LinkedIn moves to suppress AI-generated comments
LinkedIn's move against synthetic comments rewards specificity over volume, and punishes brands that automated their engagement.
Key takeaways
- LinkedIn is now detecting and down-ranking comments it judges to be AI-generated.
- Automated 'engagement at scale' tactics, common in B2B sales teams, are the main casualty, not spam accounts.
- LinkedIn has not disclosed its detection method, raising the risk of false positives on genuinely human but blandly written comments.
- Specific, judgment-driven comments now carry more algorithmic and reputational value than high-volume generic replies.
- Institutions should prioritise a few substantive comments a week over quota-driven commenting programmes.
LinkedIn has started down-ranking comments it judges to be AI-generated, according to Social Media Today, which reports that the platform is now detecting and suppressing the visibility of auto-written replies clogging its feed. The company has not published a detection methodology, which is itself telling. A platform that cannot fully explain how it is adjudicating authenticity is nonetheless prepared to act on the judgment at scale.
This matters more than it sounds. For three years, LinkedIn's growth playbook has rested on a single tactic: comment on everything, fast, with tools that generate a plausible-sounding reaction in seconds. Engagement pods formalised it. AI commenting extensions industrialised it. A cottage industry of browser plug-ins now exists purely to let sales reps and founders "engage at scale" without reading the post they are replying to. LinkedIn just told that industry its product no longer works.
The arithmetic that made AI comments attractive is the same arithmetic that is killing them
The logic was always simple. LinkedIn's algorithm rewards early comments more than early likes, because a comment signals a conversation, and conversations keep users on the platform longer than a passive scroll. Sales teams and personal-brand consultants reverse-engineered that signal and automated it: generate a generic but topical reply, post it within minutes of publication, repeat across hundreds of posts a day. Volume replaced judgment. For a while it worked, because the algorithm could not easily tell a hollow "Great insight, thanks for sharing!" from a considered one.
That gap is now closing. Suppressing AI-flagged comments is not a moral crusade against bots; it is LinkedIn protecting the one thing that makes its feed valuable to buyers and advertisers: the credibility of the conversation layer. A feed full of synthetic affirmation is a feed nobody trusts, and a platform nobody trusts cannot charge premium rates for Sales Navigator seats or sponsored content. The crackdown is a defence of LinkedIn's commercial model, not an act of platform hygiene.
Who actually loses reach here
The losers are not the obvious targets. Spam accounts running comment bots at industrial scale will adapt or get banned; that was always the cost of doing business for them. The real casualties are legitimate B2B teams who quietly outsourced "engagement" to automation because manual commenting at volume was expensive and tedious. A financial-services communications team running an exec's comment strategy through an AI tool to hit a "10 comments a day" quota is now producing content that LinkedIn's own system actively buries. The quota was always a vanity metric; now it is a liability.
This should reorder how serious institutions think about LinkedIn engagement. A multilateral agency or an industrial group with a handful of senior voices posting regularly does not need volume. It needs two or three comments a week that demonstrate the executive actually read the post: a specific number challenged, a counter-example from their own market, a question that only someone with domain knowledge would ask. That kind of comment drives the profile visits and DMs that pipeline actually depends on, because it signals a real person with real judgment, not a reply-generator running on autopilot.
The detection problem nobody is solving
LinkedIn's own position is awkward. It is a platform built partly on the assumption that professional content should read a certain way: polished, confident, faintly corporate. That register is exactly what large language models reproduce best. Distinguishing an executive's genuinely considered reply from an AI-polished one is not a technical problem with a clean answer, and LinkedIn has given no indication it has solved it. Expect false positives: real comments, written by real people who happen to write in a flat, LinkedIn-ish voice, getting quietly suppressed alongside the slop.
That uncertainty is itself a signal worth acting on. Brands that have trained their executives, through years of content-strategy advice, to write comments in the same bland register (three sentences, an exclamation point, a generic compliment) are now exposed to algorithmic suspicion they did not sign up for. The safest position is not stylistic blandness; it is specificity. A comment citing a data point, disagreeing with part of the post, or referencing a named deal or project reads as human because it is harder to generate convincingly and therefore less likely to be mistaken for synthetic filler.
The deeper shift is about what LinkedIn now treats as its scarce resource. Reach used to be rationed by attention. It is increasingly rationed by provable authorship. For institutions whose entire value proposition rests on being taken seriously, that is a rule worth building a comment strategy around, not working around.