Nobody wrote the post you just liked

More than 40 per cent of LinkedIn's long-form posts are written entirely by AI.

Pangram Labs scanned 1,002,627 posts across LinkedIn, Medium, Substack, X and Reddit between April 24 and the end of June, using a Chrome extension that reads posts as users scroll past them.

LinkedIn supplied about a third of everything scanned and produced 62 per cent of all the AI content found.

Then the company proved the point itself.

LinkedIn announced in May that it would detect and downrank AI-generated posts, in the name of “keeping conversations real”. It pulled the AI writing assistant out of the post button.

Pangram's report notes the announcement was itself AI-generated.

The pledge to crack down on AI was written by AI.

That's the shape of the whole thing.

The platform where every word is bolted to your real name, your real employer and your real face is the platform where people least want to do the writing.

Reddit — anonymous, throwaway, no career riding on it — came back 98 per cent human on replies.

Across all five platforms, one in four posts longer than 250 words flagged as fully AI-generated. Medium ran second on long-form at 31 per cent, X at 29, Reddit 13, Substack 10.

LinkedIn built the button, too. For years the AI writing tool sat inside the posting interface, one tap from an empty field. It has since been rebranded “Enhance post” and still offers to write for you.

Chief executive Max Spero calls the numbers a lower bound. People who install an AI detector aren't a random sample, and the model is better at recognising human writing than AI writing. He describes the pile-up as a tax on readers' time.

A LinkedIn spokesperson said professionals come to the platform to hear from real people and their unique insights, that the company works to reduce automated and generic content, and that AI can be used to beat the blank page problem.

The blank page was the job.