Two detectors, one scorecard, verdict first. Every number in a comparison is traceable to the same fact base as the main ledger; cells still under verification are marked pending rather than filled.
Copyleaks vs Winston AI: each vendor's own accuracy claim, its footnotes, the price tiers in eesel AI's review, and the one separate real-world test that reports lower results.
GPTZero vs Originality.ai: the RAID and Empirical Study results from eesel AI's review, GPTZero's own 99% claim, free tiers, and which team each tool fits.
Sapling's AI detector API: the eight languages and 200,000-character request limit on its agent page, eesel AI's plan prices, and the 97% figure that remains a vendor claim.
Turnitin vs GPTZero on false positives: five universities that disabled Turnitin's AI check per eesel AI, a 61.3% non-native flag rate, GPTZero's RAID result and vendor claims.
ZeroGPT's free AI checker: the 15,000-character no-signup limit, its 98.4% accuracy and under 1% false-positive claims as vendor figures, and what eesel AI found in testing.