Study finds over 40% of long LinkedIn posts are fully AI‑generated

If Pangram’s numbers hold up, generative models have already become a dominant author on parts of the social web, at least in the long‑form, high‑visibility content that drives professional reputation and lead generation. LinkedIn emerging as the most saturated platform is particularly striking: it’s where people signal competence and authenticity to employers, investors and peers. When nearly half of long posts are machine‑written, the informational substrate that knowledge workers rely on is quietly shifting.

For the AGI race, this is less about raw capability and more about distribution and data feedback loops. As AI‑authored content saturates professional networks, future models will be increasingly trained on their own outputs, raising concerns about “model collapse” and homogenised discourse. At the same time, the tools that detect and filter AI text, like Pangram’s, will become gatekeepers of what counts as “human” online.

Strategically, platforms like LinkedIn and X face a dilemma: they are pushing their own AI writing aids to increase engagement, while simultaneously needing to downrank generic AI sludge to keep feeds useful. How they resolve that tension will shape both the business case for open‑weight models and the public’s tolerance for ever‑more‑synthetic media.