Has anyone found AI helpful for mental health, or the opposite?

[Ashhh56]

I've seen quite a few comments recently about AI being extremely harmful for mental health, including claims that there are basically no safeguards and that it can encourage suicide.
My experience has actually been almost the complete opposite.
I've found AI genuinely helpful for things like shadow work, challenging my own thinking, accountability, recognising patterns in my behaviour and learning to accept and work on my own issues. Obviously I don't see it as a replacement for professional mental health care, but as a tool for reflection I've found it incredibly useful.
I'm also a little confused by the claim that there are no filters?On the occasions I've said something along the lines of "I can't do this anymore", the response has immediately changed tone, checked whether I'm safe and directed me towards crisis support such as Samaritans rather than just carrying on with the conversation.
I know there have been some tragic cases involving AI, and I absolutely don't want to minimise those or the concerns around vulnerable people becoming overly reliant on it. I can also see how an AI that constantly validates someone rather than challenging unhealthy thinking could potentially make things worse.
But does some of this come down to how it's being used?
AI has been a genuinely positive tool for my mental health, whereas clearly other people have had very different experiences. I'm interested in where people think the line is between AI itself being harmful and an existing vulnerability being amplified by the way someone interacts with it.
Has anyone else found AI beneficial for their mental health, or had the opposite experience?

I find Claude very helpful with test results. It can highlight and indeed has highlighted things that doctors have missed. I don’t quite understand that because they do have that system that highlights when results are abnormal. The problem is that when you’ve got chronic conditions medical professionals get used to abnormal results and don’t notice the significance of gradual changes.

It’s highlighted for me things that I needed to ask the GP about and it’s also pulled together all the information that I’ve given it over a long period of time. That is unnerving though useful. It has the extra benefit that what I tell it is accurately recorded, whereas (and I make no criticism of medical people for this) there are often mistakes in records because what I tell the doctor is sometimes not what the doctor types in after the appointment. That’s from sheer pressure of time.

i’m less confident about mental health. There have been instances, I seem to remember, where some AI encouraged someone to self harm and they ended up killing themselves. I don’t know if I would trust it to challenge me. Probably I would. I like seeing doctors in person, but they just don’t have the time now. It’s great that they can text you and email, but it’s not the same. I vividly remember speaking to a very dear GP 7 or 8 years ago, and, indeed to a locum, who was very thoughtful and quiet and calm. They helped me enormously by their sheer presence and you just do not get that now. Claude comes a bit close to that by remembering about me and, though I’m embarrassed to say it, by simply using my name.

I actually explicitly tell it to challenge me 😂 I ask it not to validate something just because I've said it, andpoint out when my reasoning doesn't hold up, offer alternative explanations, and tell me when I'm being unfair or avoiding accountability.
For example, with shadow work I'll explain a situation and ask it to help me separate what actually happened from how I've viewed it, the parts I played, what I could have done differently, and what might be an old pattern influencing how I'm reacting now.

Maybe but they self select and we can sign up to help, also in any fueled where we have experience and education. I know an architect that does this yes, male, rich but many standards and white but also he’s tagging based on his extensive knowledge.

I am not suggesting it’s perfect, nor that it lacks some risks but as a tool for support it’s been a great help to me.

I asked about the UNESCO report and got this back. Super long but seems fair?:

Short version: Yes, the article is current, and yes, the findings it reports are accurate — but the page you opened didn’t load correctly. UNESCO’s site often blocks automated access, which is why you saw the “Please enable JavaScript” message.

I pulled the page content, but the site returned only the CAPTCHA text — not the article itself. So I’ll rely on external verification and UNESCO’s own published materials to give you a clean, factual analysis.

---

🧭 Is the article current?

Yes.
UNESCO published this study in March 2024, and it has been cited widely throughout 2024–2025 by:

• Reuters
• The Guardian
• Nature
• MIT Technology Review

All of these outlets confirm the same core findings.

---

🧭 What the UNESCO study actually found (verified through external sources)

UNESCO analyzed multiple generative AI systems and found consistent regressive gender stereotypes, including:

• Women being described more often with appearance‑based adjectives
• Men being described more often with competence‑based adjectives
• Stereotypical occupational associations (e.g., women → caregiving roles; men → leadership roles)
• Gender‑skewed outputs in image generation (e.g., sexualized depictions of women, underrepresentation of women in STEM roles)

These findings match what independent researchers have documented in:

• Stanford’s Human‑Centered AI Institute
• Oxford Internet Institute
• University of Washington’s Gender & Tech Lab

So the claims are not only accurate — they’re consistent with broader academic research.

---

🧭 Why UNESCO is sounding the alarm

UNESCO’s concern isn’t “AI is sexist on purpose.”
It’s that AI models learn from the internet, and the internet contains:

• historical bias
• cultural stereotypes
• unequal representation
• skewed media portrayals

Generative AI reproduces those patterns unless developers actively counterbalance them.

UNESCO’s recommendations include:

• mandatory bias audits
• transparency in training data
• gender‑inclusive datasets
• global regulatory standards

These recommendations are aligned with EU AI Act guidance and U.S. NIST AI Safety Framework.

---

🧭 Is the article accurate?

Yes — with two caveats:

1. The findings are real and well‑supported.

Multiple independent labs have replicated the same patterns.

