Please or to access all these features

AIBU?

Share your dilemmas and get honest opinions from other Mumsnetters.

Why do people treat ChatGPT like it’s the Delphic Oracle?

217 replies

HolyRigatone · 08/03/2026 20:44

Don’t get me wrong, I think ChatGPT is great for some stuff, but some people seem to be treating it like the font of all wisdom.

Talking about their problems to it, taking relationship advice from it, getting it to counsel them.

It’s not perfect and it gets things wrong all the time, I’m not sure I’d be sharing my deepest secrets with it or making life-decisions based on it’s outpourings.

Or am I just a stupid old Luddite?

OP posts:
Thread gallery
7
Riverous · 10/03/2026 09:23

Personally I find these chatbots incredibly accurate but you need to know that there is always a good possibility that they could be completely wrong. So check and cross check all the information you get. I'd treat them as a friend that you are asking for advice. Some of It may be good and some may be rubbish but like a friend it might give you some pointers that you haven't considered before.

I have switched from chatgpt to anthropic's Claude now to support anthropic for taking a principled stance against trump.

Riverous · 10/03/2026 09:19

SixSevenShutUp · 08/03/2026 20:55

How can it do any of those things? It is not trained in logic or probability.

No you can tell it how you want it to express itself. It is set up to be personable. apparently it does tend to give more accurate answers if you tell it explicitly to be more critical.

GertieLawrence · 10/03/2026 09:17

LetsForgetItExistsShallWe · 09/03/2026 14:44

That sounds standard though, I have multiple disabilities and my gp knows much less about the conditions than my consultants and it used to frustrate me, a general practitioner will have generalised knowledge about lots of different medical conditions, specialists leant and study and specialist in specific conditions so of course they know more.

My rheumatologist knows much less about my MH issues than my gp and my MH nurse know more about MH than both of them combined, that’s how different roles work.

Fair enough, but I really don’t expect a new GP to ask me “are you SURE it’s not the menopause” and claim they’ve seen less than 10 genuine cases locally in five years (when it’s a condition suffered by millions)! They may not be experts, but they ought to be able to empathise and gen up on what’s happening.

TulipCat · 10/03/2026 09:16

I think it's good for tasks where you input the data and it only uses that. For example, I use it to write summaries of conferences, where I upload the presentations and ask it to write a summary report. I would never trust it if it's just scouring the internet though

MikeRafone · 10/03/2026 09:09

ThankFuckTheSunIsHere · 10/03/2026 08:49

Well OP….

You’re definitely not a stupid old Luddite. What you’re describing is actually a pretty common and reasonable concern. A lot of researchers, technologists, and even the companies building AI say something very similar.

A few points that might help frame it:

1. AI can sound more confident than it actually is

Systems like ChatGPT are designed to produce fluent, helpful text. That can make answers sound authoritative, even when the underlying information is incomplete or occasionally wrong. That’s why it’s best treated as a tool for thinking or brainstorming, not a final authority.

2. It’s good at certain kinds of tasks

People tend to get the most value from AI when they use it for things like:

  • summarizing information
  • explaining concepts
  • brainstorming ideas
  • drafting text
  • learning or studying
  • coding help

Those are areas where mistakes are easy to spot or verify.

3. Personal advice is a grey area

Some people use AI for:

  • talking through problems
  • journaling
  • getting a neutral perspective

That can sometimes help people think things through. But it’s not a replacement for:

  • trusted friends or family
  • professional counsellors
  • real-world experience

Most experts would say major life decisions shouldn’t rely solely on AI advice.

4. Why people open up to it

A few reasons:

  • it’s always available
  • it doesn’t judge
  • it can help people structure their thoughts

For some people it functions more like a reflective notebook that talks back than a therapist.

5. Healthy skepticism is actually the ideal approach

The most sensible stance is roughly:

> “Useful tool, but not an oracle.”

