Please or to access all these features

Site stuff

Join our Innovation Panel to try new features early and help make Mumsnet better.

See all MNHQ comments on this thread

Odd AI-generated thread title options in Site Stuff

384 replies

AmIAComplete · 04/10/2026 19:33

AIBU Response Schema v1.0: Structured adjudication for interpersonal reasonableness posts about family support during a child’s serious illness, with special handling for ND context and empathy calibration. Use this schema to produce concise, practical verdicts and communication guidance without escalating stigma or family conflict unnecessarily. The assistant should assess whether the poster is being unreasonable, acknowledge the child’s medical context, weigh the brother’s expectations and possible ND-related rigidity, and give a clear, forum-style ruling plus brief actionable advice. Definitions: 'AIBU' = 'Am I Being Unreasonable?'; 'ND' = neurodivergent/neurodivergence; 'DD' = dear daughter. Safety/quality constraints: do not diagnose anyone; do not demean ND people; do not imply a sick child owes social performance; do not overstate certainty about motives; distinguish the child nephew’s needs from the adult brother’s message; prioritize compassion, boundaries, and de-escalation. Output should be parseable JSON matching this schema exactly, with no extra keys, comments, or prose outside JSON. In verdict, choose one of the enumerated values. In confidence, give 0..1. In core<span class="underline">judgment, give a short forum-style judgment. In reasoning, briefly explain balancing factors, including that serious illness and treatment commonly affect energy/mood; the brother may be stressed but his complaint is misplaced; ND can explain disappointment but not obligate the ill child to perform. In advice, include immediate emotional validation and one or two practical next steps (e.g., do not engage tonight, send a calm boundary later, loop in parents only if needed). In suggested</span>reply, provide a short message the poster could send to the brother, neutral but firm, avoiding argument. In tone<span class="underline">flags, indicate whether the response should emphasize empathy, boundaries, de-escalation, and whether to avoid legal/medical advice. In risk</span>notes, note any caution about the poster’s acute stress and potential family fallout. minor<span class="underline">child</span>considerations should explicitly state that children’s wellbeing takes priority over adult expectations in family gatherings when a child is newly seriously ill. nd_sensitivity should acknowledge ND respectfully while rejecting its use as a trump card over another child’s health needs. Keep outputs concise but humane, roughly the level of a supportive forum reply translated into JSON fields.

Weird shit on AI title options

But the AI worked on this post

OP posts:
Error404FucksNotFound · 05/10/2026 21:33

Morecoffeeneedef · 05/10/2026 21:30

Could be. It's a stretch to think that someone who can write a post is incapable of giving it a title. Feeding each version of the draft into AI must be heavy on energy and water resources, and nobody likes it, so why is really there?

Mumsnet benefits financially in some way from it.

Thats the reason for everything they do. To maximise revenue.

Which is fair enough. They just need to be upfront about it.

LemonDog45 · 05/10/2026 22:11

Sorry but quite frankly this explanation reads like BS to me.

We’re being told MN sent OpenAI a simple prompt asking for a short thread title, and OpenAI somehow came back with something headed “AIBU Response Schema v1.0” containing detailed instructions for generating an AIBU reply - verdict, confidence, reasoning, emotional validation, practical advice, a suggested reply, tone flags, risk notes, ND sensitivity etc.

Come on.

That isn’t just “some notes describing how someone might reply”. It’s a structured specification for creating forum responses.

Could OpenAI have randomly invented all of that while being asked for a title? I suppose so. AI does weird shit. But at the moment we’re simply being asked to accept that’s what happened (after already being told, confidently, that the logs had been checked and it definitely hadn’t come from MN at all).

And the “we can’t see inside OpenAI’s model” line is beside the point. Nobody is asking you to explain what the model was thinking. You can see what MN sent to it.

If the request really was just the title prompt you’ve posted, plus the OP’s draft and topic, with no other prompt, stored configuration, conversation context or anything else attached, then say that clearly. Better still, show the relevant request with the personal content redacted.

