Researchers analyzed newly revealed web sites from 2022 via mid-2025 to estimate what proportion used generated textual content and the way this may have an effect on future data on-line.
The proliferation of AI-generated and AI-assisted textual content on the web is feared to contribute to a degradation in semantic and stylistic variety, factual accuracy, and different damaging developments. We discover that by mid-2025, roughly 35% of newly revealed web sites had been categorised as AI-generated or AI-assisted, up from zero earlier than ChatGPT’s launch in late 2022. We additionally discover proof suggesting that will increase in AI-generated textual content on the web carry a couple of lower in semantic variety and a rise in optimistic sentiment. We don’t, nonetheless, discover statistically vital proof supporting the speculation that an elevated price of AI-generated textual content on the web decreases factual accuracy or stylistic variety. Notably, our findings diverge from public notion of AI’s affect on the web.
So it has grown to a couple of third of latest websites that use AI-generated or AI-assisted textual content. That looks like lots?
I’m extra stunned that there didn’t seem like a major change in pretend data or a convergence in fashion.
My concept is that most individuals placing up these generated websites are both experimenting or making an attempt to make a fast buck. Both method, they only take no matter data is given to them by way of a probabilisitic mannequin and neglect about it. They don’t care what the phrases say or how it’s stated, simply so long as it fills area. So the output defaults to largely right statements.
