LinkedIn drops AI post enhancer for a plain proofreader

LinkedIn has quietly buried its "Post Enhancer," the AI tool that rewrote drafts into something with more polish and, invariably, less voice. Social Media Today reports that the platform has replaced it with "Post Proofreader," a narrower feature that checks grammar and phrasing rather than rewriting content wholesale. This is not a feature swap. It is an admission.

For eighteen months, LinkedIn pushed the opposite bet: that members wanted the platform to help them sound better, faster, by generating fuller drafts from a few bullet points or a rough paragraph. The tool made sense on paper. Writing a post is friction; friction suppresses posting frequency; posting frequency drives the engagement numbers LinkedIn sells to advertisers. Remove the friction, get more content, get more time-on-feed. Standard platform arithmetic.

The arithmetic broke on contact with the feed itself. Post Enhancer output had a signature: the same rhythmic cadence, the same bolded takeaway line, the same faux-vulnerable opener about failure and growth. Once a few thousand creators ran their drafts through it, that signature became recognisable, then mockable, then actively penalised by the audience that mattered most: the people who comment, save, and click through to a profile rather than just scroll past. LinkedIn's own community did the quality control the algorithm couldn't. Members started calling it out in comments. Engagement on visibly AI-smoothed posts softened. The platform's answer is Post Proofreader, a tool that fixes typos and awkward syntax but leaves the writer's actual sentences intact.

The tell is in what LinkedIn chose to keep. Grammar-checking is invisible when it works; nobody screenshots a corrected comma splice to mock it. Rewriting is visible by design, because the entire value proposition was "sound more like a professional," which is another way of saying "sound less like yourself." LinkedIn has effectively conceded that on this platform, voice is the product. Strip it out and you get content that performs worse, not better, because the feed's ranking signals (dwell time, replies, reshares with commentary) reward specificity and idiosyncrasy, and AI smoothing sands both away.

That has a direct cost for the institutions this matters most to. A World Bank programme officer, a Holcim regional director, an insurance CFO: these are people whose LinkedIn authority rests on sounding like they actually know something, not like a competent committee wrote their update. Post Enhancer's failure mode was making a chief economist's market note read identically to a recruiter's "excited to announce" post. For financial-services and multilateral communicators, whose credibility depends on precision and a recognisable point of view, that homogenisation was a real threat to reach, because the algorithm and the audience both, in different ways, learned to discount it.

The backlash was never really about AI. It was about disclosure and detection. A tool that helps a non-native English speaker fix subject-verb agreement is uncontroversial; nobody objects to spellcheck. A tool that generates the argument, the anecdote, and the framing is a different act, and audiences on a professional network, oddly, are better at spotting it than audiences on more casual platforms, precisely because LinkedIn users are trained to read for competence signals. Post Proofreader draws the line LinkedIn should have drawn from the start: assistance with mechanics, not authorship of substance.

For comms teams managing executive presence at scale, the operational lesson is narrower than "AI is bad." It is that AI belongs upstream (research, drafting structure, headline testing) and downstream (proofreading, formatting), never in the middle where the actual argument gets made. The output that still travels on this platform is the one with a specific number, a named source, and a sentence only that particular executive would write. Tools that blur those fingerprints will keep losing to tools, like this one, that leave them alone.

LinkedIn built a feature, watched its own users reject the aesthetic it produced, and pulled it. That is a rare instance of a platform listening to taste rather than metrics, and it is worth remembering the next time a vendor pitches an AI writing tool promising to make an executive's posts sound more compelling. On LinkedIn, compelling and generic are now mutually exclusive.