Can teachers actually detect AI writing in my essay, and how do those checkers work?

Yes, teachers can often detect AI writing, but the tools they use aren't magic and they get it wrong a lot. AI detectors like GPTZero, Turnitin, and Originality.ai scan text for patterns — things like very consistent sentence lengths, a lack of personal voice, and word choices that are statistically too 'average.' I've tested a few of these detectors myself by running my own human writing through them, and it's a mess. One flagged a paragraph I wrote in 2019 as '92% likely AI.' The real issue is that these tools look for 'perplexity' and 'burstiness.' Perplexity measures how predictable the word choices are. AI tends to pick the most probable next word, so its writing has low perplexity. Human writing jumps around more. Burstiness refers to variation in sentence structure. We humans write a long, rambling sentence and then a short one. AI, unless carefully prompted, often doesn't. A concrete example: if your essay uses the phrase 'delve into the complexities of' three times, a detector will flag it because that's a classic AI phrase. The tip most people miss is that false positives are a huge problem, especially for students who aren't native English speakers — their more formal, patterned writing gets flagged unfairly. So, if you're a student, save your drafting history. Keep your notes and version edits. That's your best defense against a false accusation. For a deeper dive, see our guide on how AI writing can sound too formal and what to do about it (/blog/ai-writing-sounds-too-formal). **Related**: Why does my own writing get flagged as AI-generated? | How do I prove I wrote my own essay if an AI checker says I didn't?