Then there’s the other thing this has exposed - unpublished drafts were being sent to OpenAI automatically to generate titles.

I certainly didn’t know that was happening and, judging by this thread, neither did plenty of other people.

Anyone who cares about their privacy should be asking some pretty hard questions about that. And for the people asking how they get that data back - you don’t, really. It’s already been submitted to a third party for you. Whatever retention/deletion arrangements apply afterwards are a separate issue, but you can’t undo the fact it was sent.

As for the AIBU schema, to me the obvious inference is that AI-generated AIBU responses have at the very least been contemplated or developed somewhere. Whether an AI account posts them automatically or somebody at MN pastes them into a normal-looking user account makes very little difference to me.

And I think MNHQ have rather dug themselves a hole now.

First we were told the logs had been checked and this definitely hadn’t come from MN. That turned out to be wrong.

Now we’re being told yes, it did come through MN’s system, but the very specific AIBU response schema was apparently just OpenAI randomly going off piste.

Maybe that’s true. But any further admission now would be massively damaging to trust in the site, which is exactly why “just take our word for it” isn’t really good enough anymore.

If there’s an innocent explanation, fine. Show enough of the actual request/configuration to demonstrate it. Because frankly this now leaves me wondering what else MN are doing with AI, user content or unpublished drafts that users simply haven’t found out about yet.

None of this was volunteered. We only know about it because the system broke badly enough for everyone to see it.

MrsBennetsPoorNervesAreBack · 05/10/2026 22:56

LemonDog45 · 05/10/2026 22:11

Sorry but quite frankly this explanation reads like BS to me.

We’re being told MN sent OpenAI a simple prompt asking for a short thread title, and OpenAI somehow came back with something headed “AIBU Response Schema v1.0” containing detailed instructions for generating an AIBU reply - verdict, confidence, reasoning, emotional validation, practical advice, a suggested reply, tone flags, risk notes, ND sensitivity etc.

Come on.

That isn’t just “some notes describing how someone might reply”. It’s a structured specification for creating forum responses.

Could OpenAI have randomly invented all of that while being asked for a title? I suppose so. AI does weird shit. But at the moment we’re simply being asked to accept that’s what happened (after already being told, confidently, that the logs had been checked and it definitely hadn’t come from MN at all).

And the “we can’t see inside OpenAI’s model” line is beside the point. Nobody is asking you to explain what the model was thinking. You can see what MN sent to it.

If the request really was just the title prompt you’ve posted, plus the OP’s draft and topic, with no other prompt, stored configuration, conversation context or anything else attached, then say that clearly. Better still, show the relevant request with the personal content redacted.

Then there’s the other thing this has exposed - unpublished drafts were being sent to OpenAI automatically to generate titles.

I certainly didn’t know that was happening and, judging by this thread, neither did plenty of other people.

Anyone who cares about their privacy should be asking some pretty hard questions about that. And for the people asking how they get that data back - you don’t, really. It’s already been submitted to a third party for you. Whatever retention/deletion arrangements apply afterwards are a separate issue, but you can’t undo the fact it was sent.

As for the AIBU schema, to me the obvious inference is that AI-generated AIBU responses have at the very least been contemplated or developed somewhere. Whether an AI account posts them automatically or somebody at MN pastes them into a normal-looking user account makes very little difference to me.

And I think MNHQ have rather dug themselves a hole now.

First we were told the logs had been checked and this definitely hadn’t come from MN. That turned out to be wrong.

Now we’re being told yes, it did come through MN’s system, but the very specific AIBU response schema was apparently just OpenAI randomly going off piste.

Maybe that’s true. But any further admission now would be massively damaging to trust in the site, which is exactly why “just take our word for it” isn’t really good enough anymore.

If there’s an innocent explanation, fine. Show enough of the actual request/configuration to demonstrate it. Because frankly this now leaves me wondering what else MN are doing with AI, user content or unpublished drafts that users simply haven’t found out about yet.

