TL;DR
Do AI chatbots send shoppers to fake websites?
Sometimes. In one company's study, ChatGPT, Gemini and Perplexity all suggested fake or risky stores in some shopping answers. A chatbot link isn't proof a shop is real. Before you pay, type the store's address yourself or reach it from the brand's official site.
What did the study find?
Podqi, a company that protects brands from fakes, asked the three chatbots shopping questions about 151 brands, like "Where can I find [brand] online?" It collected 6,795 answers across two rounds in September 2026. The Verge reported the results on Thursday, November 12.
Podqi counted a site as "risky" if it was one of three things: a fake site pretending to be a real store or selling knockoffs, an old brand web address that someone else had registered again, or a dead site that anyone could buy and reuse.
| Chatbot | Answers that showed a risky site |
|---|---|
| Perplexity | 15% |
| ChatGPT | 8% |
| Gemini | 4% |
ChatGPT was the best of the three at warning people about a risky site it showed, doing so 24% of the time. In one answer, Perplexity named the wrong page as clothing brand Stussy's official site.
Why are small brands hit harder?
Questions about bigger companies turned up risky sites about half as often as questions about smaller brands. That makes sense. Big brands have lots of well-known pages for a chatbot to find, while a small label may have one site and a few mentions, so a convincing fake has less competition.
If you sell online yourself, that's a reason to check how chatbots describe you. Ask ChatGPT, Gemini and Perplexity where to buy your products and see which links come back.
How much should I trust this study?
Treat it as a warning, not a scorecard. Podqi wrote its own questions instead of studying real shoppers' chats, and it sells brand protection, so it has a reason to find problems. The Verge piece includes responses from Google, OpenAI and Perplexity. Still, all three chatbots slipped at least some of the time, and that matches what the rest of the web already teaches: a link is only as good as the place it points.
X reactions
SEO consultant Glenn Gabe wasn't surprised, and another reader turned the study into a checklist:
Shocking (sarcasm) -> AI chatbots are recommending fake retail websites, study finds
— Glenn Gabe (@glenngabe) November 12, 2026
Gemini was better than ChatGPT showing risky sites in 4% of answers. Perplexity performed the worst showing risky sites in 15% of answers. And ChatGPT was in the middle at 8%.
"Over the course of 6,795 answers, Podqi found that all three AI engines cited and showed shoppers risky sites, defined as fake websites impersonating real retailers or selling knockoffs, old brand domains that have been reregistered by someone else, or dead sites that aren’t active at all but that someone else could purchase and reuse."
https://t.co/vsSvUB89vG
A study by IP firm Podqi found ChatGPT, Gemini and Perplexity can point shoppers to fake retail sites. So when a bot names a shop, do not tap the link it gives. Type the store name yourself, match the address, then check the contact page and return policy before you pay.
— Babátọ́pẹ́ Àjẹpẹ́ (@babatopeajepe) November 12, 2026
Our take
Holiday shopping is weeks away, and a lot of people will ask a chatbot where to buy gifts this year. That's fine for ideas. It's a bad idea for the final click, because these tools pull links from the open web, fakes included, and present them with the same calm confidence as the real thing.
The fix on the chatbot side is simple to describe and hard to build: check whether a store is real before recommending it, and say so when it can't tell. ChatGPT's 24% warning rate shows it's possible, and also how far there is to go. Until the rates drop, use AI to decide what to buy and use your own eyes to decide where.
— Artificially Nerdy
How to shop safely with AI
- Don't pay through a chatbot link. Type the store's address yourself, or start from the brand's official site. Letting an agent buy for you? Read our checklist before an AI agent buys for you.
- Check the address letter by letter. Fakes often add a word or swap the ending, like .shop instead of .com.
- Look for a real contact page and return policy before entering card details.
- Pay with a credit card rather than a debit card or bank transfer, so you can dispute a bad charge.
- Run a small brand? Ask the three chatbots where to buy your products and report any fake sites they show.
FAQ
Which AI chatbot was safest for shopping in the study?
Gemini showed risky sites least often, in 4% of answers. ChatGPT was most likely to warn you when it did show one.
What happens if I buy from a fake site a chatbot suggested?
The same as any fake store: you may get a knockoff, nothing at all, or a stolen card number. Contact your card issuer quickly to dispute the charge.
Does this mean AI shopping assistants are unsafe?
They're useful for comparing products and prices. The risk sits in the links, so check the store yourself before you buy.
Was this a study of real users' chats?
No. Podqi wrote its own shopping questions about 151 brands and recorded the answers in September 2026.
Sources
- Podqi: AI search brand risk research — Method, brands, answers and risky-site definitions.
- The Verge: Gemini, ChatGPT and Perplexity shopping study finds risky fake websites — November 12, 2026. Rates by chatbot, 24% warning rate, small vs. large brands, Stussy example, company responses.
- Glenn Gabe on X — November 12, 2026, 12:50pm ET.
- Babátọ́pẹ́ Àjẹpẹ́ on X — November 12, 2026, 12:17pm ET.
