A title tool produced a full answer
The user, AmIAComplete, had described a family dilemma involving a neurodivergent uncle and a sick child. She asked fellow users to tell her whether she was overthinking an awkward exchange with her brother’s family. Instead, a prompt titled “AIBU Response Schema v.1.0” appeared.
It instructed an AI to write a forum post about who was right in an interpersonal conflict, give a clear verdict in Mumsnet’s style and offer brief practical advice. It also defined forum shorthand, including AIBU, ND for “neurodivergent” and DD for “darling daughter.”
The prompt told the AI not to diagnose anyone, demean neurodivergent people or imply that a sick child should be sociable. It called for empathy, personal boundaries and de-escalation. Users called its appearance “creepy,” “insane” and “really disappointing”; one said they might as well do something else if posts were being written by AI.
Mumsnet first said it did not use AI to write discussions. It attributed the text to a tool that suggests thread titles and uses an OpenAI model. Some users copied the prompt into an AI and found it could generate a complete post.
The site then said the prompt had not come from Mumsnet, suggesting it had somehow transferred from AmIAComplete’s computer. Users doubted that account, with some speculating that the prompt came from an internal Mumsnet system.
Roberts later confirmed that it did come from the site’s system: user drafts are sent to OpenAI to suggest a title, but this time the system returned a prompt for a full response instead. She said it “looks like an error on their side” and stressed that Mumsnet does not use AI to write posts or replies.
Trust is the issue
Roberts joined the discussion after users challenged the site’s explanations. One participant asked why OpenAI would independently suggest such a specific task if Mumsnet had never asked it to. Another said they no longer always knew what was real.
Speaking to The Guardian, Roberts said she understood why the prompt seemed “out of place.” Mumsnet’s systems, she said, do not request, store or use anything like it. The site has never tested a tool for writing posts, she added, though it uses AI for coding, advertising operations, discussion analysis and flagging potentially problematic posts for moderators.
The dispute comes as Mumsnet’s reach is shrinking. Founded in 2000, the site reported a 19% fall in monthly unique users in 2025, to 6.1 million, while the number of posts fell by 600,000. Its latest annual report says it is adopting AI tools to improve productivity and user experience.
A recent study published in Nature found that AI use on social media can produce “semantic sludge”—content users see as less informative, “robotic” and “faceless.” That concern is especially pointed on a forum where, Roberts said, people trust advice because it comes from other parents, not a machine.
I think the explanation may account for how the prompt appeared, but it does not answer the question users were asking: what exactly did the system send to OpenAI, and why did it return instructions for writing a whole reply? Mumsnet’s wider use of AI makes that distinction more important, not less. The site’s promise that people write its posts depends on users believing they can tell where the machine’s role ends.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X