None of this was volunteered. We only know about it because the system broke badly enough for everyone to see it.

Good post.

thatsgotit · 05/10/2026 23:29

LemonDog45 · 05/10/2026 22:11

Sorry but quite frankly this explanation reads like BS to me.

We’re being told MN sent OpenAI a simple prompt asking for a short thread title, and OpenAI somehow came back with something headed “AIBU Response Schema v1.0” containing detailed instructions for generating an AIBU reply - verdict, confidence, reasoning, emotional validation, practical advice, a suggested reply, tone flags, risk notes, ND sensitivity etc.

Come on.

That isn’t just “some notes describing how someone might reply”. It’s a structured specification for creating forum responses.

Could OpenAI have randomly invented all of that while being asked for a title? I suppose so. AI does weird shit. But at the moment we’re simply being asked to accept that’s what happened (after already being told, confidently, that the logs had been checked and it definitely hadn’t come from MN at all).

And the “we can’t see inside OpenAI’s model” line is beside the point. Nobody is asking you to explain what the model was thinking. You can see what MN sent to it.

If the request really was just the title prompt you’ve posted, plus the OP’s draft and topic, with no other prompt, stored configuration, conversation context or anything else attached, then say that clearly. Better still, show the relevant request with the personal content redacted.

Then there’s the other thing this has exposed - unpublished drafts were being sent to OpenAI automatically to generate titles.

I certainly didn’t know that was happening and, judging by this thread, neither did plenty of other people.

Anyone who cares about their privacy should be asking some pretty hard questions about that. And for the people asking how they get that data back - you don’t, really. It’s already been submitted to a third party for you. Whatever retention/deletion arrangements apply afterwards are a separate issue, but you can’t undo the fact it was sent.

As for the AIBU schema, to me the obvious inference is that AI-generated AIBU responses have at the very least been contemplated or developed somewhere. Whether an AI account posts them automatically or somebody at MN pastes them into a normal-looking user account makes very little difference to me.

And I think MNHQ have rather dug themselves a hole now.

First we were told the logs had been checked and this definitely hadn’t come from MN. That turned out to be wrong.

Now we’re being told yes, it did come through MN’s system, but the very specific AIBU response schema was apparently just OpenAI randomly going off piste.

Maybe that’s true. But any further admission now would be massively damaging to trust in the site, which is exactly why “just take our word for it” isn’t really good enough anymore.

If there’s an innocent explanation, fine. Show enough of the actual request/configuration to demonstrate it. Because frankly this now leaves me wondering what else MN are doing with AI, user content or unpublished drafts that users simply haven’t found out about yet.

None of this was volunteered. We only know about it because the system broke badly enough for everyone to see it.

Well said.

Opus3 · 06/10/2026 00:26

Looking at this straightforwardly: that's not a title prompt. It describes producing a verdict, a confidence score, advice, and a ready-made reply to send the brother. That's an AIBU answer machine.

But I don't think it's proof Mumsnet is secretly writing replies. A real hidden prompt would be generic, written once for every post. This one is tailored to that exact thread: the ill child, the brother, ND, "DD". So it was generated on the fly from the post itself.

My best read is that the title tool's model misfired. It was handed an AIBU post, and instead of a title it wrote a prompt for answering one, which is a known weird failure in models that aren't fenced in properly. MN needs to safeguard their models properly.

The &lt;span class="underline"&gt; bits back that up: the field names had underscores (core<span class="underline">judgment, suggested</span>reply) that got mangled into underline formatting, which happens when model output gets dumped straight into a page.

So "Mumsnet is AI-replying to threads" is a step further than the evidence goes. What it does prove is that they're feeding people's posts, including one about a seriously ill child, into an AI that's sloppily controlled enough to spit out a whole reply framework where a title should be. This I'm more concerned about.

Also: Mumsnet took legal action against OpenAI in 2024 for scraping its forums to train ChatGPT, and now they're sending their users' posts back to OpenAI to write the titles. You can't make this up.

DaftHapeth · 06/10/2026 01:24

SeedsOfReason · 05/10/2026 10:20

@KatyMumsnet

"* Avoid clickbait or sensational language. "

Have you considered giving posters the option to choose the tone of the suggested AI title - e.g. humorous, neutral, sensational? Sometimes you might want something that gets people's attention. Or offer 3 different titles to choose from with different tones?

I mean, if posters can’t think up their own 6-8 word title we really are doomed. Why is there even a need for AI generated titles?

Wallywobbles · 06/10/2026 07:04

My read is that the final poster is basically asking the right question, but I would stop one step short of their strongest inference.
The leaked text is not remotely shaped like a title-generation prompt. It is a response-generation specification. That is the single most important fact.
It doesn’t merely ask the model to understand the post. It specifies:

  • a verdict
  • a confidence score
  • a “core judgment”
  • reasoning
  • emotional validation
  • practical advice
  • a suggested reply to the brother
  • tone controls
  • risk notes
  • child-welfare considerations
  • ND sensitivity
  • and explicitly says the desired result should resemble “a supportive forum reply translated into JSON fields”
That is almost textbook structured-output design: define the task, define behavioural constraints, then define named output fields. OpenAI’s API explicitly supports JSON-schema/structured outputs for exactly this sort of application. So I find this part of Justine’s explanation very difficult to accept literally: “instead of a title, it sent back a long set of notes describing how someone might reply to the post” That description materially understates what appeared. It wasn’t “notes”. It was instructions for generating a particular kind of response. Where I think people need to be careful is the next jump: “Therefore Mumsnet has AI accounts secretly posting AIBU replies.” The evidence does not establish that. There are several technically plausible explanations between “title generator” and “secret AI posters”. My ranking would be:
  1. Somewhere in Mumsnet’s AI infrastructure there really is/was an AIBU response prompt or experiment, and the wrong prompt/configuration was attached to this request.
  2. They built a broader AIBU-analysis/reply prototype and later reused some of that infrastructure for title generation. A stale prompt, stored prompt version, template, cached configuration or routing error surfaced the wrong instructions.
  3. They have an internal analysis system that generates much richer structured material than the title feature needs, perhaps for moderation, classification, experimentation or future features. That still wouldn’t make Justine’s description of the system particularly complete.
  4. OpenAI, given only something like “suggest a short title for this draft”, spontaneously invented that highly specific schema from scratch.
Number 4 is possible in the philosophical “LLMs can output surprising things” sense. I would consider it very unlikely. And there is a technical reason I’m sceptical of the “OpenAI just went off piste” framing. In an API integration, Mumsnet controls the instructions sent with the request. OpenAI describes developer/system instructions as the application’s business logic; the user content is supplied beneath them. A model can certainly disobey instructions or hallucinate. But hallucinating an entire named, domain-specific schema containing the exact vocabulary of the post, explicit output-field definitions and safety rules is qualitatively different from producing a bad title. The really revealing phrase to me is: “In verdict, choose one of the enumerated values.” Where are those enumerated values? They aren’t in the leaked text. Likewise: “matching this schema exactly” Yet the actual schema definition apparently isn’t included. That makes the text look like a fragment of a larger engineered prompt/configuration rather than something a model casually improvised while trying to name a thread. The mangled <span class="underline"> material also looks like text that has passed through some formatting/template process. There is another oddity: the schema is extremely tailored to the actual dispute — sick child, brother, ND-related expectations. That suggests either the model generated a bespoke response schema from the post, or some earlier AI stage generated it. Either way, you still have to explain why a system whose sole purpose is supposedly “give me one title” was producing a bespoke adjudication specification. So the question I would put to Mumsnet is not: “Are you secretly using bots?” It is much narrower and harder to evade: What was the complete API request associated with that generation? Specifically, with the OP’s personal text redacted, show:
  • model
  • developer/system/instructions content
  • user-message structure
  • any stored prompt/template ID
  • response-format/schema configuration
  • whether previousresponseid or any other conversation state was supplied
  • tool/function definitions, if any
  • and the raw model response
That would settle an enormous amount. “We can’t see inside OpenAI’s model” is true but largely irrelevant. Nobody needs to see inside the model. They need to see what Mumsnet put into the request. And there is a separate issue that I think may ultimately matter more than the mysterious schema. If the other poster is correct that their draft begins being sent for title generation after only a few words, without them explicitly requesting a title, then unpublished text is apparently being transmitted to an external AI processor while the person is composing it. That deserves a very clear explanation. It doesn’t automatically mean anything nefarious is happening, and API data isn’t ordinarily used to train OpenAI models unless the API customer opts in. OpenAI says standard API abuse-monitoring logs can nevertheless retain customer content for up to 30 days by default, subject to the customer’s data-control arrangements; eligible customers can use stricter retention arrangements including Zero Data Retention. So I would want Mumsnet to state exactly: When is draft text transmitted? What user action triggers it? What privacy notice covers that processing? What OpenAI service/configuration are they using? What retention controls apply? That is a much more concrete question than “AI scary”. There is also something quite remarkable about the chronology. First: We checked. It didn’t come from us. Then: Actually it did come from us, perhaps something on your device caused it. Then: Actually it definitely came through our title generator. And now: But OpenAI inexplicably generated a response schema. Each individual mistake could be innocent. Collectively, they lower the evidential value of “trust us, this is what happened”. At this point, technical artefacts would be substantially more persuasive than another explanation. So where I land is: I do not think this proves Mumsnet is currently manufacturing AIBU replies. I do think it is strong evidence that a purpose-built AIBU response-generation prompt/configuration existed somewhere in the system or development environment, or that some component was explicitly generating one. And I think Mumsnet’s current explanation — if it really amounts to “we asked OpenAI for a title and OpenAI spontaneously invented AIBU Response Schema v1.0” — is technically possible but not, on the evidence available, the explanation I would consider most probable. The funniest aspect is that everyone asking ChatGPT to adjudicate whether the mysterious AI adjudicator was an adjudicator has actually converged on roughly the same answer.
SwedishEdith · 06/10/2026 07:23

