The tech industry's newest shopping assistants have a serious problem - they're stuck in the past. OpenAI, Google, Microsoft, and Perplexity all rolled out AI shopping features this month, but real-world testing reveals they consistently recommend outdated products over current models.
The holiday shopping season just got a lot more complicated. OpenAI's ChatGPT Shopping Research, Google's Gemini calling service, Microsoft's Copilot price tracking, and Perplexity's product search all promise to revolutionize how we shop. But a month-long test of these systems reveals a fundamental flaw that could leave consumers with last year's tech.
When journalist Stevie Bonifield asked four different AI bots to find an Android smartwatch for her Nothing CMF Phone 1, the results were telling. ChatGPT's most thorough Shopping Research feature recommended the Garmin Vivoactive 5 - a solid watch from 2023, but not the current Vivoactive 6 that offers more storage, improved GPS, and new features like Smart Alarm.
The pattern repeated across platforms. Google's Gemini suggested the Pixel Watch 2 from 2023 over the newer Pixel Watch 4, despite significant improvements in battery life, charging, and processing power. Perplexity mixed current recommendations like the Pixel Watch 4 with the ancient Samsung Galaxy Watch 4 from 2021.
"After waiting 15 minutes, Google's AI emailed me to let me know that every store it called doesn't sell Garmin smartwatches," Bonifield wrote about Google's "Call for me" feature, which promises to contact local retailers. The feature represents ambitious AI automation but fails at basic execution.
Microsoft's Copilot showed the most promise, immediately suggesting the CMF Watch Pro 2 designed specifically for the CMF Phone 1. But even Copilot missed the more recent CMF Watch Pro 3. Its shopping sidebar proved most useful, offering price history, aggregated Amazon reviews, and price tracking alerts.
The data issue becomes clear when you dig deeper. These AI systems appear to be training on product information that's 18-24 months old, yet they present recommendations with the confidence of current market analysis. For consumers who don't know what they're looking for - the exact audience these tools target - this creates a perfect storm for outdated purchases.
