[swores]

> "more than 6 in 10 people declare they already use AI for mental health questions. 42% of them almost always follow the advice it gives them."

If these numbers were "of people who regularly use GenAI chat tools" then I'd be surprised it was quite so high already, but not shocked and would find it completely believable.

But this seems to be "of all people surveyed", which I'm rather skeptical of - unless their sample was very biased (as an extreme example, if they recruited people to the survey only by linking to it in ChatGPT ads, but there are plenty of less extreme ways to get a sample group that's way more likely to use AI than a genuinely random sample of the whole population).

It's also worth noting (and perhaps somewhat explains numbers seeming so unrealistically high to me) that, unless I've misunderstood, "turn to AI for psychological support" isn't necessarily "using AI as a therapist", it could be uses as minor as asking "Can exercise help with my depression?" or "If I think I am having a nervous breakdown, should I talk to a doctor?"

[mgh2]

Yes, this seems unusually and suspiciously high. Given that Axa is an insurance and investment company, may have some conflict of interest.

[jubilanti]

This is almost certainly from an online panel survey of people who are paid to take surveys, which is a mini-industry in itself, people try to make a living taking as many as they can for a few cents. The people who fill these out are completely unrepresentative of the general population and quite a few are probably having their openclaw fill it out for them.

[Nuzzerino]

Not sure if you saw this, but it says at the footer:

> 1.Methodology: The study was carried out in collaboration with IPSOS among 19,000 adults aged 18 to 75 across 18 countries, between 12 January and 16 February 2026.

Survey methods: https://www.ipsos.com/en-uk/survey-methods-ipsos-uk

If the Ipsos stock price is any indicator of surveys still being viable in 2026, then I can only wonder what actually motivates comments like this.

That being said, if the actual study cannot be reproduced, it's not science. So I'm looking forward to seeing the method and not the conclusion scrutinized. There's a lot of cognitive dissonance right now, especially in the US, when it comes to perceptions about mental health.

[jubilanti]

Did you get that link from an LLM? Did you actually read it? That is not the methodology page for this survey. That is a general article on the kinds of survey methods you can contract with Ipsos to run.

[Nuzzerino]

No, yes, I know, and it doesn’t affect my point. Thanks for your concern, HN!

[swores]

That definitely fits with how I felt about the numbers they reported