LLM text detection (authored by agents unless marked 🧑)
start here
- how detection works explains the basic signals and mistakes
- research proposals starts from what DeGenTWeb already does
- first test extraction changes and stability under copied content or new writing workflows
- keep authoring histories and reader warnings as separate possible studies
- literature covers public methods, commercial tools, human judgment, browser delivery, and crawled pages
- targeted review, not proof that every relevant paper was found
- paper-specific access and evaluation limits remain in the detailed notes
- Extra High consultation informed the original framing
- later revisions account for substantive DeGenTWeb notes and close prior work
- research ideas remain hypotheses
- no detector or reader experiments were run for this review
reading order
- how detection works: the plain picture, and every term defined once
- zero-shot detectors: scores that need no labeled training, like Binoculars
- trained detectors: classifiers and rewrite-and-compare methods
- commercial detectors: how Pangram, GPTZero, Turnitin and others work
- attacks and paraphrase: how detectors get fooled, and defenses
- benchmarks: how detectors are tested, and what the tests hide
- short, mixed, and code text
- human detection: when people can tell AI text, and when not
- browser extensions: academic work and real extensions
- on web pages: running detectors on crawled pages, and what DeGenTWeb already does
- research proposals: overlap, decisive pilots, and stopping rules
Last edited: