Can you spot AI-generated content? According to research from the Association for Computing Machinery in 2025, people are only able to suss out AI-generated content 51% of the time. And in the months since that paper was published, AI models have only improved.
A study published this month found that, from a sample size of 1,682 adults, participants were “no better than chance” at differentiating between the human-written and AI-generated stories. Further, in the study, AI-generated stories were even rated better than human-written stories.
But even using AI-detectors to identify AI-generated content is far from a guaranteed gotcha. From misidentifying text as AI-generated to false negatives, these tools are not yet a failsafe.
With AI-generated content becoming harder to identify, earlier this month, Article 50 of the EU AI Act came into effect, introducing rules around marking and labelling certain AI-generated and manipulated content.
And soon, with the introduction of Claude’s new watermark for AI-generated content, consumers will know if creators of content are dabbling with AI.
