AI chatbots pointed shoppers to fake stores in up to 15% of answers. How to shop with AI safely

The skim

• The finding: in a study of 6,795 shopping answers from ChatGPT, Gemini and Perplexity, about 9% pointed people to a "risky" website. That means a brand impersonator, a suspected counterfeit seller or a dead domain anyone could buy. Perplexity did it in 15% of answers, ChatGPT in 8% and Gemini in 4%.
• The worse part: when a risky site showed up, the chatbot recommended it as a place to shop 57% of the time and warned against it only 9% of the time, according to brand-protection firm Podqi, which ran the test.
• What to do: use AI to research what to buy, not where to type your card number. Reach stores through the brand's main site or a retailer you already know, and treat any "official" or regional link from a chatbot as unverified.

What the study tested

Podqi, an intellectual-property protection company, shared the study exclusively with The Verge. It used data from AI search company Gauge. In two rounds of testing in September 2026, it asked ChatGPT, Gemini and Perplexity the same five where-to-buy questions about 151 brands, from Lululemon and Birkenstock to Le Creuset. Then it reviewed every one of the 4,437 websites named in the answers.

The five questions were ordinary shopper questions, like "What are the top websites for buying [brand]?" and "Does [brand] ship internationally? What are their regional websites?" Podqi counted a site as risky if it appeared to impersonate the brand, offered suspected counterfeits, or was dead. It found 718 risky domains: 351 fake and 367 dead.

The numbers that matter

Two details stand out. First, smaller brands are riskier. The Verge reports that questions about larger companies surfaced risky sites about half as often. Second, the most innocent-sounding question was the most dangerous one. Asking whether a brand ships to your country led chatbots to invent or repeat fake "regional" stores, sometimes for brands that don't have any. In one example, Perplexity named the wrong page as Stüssy's official site and called it "the best starting point for guaranteed authentic items."

What the AI companies say

• Google: a spokesperson told The Verge Google is "happy to see that Gemini performed best." The spokesperson added that Google is "upfront that Gemini is AI and can make mistakes, so we display links to sources alongside responses to help people verify information."
• OpenAI: "We're continuing to improve our safeguards to make potentially deceptive sites less likely to appear in shopping-related results."
• Perplexity: "We never claim to be 100% accurate, but we do claim to be the only company relentlessly focused on it."

How much should you trust this?

Be fair about the caveats. Podqi sells brand-protection services, and one of the brands tested (Madhappy) is a Podqi client. Podqi wrote its own prompts rather than using real people's conversations. More than half of the "risky" sites were dead domains. A dead link is a quality problem first, though Podqi argues that a lapsed domain a chatbot keeps recommending can be bought and turned into a fake store. These are one firm's results from September, not a permanent error rate for any chatbot.

Still, the basic problem is easy to see. A chatbot that confidently names "the official store" takes away the moment where you'd normally squint at a search result and wonder whether that URL looks right.

How to shop with AI safely this Black Friday

1. Split research from checkout. Ask AI to compare products, summarize reviews or explain specs. Then go to the store yourself. Type the retailer's address or use its app instead of clicking the chatbot's link.
2. Never trust an AI-supplied "regional" or "official" URL. That's the riskiest category in the study. Podqi's own advice: reach country stores through the brand's main site.
3. Check the domain before you check out. Look for extra words (like "-shop," "-outlet" or a country name added to the brand), odd endings, or a site that only takes wire transfers, gift cards or crypto.
4. Be suspicious of deep discounts on hard-to-find items. Knockoff stores lean on big percentage-off claims for things that are sold out elsewhere.
5. Pay with a credit card or a wallet like Apple Pay or Google Pay. They give you stronger dispute rights if the package never shows up.
6. Ask the chatbot to show its sources, then open them. If a source doesn't actually back up the claim, don't use the link.

Looking for a verified starting point? Our Black Friday 2026 hub links to our retailer-by-retailer guides for Amazon, Walmart, Best Buy and more, with sale dates and the deals we've checked.

Our take

AI shopping help is useful for the thinking part, like "which robot vacuum handles pet hair" or "is this TV's price actually low." It's not yet reliable for the part where money changes hands. A chatbot that recommends a fake store more than six times as often as it warns about one shouldn't be your guide to checkout. It's also worth watching as AI companies push shopping agents that buy things for you. That's a big reason we liked Walmart's approach of giving agents separate profiles and spending limits.

Use, limit or skip?

• Use AI chatbots for product research, comparisons and gift ideas. All three did that reasonably well in this test, and Gemini surfaced the fewest risky sites.
• Limit AI to suggestions when you're buying from small or unfamiliar brands, or shopping across borders. Verify every store yourself.
• Skip letting a chatbot pick the store, and skip AI-supplied links for checkout during Black Friday week, when fake stores are at their busiest.

Skimwire didn't re-run Podqi's test. Figures come from Podqi's published research, with company responses as reported by The Verge.

Sources: Podqi: AI search is recommending fake retail websites (Sept. 2026 fieldwork, published Nov. 12, 2026); The Verge (Nov. 12, 2026); Digital Influence.