Flagxiety

Flagxiety (a blend of flag + anxiety) is a term we coined at Diglot to name something our users kept describing but had no word for: the specific dread of submitting work you wrote yourself and having a detector call it machine-generated.

It differs from ordinary plagiarism worry in one key way: you did nothing wrong, and you still can't prove it. Detectors output a probability score, not evidence, and there is no appeal process built into a percentage.

The fear is not irrational. A Stanford study (Liang et al., published in Patterns, 2023) found AI detectors falsely flagged 61% of TOEFL essays by non-native English speakers. Non-native writers tend to use the vocabulary and sentence patterns they were explicitly taught, which makes their prose statistically more predictable — the same property detectors read as machine-like (see perplexity and burstiness).

Flagxiety also changes how people write. A common coping move is to deliberately damage your own prose to look "more human":

Original: The results suggest that the intervention improved reading fluency. "Detector-proofed": So basically the results kind of show the intervention did improve how fluently kids read, I think.

That is a real cost: clearer writing traded away for a lower detection score, plus hours of anxiety around every submission.

The durable answer is not stylistic camouflage. Rewriting yourself to beat a heuristic is a moving target, and it punishes exactly the careful writing ESL students work hardest to produce. What actually helps is process evidence — a verifiable record of how a document was written, edit by edit (see Authorship Certificate). You cannot argue with a percentage, but you can show your work.