No one has a trustworthy count of workplace AI writing, but workers estimate 15% of what they receive is AI-generated and unhelpful.

This explores how much of what people send and receive at work is now written by AI. The short answer: the corpus has no single, trustworthy percentage for workplace communication as a whole. What it has are three partial measurements of different things, and each one shows how hard the question is to answer.

The closest workplace figure comes from a September 2025 survey of 1,004 U.S. desk workers. They estimated that 15.4% of the work they receive is AI-generated *and unhelpful* How much work that employees receive is actually unhelpful AI content?. That figure counts only the low-quality share, so the total amount of AI content is higher, and it is a guess by the people receiving the work, not a count. The cost is what stands out: workers report spending close to two hours on each instance, which is longer than it would have taken the sender to do the job properly. For professional writing in public, a detector run by Originality.ai flagged 81% of a fixed sample of 5,000 LinkedIn posts from July 2026 as likely AI, up from about 50% in late 2024 How much LinkedIn content is AI-generated right now?. For the wider web, about 35% of new websites were AI-generated or AI-assisted by mid-2025 How much of the internet is AI-generated now?.

These numbers don't agree, and there's a reason for that. The LinkedIn figure is whatever a detector decides counts as AI. It works as a trend line for one sample, not as a census of the platform. You might hope people could check the detectors themselves, but a review of 30 studies found that humans spot AI-generated text, images, and voices at about chance level Can people reliably spot content made by AI?. The line between "AI-written" and "human-written" also blurs in practice. Writers edited AI-drafted paragraphs only 23% of the time, and their edits left the text about 96% the same Do writers actually edit AI-generated text before publishing?. So a lot of mail that is "from a colleague" is mostly the model's wording. Where adoption goes deepest also varies by job: AI use built into actual workflows is concentrated in information-heavy work Where have workers actually delegated tasks to AI?, so any single average would hide large differences between occupations.

The corpus suggests a better question than "what percentage?": what happens to communication when a large share of it has no one behind it? Several notes argue that AI text carries the outward signs of a message without the social act of one person addressing another Does AI really communicate or just distribute information?. Readers end up doing the interpretive work for both sides Does AI generate genuine utterances or just text patterns?. On professional platforms, AI posts collect likes but don't start replies Why do AI posts get likes without inviting conversation?. That is the same pattern as the survey's two-hour cleanup: the sender saves effort and the receiver pays for it.