ChatGPT, Gemini Popular for Holiday Shopping but Unreliable and Risky, German Survey Finds

One in five consumers is now turning to artificial intelligence to hunt for holiday bargains, yet the technology is frequently failing to deliver on accuracy and security.

According to an October 2026 survey by the price comparison portal Idealo, shoppers are increasingly integrating tools like ChatGPT and Gemini into their seasonal routines, t3n.de reported.

The shift is accelerating. Broader data from Yougov in October indicates that half of German shoppers now plan to use AI for pre-Christmas deal hunting, with 42 percent of those users specifically tasking the software with finding the lowest prices.

While models like ChatGPT have become adept at product consultations—such as identifying a quiet, sub-400-euro vacuum cleaner effective against pet hair—they remain fallible. Tests show these tools often lack real-time data, frequently failing to account for shipping costs or current product availability.

Security risks are equally pressing. An analysis by the brand protection firm Podqi found that AI systems occasionally steer users toward unreliable or fraudulent websites. In 15 percent of responses from Perplexity, 8 percent from ChatGPT, and 4 percent from Gemini, the systems suggested problematic platforms.

“Immerhin in 57 Prozent der Fälle wurden diese ausdrücklich als Einkaufsmöglichkeit empfohlen, in weiteren 34 Prozent wurden sie ohne Warnhinweis erwähnt,” according to the Podqi study cited by t3n.de.

Brand protection firm Podqi, which scans more than 1,500 marketplaces, apps and social platforms for impersonating websites and files verified takedowns with registrars and hosts, conducted its analysis in September 2026 across 6,795 answers to where-to-buy questions about 151 brands.

The study tested Perplexity, ChatGPT and Gemini with identical prompts on the same days, measuring how often each engine surfaced risky domains and whether it recommended them, mentioned them without comment or warned shoppers away.

The data can be misleading even when the platforms are legitimate. During testing, the AI model Claude occasionally fixated on inflated list prices, ignoring lower, long-term street prices until prompted to re-evaluate. Even when the technology identifies a competitive price, it may direct users toward obscure vendors instead of well-known retailers offering the same deal.

Photo of author

Technology Desk

Technology Desk at Archyde.

Isaac Luechtefeld Kills Wife, Three Children, Then Himself in Plant City

Neuralink Trial Participant Plays GTA VI With Thoughts Via Brain Implant