What Should You Do if ChatGPT Gives Customers a Scammer’s Phone Number?

Why do assistants repeat scam numbers?

Assistants find phone numbers the same way search engines do: by reading pages that mention a business and a number together. Scammers exploit that. They post claims that a number they control belongs to a company's customer service on review sites, message boards and low-quality directories. They also create fake business listings and pages that look official. When those claims are repeated often enough, and the company's real number is hard to find or inconsistent, an assistant may present the fake one. Companies whose customer service is hard to reach by phone are especially exposed.

What should you do first?

1. Document the answer. Take screenshots of the question, the answer and any cited sources, with dates. Try the question several times and in other assistants.
2. Warn customers. Add a notice to your website, help centre and social accounts: "Our only customer service numbers are... We will never ask for card details by phone to process a refund." Email customers if the scam is active.
3. Brief your support team so they can recognise reports, help affected customers and escalate.
4. Help victims. Tell affected customers to contact their bank or card issuer immediately, change any passwords shared and report the fraud.

What should the customer warning say?

Keep it short, specific and easy to find. A good notice names your official numbers, explains how your staff will and will not contact customers, and says what to do if someone has already called a fake number. For example:

Scam warning: fake "[Brand] customer service" numbers are circulating online and in some AI answers. Our only phone numbers are [number] (bookings) and [number] (refunds), open [hours]. We never ask for full card details or passwords by phone. If you have called another number, contact your bank now and email us at [address].

Link to it from the home page and help centre while the scam is active.

Where should you report it?

• OpenAI: use the feedback option on the answer and describe the false number, with the correct one and a link to your official contact page.
• Google: if the number also appears in AI Overviews or Search, use the feedback option there, and report fake Business Profiles and scam websites through Google's reporting forms.
• The sites hosting the fake number: review sites, forums and directories usually have abuse or report buttons.
• Hosting companies and domain registrars for fake websites impersonating you.
• Phone carriers, if you can identify the carrier behind the scam number.
• Authorities: in the US, report impersonation to the FTC at ReportFraud.ftc.gov. Other countries have national fraud reporting services.

How do you find where the number comes from?

• Open every source the assistant cites.
• Search the scam number in quotation marks on Google and Bing to find pages that repeat it.
• Search your brand with "customer service number" and similar phrases, as customers do, and look beyond the first page.
• Check for fake Business Profiles or map listings using your name.
• Look for the number linked to other companies. In the Washington Post case, the same number appeared to impersonate several cruise lines' hotlines.

Keep a log of each source and the date you reported it.

How do you make your official number authoritative?

• Publish a clear contact page listing every official phone number, what each is for, opening hours and a statement that you have no other numbers.
• Put the main number in plain text in your site footer and help centre, not only in images or behind a chat widget.
• Add it to structured data, using Organization or LocalBusiness markup with a contact point, matching the visible page.
• Use the same number on every listing: Business Profile, Bing Places, Apple Business, review platforms and social profiles.
• Answer the question customers ask. A help article titled "How do I contact [brand] customer service?" with the number in the first sentence gives assistants a direct source.

Old or inconsistent contact details make fake numbers easier to believe. Keeping facts consistent is covered in how to make a brand's information consistent across AI-visible sources.

How do you monitor for recurrences?

Scammers move on to new numbers and new sites. Once a month, ask ChatGPT, Gemini, Google's AI features, Perplexity and Copilot for your customer service number, refund line and complaints contact. Use different phrasings, including "phone number to cancel my booking" and "number for a refund". Record the answers and sources. Add a question about phone contact to support intake, so staff notice when customers mention a number you do not use. Related false claims about your business are covered in what a business can do when ChatGPT falsely calls it a scam or invents a lawsuit.

What legal options exist?

Impersonating a business is unlawful in many places. In the US, the FTC's Rule on Impersonation of Government and Businesses took effect in April 2024. It lets the FTC seek money back for victims and civil penalties from impersonators. Trademark law may also apply to fake sites and listings using your name and logo. Legal action against the assistant providers themselves is an unsettled area, and courts are only beginning to consider cases about AI-generated false statements. Take legal advice if customers suffer significant losses.

What might an illustrative case look like?

Consider an illustrative regional airline whose customers start reporting calls with a "refund agent" who asked for card details. The number came from ChatGPT and Google's AI Overviews. The airline's own website lists no phone number on its main contact page, only a chat widget. The airline posts a warning banner and emails customers who recently cancelled flights. It files reports with OpenAI, Google and the FTC, and has fake listings removed from two directories and a forum. It adds a contact page with its real numbers, hours and a statement that it has no other lines, and puts the number in structured data and all its listings. Within a few weeks, the assistants give the official number.

What do people ask about scam numbers in AI answers?

Is the business liable if customers lose money?

That depends on the circumstances and jurisdiction. Acting quickly to warn customers and remove the source reduces harm and shows good faith.

Why doesn't OpenAI just block the number?

Assistants draw on the web, and scammers change numbers often. Removing sources and publishing a clear official number is the lasting fix.

Should we list a phone number at all?

Yes, if you offer phone support. A clearly published number is the best defence against fake ones.

How quickly do fixes take effect?

Often within days to a few weeks, once fake pages are removed and your official pages are recrawled.