Early-Signal Detector

task · first seen 5 Oct 2026

AI-generated text detection

Also: AI tells, slop scanner, AI detection, AIGC detection, AIGT detection, AI slop removal, AI slop detection, AI-text detection

Identifying whether a piece of text was produced by a language model.

AI-generated text detection is the task of deciding whether a passage was written by a language model. The sources show steady incremental work on a long-running problem: a Bayesian fusion method built on Fast-DetectGPT, soft-prompt detectors, a PAN 2025 shared-task entry, an evasion attack using entropy-adaptive decoding, and a Google DeepMind repo for interpretable detectors based on sparse autoencoders. Newer angles include detecting AI-originated ideas rather than wording (a UMass-led paper and a Penn State paper on insight provenance), evaluation on authentic partial AI use in PISA 2025 student essays, and community tools for spotting bot comments or removing AI style tells.

The subfield has been active since at least 2023, so the rise looks like normal publication volume rather than a new idea. The shift toward idea provenance and toward partial, mixed authorship may matter for academic integrity and peer-review policy if it holds up.

Analyst note (established), 17 Nov 2026.

Mentions per week

0719 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Rank 42 on 14 Nov 2026 · 25 mentions in 28 days · 25 independent groups · papers → code → discussion · first seen 6 weeks ago · acceleration +0.63

Earliest sources

Latest sources

Mentioned together with

agent skillszero-shot detectionsparse autoencoderitem response theoryadversarial training for debiasingresearch integrityentropy-adaptive samplinghardware compression acceleratorformat counterfactualsAI text humanization