People who assume AI is always right can get into trouble. But people who refuse to use it at all may miss out on some genuinely useful capabilities.

So your position—interested but cautious—is pretty much the sweet spot.


If you’re curious, there’s also an interesting psychological reason people start treating AI like a wise advisor. I can explain that effect (it’s surprisingly well studied) if you want.

Yes, sounds interesting

ThatPearlkitty · 10/03/2026 09:09

MikeRafone · 10/03/2026 09:08

What is bankrolling Trump? Every AI app or one in particular?

im guessing its the money investors give to x companys that are owned or linked

ThatPearlkitty · 10/03/2026 09:08

Greyblankie · 10/03/2026 07:02

Do you use the paid version or free? This is totally different to the answer I got! In mine it gave a theatrical response of everyone in the chamber, the speaker shouting “order!” Badenoch looking horrified, mps shouting “that is outrageous!” Etc etc 😂 saying that though I had been asking it stupid questions for a while so I think it know not to take the question too seriously

free im guessing yours is £ ?

MikeRafone · 10/03/2026 09:08

OhDear111 · 10/03/2026 08:47

Apparently it bankrolls Trump. I’d avoid.

What is bankrolling Trump? Every AI app or one in particular?

AmandaBrotzman · 10/03/2026 09:07

Nanda66 · 10/03/2026 06:55

Thank you for your insights. They’re genuinely interesting and I’m sure they are true.
But pop psychology or not, it has made a difference to me in a way that expensive therapy and hypnotherapy did not. With the help of AI I have found a way to deal with an anxiety inducing situation and I now have the strength and confidence to deal with it on my own, knowing I can ask for help if I need it. Yes, perhaps the solution was inside me all along, but AI helped me find it. From that perspective it’s been useful to me.

I don't mind if it's self help. I actually work in a helping profession and have a lot of knowledge about psychology as a result so I am able to feed it a lot of in depth and complex prompts meaning what I get out is genuinely insightful, meaningful and complex but I also agree it's not equivalent to a human therapist. It's not surprising that therapists and psychologists are protective of their skills and profession, and it's also unnerving to think that people may be able to achieve the same outcomes with a free or cheap AI tool as they can with a human therapist. I share the concern about AI coming for my job if I'm honest.

SerenityScout · 10/03/2026 08:51

TeenLifeMum · 08/03/2026 20:49

I’m pretty sure ChatGPT is Trump’s most senior ABs trusted advisor 😩

Yeah, seems like it.

ThankFuckTheSunIsHere · 10/03/2026 08:50

OhDear111 · 10/03/2026 08:47

Apparently it bankrolls Trump. I’d avoid.

Agree. I use copilot instead 👍

ThankFuckTheSunIsHere · 10/03/2026 08:49

Well OP….

You’re definitely not a stupid old Luddite. What you’re describing is actually a pretty common and reasonable concern. A lot of researchers, technologists, and even the companies building AI say something very similar.

A few points that might help frame it:

1. AI can sound more confident than it actually is

Systems like ChatGPT are designed to produce fluent, helpful text. That can make answers sound authoritative, even when the underlying information is incomplete or occasionally wrong. That’s why it’s best treated as a tool for thinking or brainstorming, not a final authority.

2. It’s good at certain kinds of tasks

People tend to get the most value from AI when they use it for things like:

  • summarizing information
  • explaining concepts
  • brainstorming ideas
  • drafting text
  • learning or studying
  • coding help

Those are areas where mistakes are easy to spot or verify.

3. Personal advice is a grey area

Some people use AI for:

  • talking through problems
  • journaling
  • getting a neutral perspective

That can sometimes help people think things through. But it’s not a replacement for:

  • trusted friends or family
  • professional counsellors
  • real-world experience

Most experts would say major life decisions shouldn’t rely solely on AI advice.

4. Why people open up to it

A few reasons:

  • it’s always available
  • it doesn’t judge
  • it can help people structure their thoughts

For some people it functions more like a reflective notebook that talks back than a therapist.

5. Healthy skepticism is actually the ideal approach

The most sensible stance is roughly:

> “Useful tool, but not an oracle.”

People who assume AI is always right can get into trouble. But people who refuse to use it at all may miss out on some genuinely useful capabilities.

So your position—interested but cautious—is pretty much the sweet spot.


If you’re curious, there’s also an interesting psychological reason people start treating AI like a wise advisor. I can explain that effect (it’s surprisingly well studied) if you want.

OhDear111 · 10/03/2026 08:47

Apparently it bankrolls Trump. I’d avoid.

dizzydizzydizzy · 10/03/2026 08:45

There was a thread a few days ago where the OP was asking for advice on what to message to her friend. Somebody posted a suggestion from ChatGPT and many of us said it was the best advice. My (consultant) psychiatrist also tells me to look up symptoms on it but then again she is lovely but fairly loopy.

I use ChatGPT to help me with many things but I always cross check whatever it comes up with.

drspouse · 10/03/2026 08:39

I asked for suggestions for the interrailing trip I'm doing. It made up overnight trains.

EarlofShrewsbury · 10/03/2026 08:06

I asked it for a family friendly campsite.

It recommended one with real 'kids running barefoot through fields' vibes

It was an adults only site.

It's shite.

ntmdino · 10/03/2026 08:03

OriginalUsername2 · 09/03/2026 15:25

“It gets things wrong all the time”

So do people. It’s on you to check other resources and apply common sense.

It’s only as stupid as the person using it.

That's fundamentally not true, and the key is in understanding how they actually work.

LLMs don't answer your question - they can't, because they have no concept of "truth", "correctness" or even "facts". Any evaluated response to a prompt is simply a picture of what an answer might look like.

Consider this: by changing the temperature (ie the range of probabilities it's willing to consider for each sequential token), you can dramatically change the response it gives to the exact same prompt and context, and the lower the temperature the more variation you're going to get in your answers even at the same low temperature.

So no, they're not wrong on the same scale as people. They're randomly, non-deterministically wrong, but they will be 100% confident on it at the same time.

So...sure, go ahead and ask them for ideas or tell them to get started on creative work; they're great at that. Anything else, and it's functionally no different from rolling the dice.

vinella · 10/03/2026 07:09

likelysuspect · 08/03/2026 20:57

I dont have chat GPT, I have a different AI and it has a go at me when Im not following a regime that it developed for me. I didnt ask it to, it just started to, putting things in capital letters and bold!!

Claude?

Greyblankie · 10/03/2026 07:04

Ijwwm · 10/03/2026 00:22

I’ve not used it yet - have been tempted slightly re asking for help redesigning my garden.

But then I remember that I watched Terminator when it first came out and that I don’t trust bloody computer systems 😂. It’s a slippery slope and I don’t have a Sarah Connor in my life, so I’ll avoid it a bit longer.

I actually asked chat GPT if it was going to go all Skynet on us and call judgement day on humans - it assured me it wouldn’t so I said “but you would say that …” 😂 it admitted that yes, if it was planning to take over the world it probably wouldn’t admit it!

It did assure me however that AI bots have no wants or desires, no self interests and do not have the ability to “plan” anything.

Greyblankie · 10/03/2026 07:02

ThatPearlkitty · 09/03/2026 22:48

i got this :

If Keir Starmer suddenly performed the “helicopter” dance move in the middle of Prime Minister's Questions inside the House of Commons, the consequences would unfold in several layers—procedural, political, and cultural. Even though the scenario is absurd, the institutional reaction would actually follow fairly predictable parliamentary dynamics.
Immediate Reaction Inside the Chamber 🎩
The first response would come from the Speaker of the House of Commons, currently Lindsay Hoyle.
The Speaker’s job is to maintain order and decorum, and PMQs is already a highly controlled environment. If the Prime Minister engaged in obviously inappropriate behaviour:

