What we're actually learning about AI detection
Long-form research and practical guides. No inflated accuracy claims — including our own.
Why Every AI Detector Claims 99% Accuracy (And What Independent Testing Actually Shows)
GPTZero, Originality.AI and Copyleaks all advertise near-perfect detection. Peer-reviewed benchmarks measure 85–92% on edited content. We trace the methodology gap, the false positive problem, and explain why averaging detector scores hides the signal that mattered.
GuideHow to Detect AI-Generated Images in 2026
Image generators got good enough that experts struggle. Which visual tells still work, which ones stopped working, and how automated detection picks up what the eye can't.
GuideHow to Detect AI-Generated Voice and Deepfake Audio in 2026
Voice cloning is cheap, fast and convincing. The spectral artefacts, breathing patterns and prosody gaps that still give it away — and why single-tool detection fails on audio.
GuideHow to Detect AI-Generated Document Fraud in 2026
Fake invoices, forged contracts and altered credentials now pass visual inspection. How error-level analysis and metadata forensics catch tampering, and where verification still needs a human.
Quick guides
How to Spot a Deepfake
Fake CEO video calls, synthetic testimonial ads and AI-generated news clips — the three patterns causing real losses, and what to check by eye and ear.
GuideInvoice Fraud: How to Spot a BEC Scam
Copied vendor invoices, urgent bank-detail redirects and lookalike supplier domains — how to catch them before money moves.
GuideVoice Cloning Scams: How to Spot AI Voice Fraud
Cloned executive requests, fake emergency calls and synthetic endorsements — what to listen for and how to verify before you act.
Try it on something of your own
Five detection modes — text, images, video, speech and documents. Three free checks a day, no signup required.
Try it free