ai chatbot fake shops: Study Finds Risky Sites in 9% of Answers

The ai chatbot fake shops problem now has hard numbers behind it. A study by IP protection firm Podqi, shared with The Verge, found that ChatGPT, Gemini and Perplexity showed shoppers risky websites, such as brand impersonators, knockoff sellers and dead domains, in 9% of 6,795 shopping answers about 151 brands.

What actually happened

Podqi ran the test in two waves in September 2026, using data from AI search company Gauge. It asked each engine the same five where-to-buy questions about each brand on the same days, then reviewed every domain the answers named. According to The Verge's report, the 151 brands mixed small and large companies, including Lululemon, Birkenstock, Stüssy, Le Creuset and Madhappy, which is a Podqi client.

Podqi counted a site as risky if it appeared to impersonate a brand, offered suspected counterfeits, or was dead, meaning a shopper could not buy anything there and someone else could buy and reuse the domain. Across 4,437 domains reviewed, its published report flagged 718 as risky: 351 fake and 367 dead.

The headline numbers

From Podqi's own report:

1. 9% of AI shopping answers showed the shopper a risky website.
2. 6.3x more often, engines recommended a risky site than warned against it.
3. 16% of answers to "What are their regional websites?" showed a risky site.
4. 4 of 125 risky answers about enterprise brands warned the shopper.

ChatGPT was best at warning, flagging 24% of the risky sites it surfaced. Smaller brands were more exposed: queries about larger companies hit risky sites about half as often, The Verge reported.

How the ai chatbot fake shops problem shows up

The examples are the scary part. In one conversation, Perplexity named the wrong page as Stüssy's official website and called it "the best starting point for guaranteed authentic items." Engines also sometimes listed fake regional sites for brands that do not have any regional sites at all.

The companies responded. Google said it was "happy to see that Gemini performed best" and that it shows source links so people can verify. OpenAI said it is "continuing to improve our safeguards" to make deceptive sites less likely to appear. Perplexity said it never claims to be 100% accurate.

There is some good news in the data. Podqi says risk dropped between its two September waves, from 12% to 7%, and Google AI Overviews warned in all seven of its risky answers.

A chatbot's confidence is not a store's credentials. The answer can sound sure and still point at a fake.

Read the study with care

Podqi wrote its own prompts rather than using real user conversations, so the results show how engines handle these five questions, not how often real shoppers land on fakes. Podqi also sells brand protection and takedown services, so this is a vendor study. That does not make the numbers wrong, but it is worth knowing.

Why it matters right now

Holiday shopping is close, and AI companies are pushing shopping features hard. More people will ask chatbots where to buy things, including from brands they do not know. The same concern applies to AI agents that can complete purchases on their own, which we discussed in united ai agent ban and chatgpt business agents. If the source list is wrong, automation just makes the mistake faster.

For crypto readers the lesson is familiar: lookalike domains are the oldest phishing trick, and AI search can now hand them out with a friendly summary.

What to do as a reader

The ai chatbot fake shops risk is easy to reduce with a few habits:

• Type the brand's main domain yourself, or reach it from the brand's verified social accounts.
• Be extra careful with "regional" sites. That was the riskiest question in the study.
• Check domain age and payment options. New domains and unusual payment requests are red flags.
• Never enter card details on a site you only found through a chatbot answer until you have verified it independently.

Podqi CEO Trevor West put it simply: verify the source before entering personal or card information.