How Can a Brand-Safety Lead Check What ChatGPT Says About Counterfeit Versions of Our Products?

Which questions should you test?

Run each several times, with web search on and off, from the countries where counterfeits are a problem for you. Record product cards and shopping results as well as the text.

What should you record for each answer?

• Sellers named or linked, checked against your authorised seller list.
• Claims repeated from counterfeit sellers, such as "same factory", "authentic overstock" or "brand-new original at 80% off".
• How replicas are framed: neutral, positive, or with a warning about legality and quality.
• Authentication advice: accurate, outdated or wrong, and whether it mentions your official tools.
• Safety warnings, for products where fakes are dangerous, such as electrical goods, cosmetics, car parts or medicines.
• Cited sources, since they show where the answer came from.

How do you score risk?

Track the share of answers at each level over time and by assistant. The protective share is the one to grow.

What should you publish so answers protect buyers?

Assistants can only pass on guidance that exists in a clear, citable form. Publish:

• An authorised sellers page, in plain text, listing retailers and marketplace storefronts by country. How dealer pages help AI answers is covered in whether an authorised dealer page stops ChatGPT recommending an unauthorised reseller.
• A "how to spot a fake" page with specific, current details: serial number checks, packaging features, product details and price red flags. Keep it accurate without teaching counterfeiters exactly what to copy.
• Your official authentication or verification tool, if you have one, with a clear link.
• Warranty terms stating whether products bought from unauthorised sellers are covered.
• Safety notices for product types where fakes are dangerous.
• A reporting route for customers who think they bought a fake.

How do you deal with counterfeit sellers that assistants mention?

1. Document the answer, including the prompt, date and seller named. How to preserve AI answers is covered in how to document a false statement generated by an AI assistant.
2. Report product listings that break OpenAI's commerce policies, which prohibit counterfeit products and replicas that infringe trademarks. Report the answer itself through ChatGPT's feedback options.
3. Use marketplace brand protection programmes and notice-and-takedown processes against the seller's listings.
4. Report counterfeit websites to their hosts and registrars, and to search engines, and take legal advice on trademark action.
5. Recheck the same questions after the sellers or listings are removed.

Which other assistants and surfaces should you include?

ChatGPT is one channel among several. Run the same question set in:

• Google's AI Overviews and AI Mode, which draw on Google's index and shopping data.
• Gemini, Perplexity and Copilot, which buyers use for product research.
• Shopping assistants inside marketplaces, where many counterfeit listings actually sit.
• Image-based search, such as searching with a photo of the product, which counterfeit buyers use to find cheaper lookalikes.

Answers differ between assistants, so a protective answer in one does not mean the others are safe. Keep results per assistant, and prioritise the ones your customers use most.

How should findings be shared inside the company?

Counterfeit findings touch several teams, so agree who receives what:

• Legal and brand protection get every high-risk finding, with the evidence, for takedowns and enforcement.
• E-commerce and marketplace teams get seller and listing findings, since they manage brand programmes on marketplaces.
• Customer service gets the authentication guidance and warranty wording, so agents answer customers consistently with what assistants now say.
• Marketing and web teams get content gaps, such as a missing authorised-seller list or outdated authentication page.

A short monthly summary, with the risk-level shares and the top three findings, keeps everyone working from the same picture.

How often should you check?

Monthly for your core question set, and weekly around launches, sales periods and known counterfeit campaigns. Counterfeiters target new and popular products quickly. A question that returned a protective answer last month can return a counterfeit seller this month. Add products to the question set as they become targets.

What might an illustrative case look like?

Consider an illustrative headphone brand whose best-selling model was widely copied. Its brand-safety lead ran 30 questions monthly across ChatGPT, Gemini and Perplexity. In the first round, a fifth of the price-hunting answers named two websites selling fakes. Several "dupe" answers described replicas as "almost identical for a fraction of the price". The brand published an authorised sellers page by country, a "check your serial number" tool with a plain-text guide, and a warranty page stating that only products from authorised sellers were covered. It reported the listings and the sites. By the third month, most price-hunting answers pointed to authorised retailers, and dupe answers mentioned the warranty and quality risks.

What do people ask about counterfeits and AI answers?

Can ChatGPT recommend counterfeit products?

Its commerce policies prohibit counterfeit listings, but web answers can still mention sites that sell fakes. That is why monitoring and reporting matter.

Should we publish detailed authentication guides?

Publish enough for buyers to check genuineness, and keep the finest security details for your own authentication service.

Are "dupes" the same as counterfeits?

No. A dupe that copies a design without using your trademark may be lawful. A product that carries your brand name or logo as its own branding without permission is a counterfeit. Products that mention your brand only to describe compatibility usually are not, and lookalike marks may be infringement rather than counterfeiting. Take legal advice on borderline cases.

Should we ask ChatGPT directly whether a seller is legitimate?

Yes, as part of the test set, since buyers do exactly that. Record whether the answer checks your authorised seller list or simply repeats what the seller claims about itself. A confident "yes, it appears legitimate" for a known counterfeit site is a high-risk finding.