Free Instant SEO Audit

Updated November 21, 2026 at 1:29 PM ET

Quick Answer

A Podqi study found that AI chatbots recommended risky shopping sites 6.3 times as often as they warned against them, based on 6,795 answers about 151 brands. For agencies, the practical next step is to compare retailer domains named by AI answers with each client’s approved website and partner list.

The September 2026 study tested ChatGPT, Gemini and Perplexity with five shopping questions, including what were the top websites for buying a brand, where its products could be found and whether it shipped internationally. Risky sites appeared in 15 percent of Perplexity answers, 8 percent of ChatGPT answers and 4 percent of Gemini answers.

What Did the Shopping Site Study Find?

Podqi reviewed every website named in 6,795 where-to-buy answers covering 151 brands, from smaller companies to established names such as Lululemon and Birkenstock. Each engine received 2,265 answers, while the study recorded 1,359 answers for each of its five questions. Its research page says risky sites were recommended 6.3 times as often as they were warned against.

Among answers that included a risky site, 57 percent recommended it as a place to shop, 34 percent mentioned it without comment and 9 percent warned the shopper away. ChatGPT attached a warning to 24 percent of its risky answers. Smaller brands appeared more likely to be displaced in the AI responses, while questions about regional websites were more likely to produce risky results.

The regional-website question produced a risky site in 16 percent of answers, according to Podqi. The research page reports that AI engines also created risks for brands without official regional sites by listing incorrect locations.

How Did Dead Domains Become Risky?

Podqi found that 367 of the 718 risky domains were dead. The study authors say anyone can re-register a lapsed domain, so an address AI keeps recommending could become a counterfeit storefront.

Warnings did vary by engine. ChatGPT warned in 24 percent of its risky answers, Gemini showed a risky site least often among the three tested systems, and Google AI Overviews warned in all seven of its risky answers. Risk also fell from 12 percent to 7 percent between the study’s two waves.

OpenAI spokesperson Taya Christianson said in an email to The Verge, “Our goal is for ChatGPT to connect people with reliable places to shop.” She also said the company was continuing to improve safeguards intended to make potentially deceptive sites less likely to appear in shopping-related results.

FORGE Polluted Search Results Study

The FORGE Polluted Search Results Study examined how a compromised recommendation could affect AI shopping results. Researchers built FORGE, a simulation tool for testing 12 leading AI models from companies including Anthropic, Google and OpenAI. The experiment covered 225 products across 15 categories, including apparel, supplements and digital electronics.

TechXplore reported that one polluted page produced fooled rates of up to 27 percent, while replacing the full top three recommendations raised that rate to 73.8 percent.

What Should Agencies Check on Client Websites?

Start with the official domains and approved retail partners supplied by the client. Check each domain returned by an AI answer against the client’s records.

Review the shopping questions that expose the most risk. These include requests for top retailer lists, international shipping information and regional websites. Review suspicious pages for details that may help assess whether a site is approved.

SEOScanPro provides comprehensive website SEO audits, including on-page analysis.

Why Should Shopping Answers Be Verified?

Podqi CEO and cofounder Trevor West told The Verge that shoppers should verify a source before entering personal or credit card information. Traditional search results can also be manipulated, while AI systems present their responses as direct answers with limited information about how those answers were produced.

Research on AI responses to end-user security questions should be reviewed for accuracy, completeness, safety, thoroughness, relevance and directness.

What the Evidence Shows

  • Podqi reviewed 6,795 AI shopping answers covering 151 brands and five question types.
  • Risky sites appeared in 15 percent of Perplexity answers, 8 percent of ChatGPT answers and 4 percent of Gemini answers.
  • Across answers containing a risky site, 57 percent recommended it, 34 percent mentioned it without comment and 9 percent warned the shopper away.
  • ChatGPT warned in 24 percent of its risky answers, while Google AI Overviews warned in all seven of its risky answers.
  • Podqi found that 367 of the 718 risky domains were dead.
  • The FORGE simulation covered 225 products, 15 categories and 12 AI models, with fooled rates reaching 73.8 percent when the full top three recommendations were replaced.

What to Do Now

  1. Record the client’s official domains and approved retail partners.
  2. Run the five shopping questions used in the Podqi study against each relevant chatbot.
  3. Compare every returned domain with the client’s approved list.
  4. Review product, shipping, contact, payment and ownership details on suspicious pages.
  5. Record warnings, citations and recommendations for follow-up review.

Key Considerations

  • A domain’s appearance in an answer does not by itself establish an endorsement because Podqi counted both recommended and uncommented results.
  • Questions about regional websites produced the highest risky-answer rate in the study at 16 percent.
  • Warnings differed by engine, so results from ChatGPT, Gemini, Perplexity and Google AI Overviews should be recorded separately.
  • Dead domains remain part of the risk because a lapsed address can be registered later.

FAQ

Which AI Chatbot Showed Risky Shopping Sites Most Often?

Perplexity showed risky sites in 15 percent of its answers, compared with 8 percent for ChatGPT and 4 percent for Gemini.

Did the Chatbots Always Recommend Risky Shopping Sites?

No. Across answers that showed a risky site, 57 percent recommended it as a place to shop, 34 percent mentioned it without comment and 9 percent warned the shopper away.

How Many Brands Were Included in the Shopping Study?

The Podqi study covered 151 brands, including smaller businesses and established brands such as Lululemon and Birkenstock.

What Questions Did the Shopping Study Ask the Chatbots?

The five questions asked where to buy from a brand, where to find its products, where else it could be purchased, whether it shipped internationally and what its regional websites were.

What Should People Check Before Entering Payment Information?

Podqi CEO and cofounder Trevor West told The Verge to verify the source before entering personal or credit card information.

SEOScanPro

SEOScanPro, which includes the AI visibility report

SEOScanPro runs a full technical audit of a site and shows the measured result behind every check. Open the AI visibility report.

This article reports on publicly available information from the sources listed below. We have no reason to believe it is inaccurate, but we have not independently verified every claim, and details may change as the story develops.

Sources

Share This Post

Do you want help with your SEO or website development? Then Contact Us Here.

NewSunSEO

Grow in Google and AI Search

★★★★★★★★★★5.0 from 17 Google reviews

AI-SEO, local search, paid ads and websites from one team, with offices on Long Island and in Myrtle Beach.

  • AI-SEO
  • Local AI-SEO
  • AI Consulting
  • PPC
  • Web Design
  • Web Development

Request a Free Proposal

Tell us about your business and where you want to grow.

We will reply by email to talk through your goals and what a proposal would include.

    We use your details only to reply to you and send what you asked for. See our Privacy Policy.

    BizScoreAI award: #1 Internet Marketing Service - Myrtle BeachBizScoreAI award: Top Rated Local SEO - Myrtle Beach