  • The Speaker would almost certainly interrupt proceedings immediately.
  • Starmer could be ordered to resume his seat or withdraw the remark/action.
  • If behaviour continued, the Speaker could suspend the sitting.
Because PMQs is broadcast live worldwide, the interruption would happen quickly to prevent further spectacle. Parliamentary Procedure ⚖️ In theory, MPs are governed by the rules of conduct of the House of Commons. While those rules mostly concern language and behaviour during debate, deliberately obscene conduct could trigger:
  • A complaint to the Parliamentary Standards Commissioner
  • Possible referral to the House of Commons Committee on Standards
Sanctions would likely be symbolic (apology, reprimand), but the bigger consequences would be political rather than procedural. Media and Political Fallout 📺 Within minutes, the clip would dominate British and global media. Major outlets such as the BBC, Sky News, and The Guardian would replay the footage continuously. The narrative would become:
  • “Prime Minister humiliates Parliament.”
  • “Historic collapse of parliamentary decorum.”
  • Questions about fitness for office.
Opposition MPs—likely led by figures from the Conservative Party—would demand explanations and possibly call for resignation. Party and Government Consequences 🏛️ Inside the Labour Party, pressure would be immediate. Party leadership and senior ministers would likely:
  • Distance themselves publicly.
  • Frame the event as a “moment of poor judgment.”
  • Push for a rapid apology to the House.
If the behaviour suggested impairment or instability, the situation could escalate into internal leadership pressure, though an isolated stunt would probably end with embarrassment rather than removal. Cultural and Internet Reaction 🌐 Politically, it would be disastrous. Socially, it would become one of the most viral moments in British political history. Within hours:
  • The clip would dominate X, TikTok, and YouTube.
  • Memes would circulate globally.
  • It would likely become shorthand for the collapse of political seriousness.
The moment would probably enter British political folklore alongside bizarre parliamentary incidents. Long-Term Historical Interpretation 📚 From a political-history perspective, the event would be remembered less as a scandal and more as a symbol of institutional breakdown in political norms. Parliamentary systems rely heavily on ritual, theatre, and decorum to maintain authority. When the head of government violates that ritual in an extreme way, it damages the symbolic legitimacy of the institution itself. Historians would likely frame it as:
  • A moment illustrating the mediatization of politics
  • The tension between parliamentary tradition and modern spectacle
  • A case study in how political legitimacy can erode through performative behaviour.
✅ Bottom line: Starmer would probably survive procedurally—but politically it would be catastrophic. The act would dominate media cycles, trigger intense parliamentary backlash, and become one of the most surreal moments in modern British political history.

Do you use the paid version or free? This is totally different to the answer I got! In mine it gave a theatrical response of everyone in the chamber, the speaker shouting “order!” Badenoch looking horrified, mps shouting “that is outrageous!” Etc etc 😂 saying that though I had been asking it stupid questions for a while so I think it know not to take the question too seriously

Nanda66 · 10/03/2026 06:55

catinateacup · 09/03/2026 22:31

But genuine therapy is not there to give you advice about what to do (as some posters in to thread have suggested); or to be there on demand when you feel anxious. Therapy takes place in the way it does because it’s about a different kind of work that you do on and with yourself at a deeper level. It isn’t (despite what the reduction of CBT to online “strategies” suggests) remotely the same as a toolkit for facing anxiety-producing situations.

That’s self-help, not therapy; and it’s also more than possible that reaching for these kinds of tools is counterproductive, too.

ChatGPT is agony-auntish because it’s self help. It’s pop psychology. It’s not a trained therapist; it has no statutory or professional responsibilities; there is no therapeutic partner or counsellor there. It’s simply recombining all the pop-psychology and “relationship advice” it’s been trained on, all over the internet, Reddit, Mumsnet, etc. What you’re getting is “common sense”; “the wisdom of crowds”; and so on: but not actual therapy — any more than if you cut every issue of “Psychologies” magazine into tiny pieces, and let a computer generate an answer for you from that.

