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Perplexity all rolled out AI shopping features this month, but real-world testing reveals they consistently recommend outdated products over current models.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who don't know what they're looking for - the exact audience these tools target - this creates a perfect storm for outdated purchases.
"If you don't know what you're looking for, it could be easy to end up with an outdated product, or even just miss out on newer options the AI doesn't realize exist," according to The Verge's analysis.
The timing couldn't be worse. Tech companies rushed these features to market for the critical holiday shopping period, but they're fundamentally unreliable for their core use case. Only when users specifically asked for "current" or "latest" models did the AI systems surface newer products - knowledge most casual shoppers wouldn't possess.
This reflects a broader challenge in AI development: the gap between training data and real-time information. While these systems excel at comparative analysis and feature explanations, they're working with stale product catalogs in a fast-moving consumer electronics market.
The implications extend beyond individual purchases. If millions of holiday shoppers trust AI recommendations, retailers could see artificial demand for discontinued models while current inventory sits unsold. Manufacturers investing in new features and improvements might find their innovations buried under AI systems promoting older alternatives.
ChatGPT's Shopping Research remains the most sophisticated, offering personalized questionnaires, product ratings, and detailed comparison charts. But sophistication doesn't matter if the underlying product database is outdated. Microsoft's Copilot shows promise with its comprehensive shopping sidebar features. Google's calling service represents ambitious automation that currently fails basic functionality tests.
The race to deploy AI shopping assistants has exposed a fundamental weakness in how these systems access and process current market data. Until tech companies solve this data freshness problem, human-written reviews and buying guides remain more reliable for finding the latest products.
For now, consumers should approach AI shopping recommendations with skepticism, especially for fast-moving categories like consumer electronics. The technology shows promise, but it's not ready to replace traditional product research methods.
The AI shopping assistant revolution has arrived with a critical flaw - they're recommending yesterday's products to today's shoppers. While features like ChatGPT's personalized research and Copilot's price tracking show genuine innovation, outdated product databases undermine their core value proposition. Until tech companies fix this data freshness problem, consumers are better off sticking with human-written buying guides and current product reviews. The holiday shopping season will be the real test of whether users notice and care about these limitations.
