RealityHackerOpen in RealityHacker ⇢
Tools · Chatbots & Assistants · published 2026-11-12T00:00:00+00:00 · via The Verge

Study Reveals AI Chatbots Direct Shoppers to Risky Fake Retail Sites

Image via The Verge
Image via The Verge

A new study by IP protection firm Podqi found that AI chatbots like ChatGPT, Gemini, and Perplexity recommended risky or fraudulent websites in up to 15% of shopping-related queries. The test involved 151 brands, including Lululemon, Birkenstock, and Le Creuset, with Perplexity showing the highest rate of unsafe links. The findings highlight growing concerns about AI reliability when guiding consumers to online retailers.

Expanded Detail

A study by IP protection firm Podqi tested three major AI chatbots—ChatGPT, Google Gemini, and Perplexity—on 151 shopping-related queries involving well-known brands like Lululemon, Birkenstock, Le Creuset, and Stüssy. Over two testing periods in September 2026, the researchers analyzed 6,795 responses and found that up to 15% of answers from Perplexity included “risky” URLs—defined as fake websites impersonating real retailers, defunct domains repurposed by third parties, or sites selling counterfeit goods. Gemini showed risky links in 4% of responses, while ChatGPT did so in 8%. Notably, ChatGPT was the only one to issue warnings about such sites 24% of the time. The study revealed that queries about regional websites triggered more unsafe results than other prompts, and smaller brands were disproportionately affected, with risky links appearing half as often for larger companies. This raises concerns about AI’s reliability in guiding consumers during high-stakes shopping periods like the upcoming holiday season.

The findings come amid growing scrutiny of AI systems’ accuracy and safety in real-world applications. While companies like Google and OpenAI market their chatbots as helpful tools for product research and discovery, this study exposes a critical flaw: AI may inadvertently steer users toward fraudulent or deceptive sites. Podqi CEO Trevor West emphasized the need for consumers to verify sources before sharing personal or financial data. Google and OpenAI acknowledged the limitations of their models, with Google highlighting Gemini’s relatively strong performance and commitment to source transparency, while OpenAI stated it is actively improving safeguards against deceptive content. Perplexity defended its accuracy in research benchmarks but conceded that no system is perfect. What remains unresolved is whether regulatory bodies will step in to enforce accountability for AI-generated recommendations, especially as consumer reliance on these tools grows.

Context

The study underscores a broader societal risk: as consumers increasingly rely on AI to navigate complex online markets, the potential for fraud and misinformation escalates—especially during peak shopping seasons. Vulnerable users, particularly those unfamiliar with brands or new to e-commerce, may be misled by seemingly authoritative AI responses, leading to financial loss or data breaches. This could erode trust in digital retail ecosystems and harm both established brands and smaller businesses that depend on authentic online presence. The issue mirrors earlier concerns around search engine manipulation and clickbait ads, where algorithmic recommendations led users astray—though AI’s generative nature makes the problem more insidious, as it fabricates plausible but false information. If left unaddressed, such inaccuracies could prompt regulatory scrutiny or push platforms to adopt stricter content verification protocols. What may emerge next is a demand for standardized transparency labels on AI-generated shopping advice, similar to those used in advertising, and greater consumer education about verifying digital sources.

Expanded detail and Context are AI-generated analysis; the linked article remains the authoritative source.
Read the full article at The Verge →
This summary is Al-enhanced to contain extended analysis and broader social context. The original is {NAME); the linked article is the authoritative source. Original headline: “AI chatbots are recommending fake retail websites, study finds.” Browse more stories.