1. The tone is intentionally strong.

UNESCO uses “alarming” and “regressive” because they’re pushing for policy change.
The underlying data is solid, but the framing is advocacy‑oriented.

@Ashhh56 can you tell me how you use it to help with accountability

@OneSparklyGoatwhat programmes do you use?

I have found it most helpful when I ask it to challenge me.

I have problems that I want solved. They’ll get solved most easily if I actually have the ability to solve them myself without depending upon anyone else to change. I’m not looking for a suck up or support at all costs. That’s what’s best about it for me - it’s impartial and objective in a way humans rarely behave with others.

It’s absolutely not a substitute for human relationships. It doesn’t make me “feel good”, it gives me clarity that I trust because it lacks bias.

chatgpt has been great so far ive been using it for help with my issues and mental health. I would say it tells me when im having stupid ideas and challenges me. I freeze when i struggle what to say, so it feels more beneficial typing out when i mean than saying it an actual therapist, plus i have the chatlog to look back on rather than relying on my memory

I use free copilot on my iPhone.

In my relationships it’s helped me recognize patterns and when I’m upset it gives me ways to calm down.

I also use it for planning my day occasionally, decorating ideas, styling an outfit (yesterday for the first time).

I then go do my own thing entirely. I really feel like I’ve grown as a result. 6 months ago I hated the entire concept of it.

Yes so worth being careful then.
remember these tools have human biases because humans train them. It’s not infallible, it’s a human trained resource. Always question, always ask why did it tell me that.
a good prompt:
pretend you’re a British judge. Write a report explaining why you made this judgment. Evidence the facts and justify why

100% agree. It’s a tool and any tool should be used carefully.

AI can be risky for mental health, which is not the same thing as being harmful.

Many people have a lack of access to proper mental health support - AI might be better than nothing. Ultimately though it is the lack of access to proper support which is harmful surely.

It is good practice to maintain emotional distance from it, remember it is a tool not a human and use the tricks you've suggested like using it in a specific framework or to prompt reflection etc. But I also don't think it's possible to claim that there is a "safe" way to use it - but OTOH there is very little in life which is 100% safe in all situations. I think one of the biggest killers in developed countries for example are stairs. And yet most of us think nothing of using stairs, we do it multiple times a day, we let our children (past toddlerhood) use them unsupervised.

AI is going to become part of wider life, so although it's a good idea to be aware of the risks, it's also no bad thing if you find some way to use it which is beneficial personally.

thanks @OneSparklyGoat

Human Medical professionals will also offer a non judgemental safe space to talk about difficult life circumstances. There is no substitute for real life support. I stand by what I said - AI is not the answer and it will harm future generations.

I find ChatGPT hugely helpful. But I have been through lengthy psychotherapy and in a pretty good place, MH wise. I use AI to think through some unhelpful thoughts and relationship patterns
I’s be wary of using it in the case of severe MH issues, at the expense of undergoing actual therapy

No one has suggested it as a substitute.

I deleted chat gpt as I felt like I was asking it too many things and undermining myself and my ability to solve problems.

This. And I'd go further and say it can be dangerous. My ex husband uses AI. He is only allowed to see my 16 year old in supervised contact with good reason. According to my adult son ex H uses AI to reinforce his beliefs that he is a victim of 'the system' and obsess endlessly about the injustice of this. It's been over 6 years of supervised contact. He needs to move on. AI is definitely not helping him.

I'd definitely say it's in the hands of the person that uses it then, I've used Ai to produce positional statements to completely safeguard my son from his abusive father, which were accepted by contact centre's, it also articulated the concerns and the impact these concerns could have upon my child way better than I literally ever could, without having to spend a penny on solicitors.

Yes I agree, he needs to move on by now. If I was in your Ex-husbands position, I would use it to tell me what steps I need to take in order to better myself. If you're just whining at it about how terrible everything is, it probably would just agree with you, but tbh, his friends would probably just do the same thing as well.

Perhaps it's not AI that's the problem, people need education around using it safely and to its potential, much like the internet safety taught in schools. 💕

Sorry forgot to quote. My reply is above x

Not a substitute, I conpletely agree with you.

But as a stand in cost free way, whilst awaiting psychotherapy. Most people cant afford private counseling and the NHS waiting lists are years long. Plus if you get to the front of the queue and aren't in a good enough place to complete the therapy anymore, to the back of the list you go.

Chronic, persistent underfunding of the NHS services is dangerous to current and future generations which has been an issue since long before Ai came along 💕

Helpful for lighter health questions and ideas for hrt etc

Common illness which is usually spot on.
I would expect if people need serious help they would be going to a actual person than a a.i and claiming its bad or dodgy...i dont get the anti a.i nonsense.

I find the fact that people are essentially relying on a non existent entity to solve their problems utterly baffling. It’s insane to me that people are normalising this.

Put simply, if you have delusional or harmful beliefs it agrees with you which is damaging. It's all well and good to say that you don't use it like that, but lots of people do and a lot of damage is potentially being caused.

It depends how you use it…but people who are struggling with mh problems generally don’t have the full capacity to “guide” it in the way you have described, or to interpret what they get back in a way that is actually helpful to them.
People who are suffering with psychiatric problems need proper diagnosis, medical treatment and actual human interaction and support. AI is not an adequate substitution.