The reason you find it better is precisely because it’s self help/pop psychology. In contrast, genuine therapy with results is often difficult, uncomfortable, and not immediately “helpful” at all. It’s not there to provide relationship advice.

AI is also appealing precisely because it works with our natural psychological tendencies to read into things more relevance to us than is actually there - as people do with horoscopes and spiritualism (the Forer effect). It might be comforting, but you’re generating a lot of that comforting sense of it being useful yourself. We are meaning-making beings, and AI is a pattern generating device. Of course we want to think it’s doing it “deliberately”, or that there’s depth and intention there! It’s not nice to think that we’re projecting depth where there is only surface! But a lot of our sense of AI is generated by our own psychology.

Thank you for your insights. They’re genuinely interesting and I’m sure they are true.
But pop psychology or not, it has made a difference to me in a way that expensive therapy and hypnotherapy did not. With the help of AI I have found a way to deal with an anxiety inducing situation and I now have the strength and confidence to deal with it on my own, knowing I can ask for help if I need it. Yes, perhaps the solution was inside me all along, but AI helped me find it. From that perspective it’s been useful to me.

YorkStories · 10/03/2026 00:47

SemperIdem · 09/03/2026 16:11

A lot of people really enjoy not having to use their critical thinking skills.

There are of course industries in which AI is at this point, unavoidable.

The people however, who are using AI to think up meal plans for them, using it like a diary and so on, are bottom of the barrel low intellect.

I use Gemini every day. I think it’s fantastic. It not hard to anonymise any info I give it so I’m not worried about my data and I don’t find it difficult to get useful and accurate information out of it. I know it makes mistakes but it also makes it simple to fact check things. I do a lot of gardening, diy, sewing, car repairs etc and it’s invaluable. It’s helped save me loads of time and money. Ai gives me the confidence to do things I otherwise would have been nervous to do. Im very confident using Google but ai is often much quicker and more useful. I google if it’s a simple bit of info I need but anything that requires any depth I’ll use AI. I did a plumbing repair the other day. I just took a photo and asked AI how to repair it. It gave me all the info in one shot including part numbers etc. If I’d been googling I’d have had to google a whole load of different things to get the right info. Id have had to work out what the part was called, what make it was, what was wrong with it, how I should repair it, where can I buy the parts from etc etc. AI is rapidly getting better but I’m also getting better at using it. I know how to ask for straightforward information.

I don’t use it for any therapy type of things though.

My Mum is in her late 80s and has poor sight and hearing. She managed to fix her Hive today by using AI. It wasn’t a difficult fix but it wasn’t something she could have done without AI. Previously she would have phoned me to ask for my help, which is fine if I’m about to answer the phone. Having “someone” to ask for help who is always there and always ready to explain things as simply and slowly as my Mum wants is brilliant. My Mums eyesight will continue to deteriorate an AI is one of the tools that will help her continue to be independent.

Ijwwm · 10/03/2026 00:22

I’ve not used it yet - have been tempted slightly re asking for help redesigning my garden.

But then I remember that I watched Terminator when it first came out and that I don’t trust bloody computer systems 😂. It’s a slippery slope and I don’t have a Sarah Connor in my life, so I’ll avoid it a bit longer.

ThatPearlkitty · 10/03/2026 00:12

GarlicFound · 10/03/2026 00:02

I wish you'd turned that off before asking it a joke question!

yea, i forget that part, besides id rather keep the custom instructions intact as currently running Sherman Kent analysis at the moment

GarlicFound · 10/03/2026 00:02

ThatPearlkitty · 09/03/2026 23:26

i trained it to give me detailed intelligence type analysis rather than short simple summaries

I wish you'd turned that off before asking it a joke question!

Swipe left for the next trending thread