Why ChatGPT is a terrible personal shopper.

The critique surfaced on 22 March, and it targets something most people have already stumbled into without quite naming it: ChatGPT is a convincing personal shopper in the same way that a confident stranger giving you directions is convincing — right up until you end up somewhere completely wrong. The structural problem is that large language models aren't connected to live inventory, real-time pricing, or the mundane truth that products get discontinued, go out of stock, or simply never existed in the form that a 2022 product description implied. Ask for a waterproof jacket under £150 and you'll receive a list delivered with the quiet authority of someone who has done the research. They haven't. The model has pattern-matched against text that described products at a moment in time that has long since passed. The prices are approximate. Some items are ghosts. The model doesn't know this, and more importantly, it doesn't tell you it doesn't know this. That combination — confident tone, stale data, no disclosure — is where the shopping advice quietly falls apart.

If confirmed, what this means is less about ChatGPT specifically and more about a broader misapplication of AI tools that users are already making at scale. Retailers who assumed that AI assistants might eat into their traffic from Google Shopping may be less exposed than feared — not because people aren't asking AI for product recommendations, but because the conversion rate on hallucinated inventory is predictably terrible. For consumers, the real cost is the gap between the recommendation and the reality: time spent tracking down products that don't exist at the price or specification described. For OpenAI and competitors, this is a product design problem that plugins and browsing integrations were supposed to solve, and haven't fully solved — the underlying model still responds to shopping queries with the same confident register it uses for questions that actually lie within its knowledge. The fix isn't just real-time data access; it's calibrated uncertainty, which remains genuinely hard.

Watch for any data from e-commerce platforms on AI-referred traffic and its conversion rates — that would either substantiate or complicate this critique with numbers that matter.

NewsHive monitors these sources continuously. All signal titles above link to the original reporting.

Intelligence by NewsHive. Need help navigating what this means for your business? Contact GeekyBee →