Artificial intelligence has become a standard tool for holiday shopping. It compares prices, aggregates gift ideas, and in some cases completes purchases autonomously. What it does not do, reliably, is vet the seller on the other side of the transaction. That gap between convenience and due diligence is where losses accumulate.
The scale of the problem is quantifiable. A Norton survey of 1,000 U.S. adults, conducted by Dynata between Sept. 10 and 21 on behalf of Norton’s parent company Gen, found that 16% of respondents had been scammed while shopping online during the holidays. Among those victims, 88% lost money, with an average loss of $394. For a household budget, that is a material hit. For the aggregate economy, it represents billions in misdirected consumer spending during a single season.
The mechanics of the threat have changed with the technology. Leyla Bilge, global head of scam research at Norton, notes that AI’s largest current impact is on coding. Fraudsters can now generate functional e-commerce sites at a pace previously impossible — registering thousands of fake online shops, complete with convincing product photography and fabricated reviews. The practical consequence for consumers is simple probability: the more fraudulent storefronts in circulation, the more likely any given shopper encounters one, particularly through AI agents optimized to surface the lowest price regardless of the seller’s credibility.
The financial loss is only the first-order cost. A survey of 3,110 U.S. adults released in June by the Global Anti-Scam Alliance found 82% of respondents had encountered a scam within the past year. Fifty-nine percent said it affected their mental well-being, and more than half reported being victimized repeatedly. From a risk management perspective, the repeat-victimization figure deserves particular attention. If personal data — an address, a phone number — is entered into a fraudulent AI-recommended site, that information circulates. Bilge notes victims may later face follow-up attacks that are more sophisticated, more targeted, and more personalized. A single $394 loss can become the entry point for a far costlier identity-fraud event.
Among consumers who have used an AI shopping assistant or agent, 56% reported that the tool recommended or completed a purchase from a retailer they did not recognize, per the Norton data. That is the structural weakness. An agent built to minimize price will rationally favor obscure, unvetted sellers — including fraudulent ones — because they undercut legitimate operators who carry real overhead. A low price from an unknown merchant is not a bargain; it is an unpriced risk.
The corrective is straightforward and should be treated as non-negotiable, in the same way a disciplined investor never buys a security without reading the filings. Verify the retailer directly, as Matt Schulz, chief consumer finance analyst at LendingTree, puts it. The AI recommendation is a starting point for research, not a completed due-diligence process.
The behavioral data suggests consumers know better and proceed anyway. The Norton survey found 35% of Americans had purchased from a website or seller they suspected was a scam because the deal seemed too good to pass up. In my years reviewing corporate financials, I saw the same pattern internally: pressure to act on an attractive number overrides process. Price-driven urgency, whether in a boardroom or a checkout cart, is where discipline fails.
Before completing any purchase from an unfamiliar seller, apply a structured checklist:
- Navigate directly to the retailer’s official website or app rather than relying on a link from an AI tool or social media advertisement. Control the entry point.
- Search the seller’s name paired with terms like “complaint” or “scam.” Reputation data is cheap to gather and disproportionately valuable.
- Compare reviews across multiple platforms. The Federal Trade Commission cautions that reviews and ratings can be fake or misleading — volume and star averages are not evidence of legitimacy.
- Do not treat “https” as validation. Encryption confirms the connection is secure, not that the seller is honest. Conflating the two is a common and costly error.
- Impose a mandatory delay. Schulz recommends the 24-hour rule: place the item in the cart, step away, and reassess the following day. The pause defuses urgency and creates space for the verification steps above.
Payment structure matters as well. When possible, transact by credit card, which offers dispute rights the FTC confirms in cases such as merchandise that never arrives. Debit cards, wire transfers, and peer-to-peer payment apps generally lack comparable recourse. Retain order confirmations and receipts — documentation is what converts a dispute from an allegation into a claim.
AI shopping tools are legitimate instruments with real utility in narrowing options and optimizing price. But the efficiency they deliver shifts more of the verification burden onto the consumer, precisely at a moment when fraudulent sellers are easier to create than ever. Treat every unfamiliar merchant the way you would treat an unvetted counterparty in any other financial transaction: verify first, pay second.

