Artificial intelligence (AI) platforms are rapidly integrating shopping features, but a recent test reveals these assistants often recommend outdated products. Despite advances from OpenAI, Google, Perplexity, and Microsoft, AI-powered shopping tools struggle to prioritize the latest models, sometimes pushing consumers toward older tech.
The Test: Searching for a Smartwatch
To evaluate these new features, a search was conducted across ChatGPT, Gemini, Perplexity, and Copilot for a suitable Android smartwatch for a Nothing CMF Phone 1. The results showed consistent inaccuracies: all four AI models frequently suggested products from 2022 and 2023, despite newer versions being available. For example, ChatGPT recommended the Garmin Vivoactive 5 instead of the more recent Vivoactive 6, omitting key improvements like increased storage and enhanced GPS.
Inconsistent Accuracy & Odd Suggestions
- ChatGPT provided detailed comparisons but still favored older models. It took 10 minutes to compile a list, including the Fitbit Versa 4 and Google Pixel Watch 3 alongside more current options.
- Gemini struggled to identify current stock. Its “Call for me” feature, designed to check local availability, failed entirely, reporting that no nearby stores carried Garmin smartwatches after a 15-minute wait.
- Perplexity was the most unpredictable, mixing recent models like the Pixel Watch 4 with the Samsung Galaxy Watch 4 (from 2021). It also displayed irrelevant products, including cheap, off-brand watches and even a mobile phone in its recommendations.
- Copilot was the closest to accuracy, immediately suggesting the CMF Watch Pro 2 (designed for the CMF Phone 1) but still overlooked the newer Pro 3. It offered a useful price history and review summaries.
Why This Matters
The persistence of outdated recommendations isn’t just a minor inconvenience; it highlights the challenges of real-time data integration in AI shopping. The algorithms appear to rely on product information that lags behind current releases, potentially leading consumers to make suboptimal purchases. This is particularly problematic given the rapid pace of tech development, where even a year-old model can lack critical features or performance upgrades.
The Bottom Line
While AI shopping assistants show promise—especially in price tracking and feature comparisons—they are not yet reliable enough to replace human-curated buying guides. The current tendency to prioritize older products makes these tools more likely to hinder informed decision-making than facilitate it. For now, manual research remains the more effective approach.
