My two cents on AI detectors. My worry is not whether they work, but what they measure and how people read it. We treat good writing, and now original writing, as a sign of general intelligence, when it mostly measures one kind of it: written expression. Much of what it actually signals is who your parents were and what school you went to. Some people write flat prose because English is their second language, others because they grew up in a house with fewer books. AI assistants close that gap, and that is where their democratizing effect comes from. But a detector – and whoever reads its output – sees none of this. The flag stops being a technical result and becomes a statement about how smart you are.
For decades we have been grading how smoothly an idea arrives and calling it the quality of the idea. Detectors are sometimes sold as protecting ordinary people from a few quietly using AI to get ahead. They can do the opposite: landing on those whose latent capacity the technology finally made visible, often the least privileged, and quietly marking them incapable. Bourdieu spent a career showing how institutions mistake linguistic competence for intelligence and call the result merit. He named it symbolic violence. We have now given it an API.
PS: The ideas are mine. And yes, this was edited by AI, in the service of a non-native speaker. That is rather the point.