Pieter Levels Asked X for a Lifetime Ban on AI Replies
Pieter Levels wants one AI reply to cost someone their account.
On 29 October 2026, at 13:58 UTC (00:58 on 30 October in Sydney), he posted the rule in plain lines. “X should have a zero tolerance policy for AI replies”. Then: “Like you do one and your account is suspended for life”. Then: “Replies are a nightmare now and it’s never ever been this bad”. He called AI replies an “existential threat to the future of this platform”, because “people will simply stop using it if the vast majority of replies are AI bots”.
About seven hours later, around 21:05 UTC, that post had roughly 3,500 likes, 930 replies, 108 quotes, and about 180,000 views. Those counts are approximate, from a read of the X API at the time of writing. They move. The posts gathered here include no official reply from X, and no new rule.
The AI-reply complaints started earlier in October
The lifetime ban was the last step in a complaint he had already been counting.
On 16 October he wrote, under one of his own posts, “150+ AI replies under this”. The next line: “AI replies have never been this bad!”
On 23 October the same complaint left the reply column. “AI replies was a thing but now I’m getting lots of AI emails”.
Who leads product at X had just changed, in public
The ban post arrived in the same stretch of weeks as a change of product lead, told in the leads’ own posts.
On 3 October, at 02:22 UTC, Nikita Bier wrote “Two hours ago I was officially disconnected from my X laptop.” The same day, writing about spam, he said “a tsunami is coming and I’m going to drop everything to focus on that.”
On 5 October he named a different fix. He argued the key problem is “a broadly accepted standard for agents to identify themselves”. That is a label for agents. Levels later asked for a lifetime suspension after a single AI reply.
On 26 October, at 20:02 UTC, Peter Sellis posted “I’m joining the team at @X to lead Product.” At the time of writing that post had about 22,700 likes.
An hour later, at 21:02 UTC, Rob Hallam replied. Hallam (@robj3d3) is the SuperX founder. His words: “First request: please fix the AI reply guys… they had a massive resurgence since @nikitabier was gone”. That resurgence is his view of the month. His Attention Playbook pre-sale is a separate story. The longer record is SuperX, the tool, beside the course.
The pushback: AI photos, useless replies, and false positives
The first sharp answer was about Levels, and about what he sells.
At 14:45 UTC, @beaulawsonco wrote “Guy who sells AI photos is mad about AI replies”. At the time of writing that post had about 330 likes. Levels runs the AI photo app Photo AI. Three minutes later, at 14:48 UTC, he answered “Because AI is actually good at photos, it sucks at writing useful replies”.
Usefulness is the line he drew. A reply under it moved the line. @bezibalazs wrote “then you are not arguing against AI replies but useless replies?” A ban on useless replies is a judgment about quality. A ban on AI replies is a judgment about authorship. A lifetime rule has to pick one, and then apply it twice and get the same answer.
Detection is where that breaks. @infiloop2 asked “how do you distinguish ai replies from human even”. @yuvadm named the false positive on the other side of a style test: “people are actually starting to sound like AI IRL”. If people already write like the models, a ban that keys off the sound of a sentence will suspend humans.
Scope is the third crack. @sethmosk asked whether an AI photo of yourself would also mean a permanent ban. Levels had just said AI is good at photos. A rule written for reply text still has to say what it does to an image, and to a founder whose product is those images.
@valigo (Valentin Ignatev) quote-posted at 15:33 UTC: “Reading and writing prose is, and should stay human!”
The same day showed the other half of the trade, without a verdict. @tibo_maker posted a contest for his AI video product, Revid: “which one of them is a real human?” Founders are shipping realistic AI and arguing about AI noise in the same hours.
Hallam posted again on 29 October, this time about another network. “The normie hate for AI is on another level.” He was writing about Threads ratio-ing AI videos. It sits next to his request that X fix the reply guys. Both are his read.
What a lifetime ban still has to answer
Bier’s sentence from 5 October is the other design already in public: agents that identify themselves. Levels asked for a lifetime suspension after one AI reply. The posts in this piece do not say which design X will use. They include no official reply, and no policy change.