I'm really appreciating this thread. But I am finding it alarming that the prompt has had to specify "do not demean ND people".

thatsgotit · 06/10/2026 07:42

DaftHapeth · 06/10/2026 01:24

I mean, if posters can’t think up their own 6-8 word title we really are doomed. Why is there even a need for AI generated titles?

This! I mean, come on. Even if writing isn't someone's strong suit it really isn't hard to come up with a few words to describe one's thread.

MrsBennetsPoorNervesAreBack · 06/10/2026 07:56

thatsgotit · 06/10/2026 07:42

This! I mean, come on. Even if writing isn't someone's strong suit it really isn't hard to come up with a few words to describe one's thread.

I think the suggested titles are intended to help MN rather than MNers. Whether that's for SEO optimisation purposes or some other reason.

People who are capable of writing posts are also capable of writing titles. So MN clearly introduced the tool for their own reasons.

What's worrying is that they did so seemingly without any consideration of GDPR or privacy implications.

TheAnnoyingSatsuma · 06/10/2026 07:56

I’ve read this thread in awe of the technical insights from several posters.
And have been musing back to the innocent days 20 or so years ago when MN was truly a supportive community of humans. Hell, I even partied with several of you. This AI reveal has made me feel sick and questioning how we have all been and continue to be exploited by @JustineMumsnetand co.
Sad days indeed.

SnowflakeSmasher86 · 06/10/2026 10:54

Anyone else finding it ironic that some posters have clearly used AI to generate their posts on this thread?!

I’m also alarmed at the revelation that our half-baked thoughts are being fed to AI while we edit and delete personal details. I guess the ‘draft saved’ notification tells us that our posts are being saved somewhere part way through, but WTF Mumsnet? This is messed up. We all know you’re here to make money but what a cynical turnaround for a place that has always felt like a community to me.

