7 reasons to think twice about using ChatGPT, or any AI chatbot, for purchasing decisions

So you want to buy a thing? Congratulations! Only, you’re not sure which thing. Oh no. You could spend time researching what’s best phone / tablet / watch (or whatever you’re after) for you. But these days, surely ChatGPT or some other AI chatbot can do all the hard purchasing decisions on your behalf?

“Tell me the best thing to buy”, you might type. And the AI will merrily churn out perfect recommendations by the time you’ve had a sip of coffee, right? Well, no.

In fact, here are 7 reasons to think twice before relying on an AI chatbot when making purchasing decisions.

1. Chatbots aren’t human

Duh, obviously. But this is important. A chatbot will draw from a range of sources and mash its findings up into a succinct reply. What it can never have is actual firsthand experience of using products.

In other words, ChatGPT and its ilk lack the fundamental foundation of reviews since Ug the Caveman scrawled on a wall that his latest stick was a bit rubbish.

2. Chatbots get things wrong

I mean, reviewers may get things wrong too. But screw-ups are embedded deep into the core of chatbots. They churn out ‘hallucinations’ – a nice word for factual errors. Chatbot output is based on probabilities. That means incorrect specs, features and availability may be injected into responses – a problem, given that such details can change at speed and drive purchasing decisions.

3. Chatbot sources aren’t equal

I’ve tested ChatGPT and other chatbots for purchasing decisions. One regurgitated opinions of random YouTubers as the most authoritative recommendations imaginable. Others obfuscated sources entirely.

Ultimately, you never know whether you’re getting an average of online reviews, regurgitated marketing tripe, coordinated hype, or the result of a feedback loop where LLM recommendations are fed back into LLM training data, until the only possible recommendation will be “buy more AI”.

4. Chatbots don’t know you

You may have had conversations with ChatGPT, but it doesn’t know you. Not really. It can ask you questions but can never understand what you truly prize and will care about months down the line.

Often, it’ll plump for consensus, which is fine if you want the most popular smartphone of the hour, but not so much if a niche product would suit you better.

5. Chatbots are too eager to please

An under-discussed aspect of chatbots is how they’re designed for engagement. They want you to use them for everything and forget the rest of the world exists.

Replies are presented confidently, no matter their accuracy. You’ll rarely be challenged. Chatbots want you to think you’re a genius rather than risk making you feel bad by arguing against a purchasing decision.

6. The world is burning

Let’s not kid ourselves: all technology use has an impact on resources and the environment. But there’s a world of difference between streaming a song and making a single AI query.

Chatbot training and searches burn through resources at a staggering pace, contributing to water, energy and component scarcity. At scale, that’s a Very Bad Thing™ compared to lighter tools and local reasoning.

7. Chatbots stop you thinking for yourself

There is increasing – worrying – evidence that suggests over-reliance on chatbots erodes independent thinking. Outsourcing basic decisions may reduce your ability to make them and certainly reduces curiosity.

Again, this doesn’t mean avoiding AI chatbots like ChatGPT during every purchasing decision. But for anything important, you’re probably best using your own head and authoritative sources you trust, before the former forgets how and the latter are crushed by the remorseless advance of AI, which seems increasingly determined to insert itself between you and the open web.

• Now read: AI is eating the tech world – and your favourite gadgets are next