A zero-tolerance line is easy to repeat. Running it means a detector that can be wrong, a choice between an AI reply and a useless one, and a boundary around photos and video. The replies under the original post are already those three arguments.
What it means if you build in public on X
For indie hackers, the reply column is distribution. You post the work, and a reply under someone else’s post is how a smaller account enters a larger conversation. Build in public turned that habit into a tactic. A following is one channel, and on X that channel often runs through replies. The thing being sold, underneath the feed, is still a product the maker owns.
Levels’ claim is that people leave if most of those replies are bots. If readers stop opening the replies, the look gets worse for the humans posting there too. If X did suspend an account for life over one AI reply, those same people would live with the test the replies already asked for: how to tell a bot from a person who sounds like one.
He has argued from the product side in public as well. The week he wrote that software is dead, and the vibe-coded SaaS replacements, belong to a founder who also sells AI photos. The short record keeps those posts together. Whether indie hacking itself is fading is a wider question than whether the reply column is still worth reading.
A short timeline
- 3 October. Nikita Bier says he was disconnected from his X laptop, and writes that a spam tsunami is coming.
- 5 October. Bier argues the key problem is a standard for agents to identify themselves.
- 16 October. Levels counts “150+ AI replies” under one post.
- 23 October. Levels says AI emails are joining the AI replies.
- 26 October. Peter Sellis says he is joining X to lead Product. Rob Hallam asks him to fix the AI reply guys.
- 29 October. Levels calls for zero tolerance and a lifetime suspension. About seven hours later the post had roughly 930 replies and 108 quotes.
A note on the numbers
Every engagement figure here is approximate, from a read of the X API around 21:05 UTC on 29 October 2026 (08:05 on 30 October in Sydney). Levels’ ban post, published at 13:58 UTC that day: about 3,500 likes, 930 replies, 108 quotes, and about 180,000 views. The reply that called him a guy who sells AI photos: about 330 likes. Peter Sellis’s joining post, from 20:02 UTC on 26 October: about 22,700 likes. The wording of each post is in the links. No official response from X, and no policy change, appears in these sources.
Should X ban AI replies outright, and does it change how you build in public?
FAQ
What did Pieter Levels ask X to do about AI replies?
On 29 October 2026 at 13:58 UTC he posted that X should have a zero tolerance policy for AI replies, and that one AI reply should mean a lifetime suspension. He wrote that replies had never been this bad, and that people will stop using the platform if most replies are bots. At the time of writing, about seven hours later, the post had roughly 3,500 likes, 930 replies, 108 quotes, and about 180,000 views.
Has X banned AI replies?
No official response and no policy change appear in the posts this story uses. On 26 October 2026 Peter Sellis wrote that he was joining X to lead Product. On 3 October Nikita Bier wrote that he had been disconnected from his X laptop. Those are their own posts, not a new reply rule.
What were the main arguments against a lifetime ban?
Replies raised detection, false positives, and usefulness. One asked how anyone tells an AI reply from a human reply. Another said people are starting to sound like AI in real life. Another said the complaint is about useless replies, not AI replies. A quote post said the man who sells AI photos was angry about AI replies. Levels answered that AI is good at photos and bad at useful replies. Another post asked whether an AI photo of yourself would also mean a permanent ban.
What did Rob Hallam ask the new X product lead?
On 26 October 2026 at 21:02 UTC, Rob Hallam, the SuperX founder, replied to Peter Sellis and asked him to fix the AI reply guys, who Hallam said had resurged since Nikita Bier left. That is Hallam's view. On 29 October he also wrote, about Threads ratio-ing AI videos, that normie hate for AI was on another level.
Why does the AI-reply fight matter if you build in public?
Indie hackers who grow on X often do it by posting the work and joining other conversations in the replies. Levels argues people will stop using the platform if most replies are bots. A lifetime ban, which X has not announced, would also have to deal with detection, false positives, and the line between a useless reply and an AI reply. Those objections are already in the replies under his post.