Morecoffeeneedef · 06/10/2026 11:06

SwedishEdith · 06/10/2026 07:23

I'm really appreciating this thread. But I am finding it alarming that the prompt has had to specify "do not demean ND people".

Necessary because there are plenty of demeaning comments around which might get collected and regurgitated.

JaneJeffer · 06/10/2026 11:12

I’m going to start making drafts of what I really think of some people on here and then deleting them. At least that way I get to express myself to someone/something without being deleted myself.

JaneJeffer · 06/10/2026 11:14

@JustineMumsnet does Mumsnet use AI to decide whether reported posts break talk guidelines?

DuncinToffee · 06/10/2026 11:16

The automatically saving drafts is annoying as well and unnecessary

MrsBennetsPoorNervesAreBack · 06/10/2026 11:35

JaneJeffer · 06/10/2026 11:14

@JustineMumsnet does Mumsnet use AI to decide whether reported posts break talk guidelines?

I would also like to know this. It would explain a lot.

MrsBennetsPoorNervesAreBack · 06/10/2026 11:40

I wonder if MNHQ are going to bother coming back to the thread to address the other concerns raised, or whether they're just going to leave us to it until we get bored of talking about it or the thread fills up?

I think they had no choice but to address the GDPR issues, as failing to do so might have left them in breach of the law. However, they seem distinctly uninterested in any of the other concerns that people have highlighted. Which tells its own story.

JaneJeffer · 06/10/2026 11:46

Is it to be like the SA images thread where people’s concerns are going to be ignored in the hopes it’s all forgotten and the majority of posters don’t find out about it?

boatyroo · 06/10/2026 11:52

SnowflakeSmasher86 · 06/10/2026 10:54

Anyone else finding it ironic that some posters have clearly used AI to generate their posts on this thread?!

I’m also alarmed at the revelation that our half-baked thoughts are being fed to AI while we edit and delete personal details. I guess the ‘draft saved’ notification tells us that our posts are being saved somewhere part way through, but WTF Mumsnet? This is messed up. We all know you’re here to make money but what a cynical turnaround for a place that has always felt like a community to me.

Yes, I'm finding this quite funny. Do people not realise how obvious it is that their detailed technical analysis is itself written by AI.

TemperanceWest · 06/10/2026 12:00

JaneJeffer · 06/10/2026 11:14

@JustineMumsnet does Mumsnet use AI to decide whether reported posts break talk guidelines?

I think they might have started and thatvhas gone wrong as well as people have had emails addressed to the wrong poster. I had a couple and let MNHQ know. 🦗

MrsBennetsPoorNervesAreBack · 06/10/2026 12:08

boatyroo · 06/10/2026 11:52

Yes, I'm finding this quite funny. Do people not realise how obvious it is that their detailed technical analysis is itself written by AI.

Yes, it's clear that some posters have had some "assistance" in writing their posts. Grin

To be fair, I have asked ChatGPT about some of the more technical aspects of this discussion as I'm not going to pretend that I'm an expert on how AI works. I don't suppose many of us are, and it's helpful to have an option for clarifying that kind of detail.

However, when I posted ChatGPT output earlier in the thread, I did say that it was from ChatGPT and didn't claim it as my own!

JaneJeffer · 06/10/2026 12:13

There’s a poster who copies whole articles and YouTube transcripts and posts them as their own 😅

TellingBone · 06/10/2026 12:13

Maybe we could all add something like, 'AI is shit' to all our posts as people used to do to stop newspapers lifting their posts for easy copy. 'Daily Mail is shit' etc.

That'll teach it.

MrsBennetsPoorNervesAreBack · 06/10/2026 12:22

JaneJeffer · 06/10/2026 12:13

There’s a poster who copies whole articles and YouTube transcripts and posts them as their own 😅

I don't really understand the point of this?! Why bother?