A new analysis cited by the Wall Street Journal found that 41% of LinkedIn long-form posts, and 81% of public posts examined more broadly, showed more than a moderate amount of AI generation. LinkedIn itself has been pushing writing assistance tools that make this worse, not better. The platform built to prove you know your industry is now full of posts nobody actually wrote.
This is not a minor content-quality complaint. It changes the economics of thought leadership on the one channel most B2B marketers have treated as their default.
The Feed Has Stopped Doing Its Job
LinkedIn worked as a thought leadership channel because a post carried a signal: this person thought this, wrote this, and was willing to put their name on it. That signal is what made a good post worth a follow, a connection request, a sales call.
When 4 in 5 posts in a sample show heavy AI involvement, the signal collapses. A reader scrolling past a well-structured post about supply chain resilience or pipeline forecasting has no reliable way to know if a person spent thirty minutes thinking it through or thirty seconds prompting a model. Once that uncertainty sets in, engagement stops rewarding insight and starts rewarding format: the right hook, the right line breaks, the right contrarian opener. Substance becomes hard to tell apart from performance of substance. Buyers who are trying to shortlist vendors based on who actually understands their problem are left with less to go on, not more.
AI Made Text Cheap. It Cannot Do This
The mechanism that broke LinkedIn is the same one that makes voice more valuable right now. Generating a plausible 800-word post takes a prompt. Generating forty-five minutes of a specific person answering unscripted questions about a real client problem, in their own cadence, with their own hesitations and corrections, takes an actual person who knows the subject. That is much harder to fake convincingly, and audiences are getting better at sensing when it has been faked.
This is the argument for podcasting as a thought leadership format, not as a preference but as a mechanism. A recorded conversation is expensive to fabricate at scale in a way a LinkedIn post no longer is. It’s also why, at B2B Better, a podcast production agency, we tell clients that the format itself is now doing part of the credibility work that used to belong to the platform. You are not just choosing where to publish. You are choosing whether your content can be mistaken for something a machine produced. Six months ago that distinction barely mattered. Now it decides whether a prospect trusts what they are looking at.
Distribution Still Runs Through LinkedIn, Just Differently
None of this means abandon LinkedIn. It means stop treating it as a place to publish original thinking in text form and start treating it as distribution for thinking you did somewhere harder to fake. Clip the podcast episode. Post the transcript excerpt with the audio attached. Let the video pull-quote carry the credibility that the caption alone can no longer supply. The feed rewards native text less than it did two years ago, but it still rewards a founder or VP appearing on video, saying something specific, with a name and a voice attached to it.
The practical shift is sequencing. Record the conversation first. Let LinkedIn distribute proof of it, rather than asking LinkedIn text to carry the argument on its own.
If your LinkedIn strategy right now is a content calendar of solo text posts, the maths has changed against you: your competitor’s AI post and your AI-assisted post are now competing for the same shrinking pool of reader trust. The way out is not writing better prompts. It is producing something a prompt cannot produce, and using LinkedIn to point people at it.
