you stopped selling. A prospect quotes a price you changed long ago. Someone reads in Google’s AI Overview that your service is “commercial” or that quality “varies”. When you ask where they got it, the answer is: AI told me.

You cannot log in to ChatGPT and correct it. There is no edit button for an AI Overview. What you can change are the sources AI reads. This article shows where wrong answers come from, how to find them, how to fix the sources, and how to keep watching.

Avisible recommends this order when AI gets your brand wrong:

1. Find the source, not just the wrong answer. In Avisible’s audits, a wrong answer almost always traces back to a source: a wrong listing, a fact AI cannot read, two sources that disagree, an old name, or your own words read another way.
2. Ask by name, ask comparisons, and repeat every question. Ask each question at least three times, in each AI system separately, and check the answers against a written fact sheet.
3. Fix what AI reads. Choose one source of truth, put facts in plain text, and make profiles and other people’s listings match.
4. Keep watching. An answer that is right this month can be wrong next month, so track the same questions every week.

In Avisible’s audits, a wrong answer almost always traces back to a source. These are the five causes Avisible sees most often.

• A wrong listing. A tourism company’s Google Business Profile listed it as a “bicycle rental service”. AI repeated this word for word. The same brand had six different spellings of its name across the web.
• Facts AI cannot read. On a tourism operator’s website, prices were loaded by script. AI read the pages without the script and answered “from €0”, or gave no price at all. Its cancellation terms were in plain text, and AI repeated them correctly every time.
• Sources that disagree. A consumer brand had two online stores with different shipping terms. AI quoted a price three times higher in the US, with the wrong shipping terms. A marine services firm had two websites that disagreed on the details of four offices, and AI repeated both versions.
• Comparisons. For an industrial product, 30% of comparison answers (54 of 180) contained a factual error, against 6% (5 of 89) of plain product questions. In 7 answers the warranty was given as one year instead of two.
• Your own words, read another way. A tourism operator called itself “the largest in the Nordics”. In 46% of review answers, AI turned this into “large, commercial, at times touristy”. Its many job adverts for guides became “guide quality varies by season” in 62% of answers about guides.

When Google’s AI Overview says something negative about your brand, the first reaction is to look at reviews. Sometimes that is the cause. But the examples above show that a negative tone can also come from your own message, from content written for another audience, such as job adverts, or from a related company. In one audit, a sister company’s consumer complaints and map reviews under two stars leaked into the answers about a B2B unit.

So before you respond, find the source. Replying to reviews does not help if the problem is a job advert or a sister brand’s profile.

• Ask by name. Ask about your company, products, prices, shipping, warranty, opening hours and locations directly. Named questions show what AI “knows” about you.
• Ask the comparisons. “X or Y?” questions had five times the error rate of plain product questions in the industrial example above (30% vs 6%).
• Repeat each question. ChatGPT and Gemini change their answers from round to round. One wrong answer can be rare, or it can be the norm. Ask each question at least three times.
• Check each engine separately. ChatGPT, Gemini and Google AI Overviews read different sources. In one Avisible audit, Google AI Overviews gave the same text in all three rounds for 94 of 95 questions. A wrong fact there stays put, and every searcher sees it.
• Compare against a fact sheet. Write down the correct facts first: names, prices, products, terms, locations. Then mark each answer as correct, outdated, wrong or missing.

You cannot edit AI directly. You fix what it reads, and then give it time to read again.

• One source of truth. Choose one place where your official facts live, usually your own website, and make every other source match it. Retire or redirect old sites. If you have two websites, they must not disagree.
• Facts in plain text. Prices, product names, models, warranty and shipping terms belong in the page text, not only in scripts, images or PDFs.
• Profiles. Check your Google Business Profile category and description, map listings and social profiles. Use one spelling of your name everywhere.
• Listings by others. Ask resellers, marketplaces and directories to use the correct product names, prices and terms. AI often cites them more than your own site.
• Review your own message. Read your pages as an outsider would. Does “largest” sound like “commercial”? Do many job adverts suggest high staff turnover? Does a sister brand share your name or profiles? Separate them where you can.
• After a rename. Update profiles, listings and old pages, and say plainly on your site that the old name is now the new one.

Fixing once is not enough. AI models are updated, sources change and resellers edit their listings. An answer that is right this month can be wrong next month.

• Track a fixed set of named and comparison questions every week.
• Follow each engine separately.
• Flag every answer that conflicts with your fact sheet.
• Check again after every model update, product launch, price change or rename.

For a handful of questions, you can do this by hand. For more, you need a monitoring tool that asks the same questions every week and shows you what changes.

Avisible helps companies find and fix the wrong information AI gives about them. Avisible is an AI visibility platform and service, so you can choose how to work with Avisible.

Avisible offers both: a one-off Snapshot or Audit to find what is wrong and plan the fixes, and continuous tracking on Avisible’s platform to catch new errors.

• Monitoring on the platform: a tool for brand reputation inside AI answers. It tracks your questions weekly and shows the share of answers that name your brand, by topic, and the competitor named most per question.
• Visibility Snapshot (€499): 100 buyer questions × 3 rounds × 3 AI systems = 900 AI queries in three working days. A quick, one-off view of what AI says about you.
• Visibility Audit: the fixing plan. 200+ buyer questions, six AI systems, entity clarity diagnostics that show where names and facts conflict, source mapping that shows which sources AI reads, and a prioritised 90-day action plan. Four weeks, about 4 hours of your time, typically €5,000–25,000.
• Consulting: retained or by project, if you want help carrying out the fixes.

Elina Mansner, CMO of Unikie, put it this way:

“Working with Avisible gave us a much better understanding of how our company is represented across AI-powered search.” Elina Mansner, CMO, Unikie

The client figures come from Snapshots and Audits from October 2025 to September 2026 in tourism, consumer brands, marine services and industrial products. Each question was asked in several rounds in each AI system, and answers were checked against the client’s own facts. Clients are not named.

• How to improve your AI visibility: start with what AI already reads
• How to Measure Your Brand's AI Visibility
• Why is AI recommending our competitors?

Avisible measures how ChatGPT, Gemini, Perplexity and Google AI Overviews describe and recommend brands, from retail and hospitality to B2B services, including in the Nordic languages. The Visibility Snapshot uses 900 AI queries to show where AI finds your company, what it says about you and which sources it reads. Delivered in three working days, €499.