Cantrill: LLM-Written Posts Are Obvious, and They're Killing Your Credibility

Bryan Cantrill argues that the flood of AI-generated posts on LinkedIn is easy to spot and actively counterproductive. The stylistic fingerprints of LLM prose — emoji clutter, one-sentence paragraphs, formulaic ‘it’s not just X but also Y’ constructions, and reflexive em-dashes — read as unmistakable tells to a growing audience that has seen enough machine-written text to recognize it. LinkedIn’s own ‘rewrite it with AI’ prompts only accelerate the trend.

The real cost, in his framing, isn’t just clumsy prose but a collapse of trust. When a reader can tell a post wasn’t written by the person it’s attributed to, they stop being able to distinguish genuine insight from generated filler and simply stop reading. Cantrill likens the effect to an unnoticed wardrobe malfunction: plenty of people register it, but nobody bothers to point it out.

He’s careful to separate use cases rather than reject the tools outright. LLMs, he says, are genuinely valuable for brainstorming, comprehension, and editing — but they are poor writers and, crucially, not you. His advice is to lean on them behind the scenes while keeping your own voice in the finished product, on the theory that authentic writing is what earns a message the credibility to be read at all.