All AI detectors in the index

seventy-two AI detectors sit in this hub, one card each, and every card carries the same fourteen fields in the same order: what the tool detects, the generators its vendor names, the languages it states, input limits, the free tier, paid plans with the date each price was checked, API access, integrations, the vendor's own accuracy claim, the independent studies that tested it, the data retention statement, launch year and ownership, and the rankings the card appears in. A field with no published value has no row. The index never fills a gap with an estimate, and it never prints a score of its own. The rules behind every row are set out on the methodology page, and the AI detector index on the home page shows the 72 cards as one board.

The table below is the hub's own view of that board. Each row names the card's register ID, what the detector covers, how many languages its vendor states, whether a free tier exists, whether the vendor documents an API, whether it names a learning platform, and how many study records in the index tested it. Every one of those values is read from the card, which in turn names the vendor page or the paper it came from and the date it was read. On 8 November 2026 the first sixteen vendors stated between 1 and 30 languages, eleven documented an API, seven named at least one learning platform, and eight had been tested in at least one of the nine study records. Six cards added on 9 November 2026 brought the API count to sixteen, the learning platform count to eight and the tested count to ten; six more that afternoon and ten more in the fourth wave that evening (Smodin, BrandWell, Surfer, Compilatio, Quetext, TruthScan, DetectGPT, Sightengine, Deepware and Sensity) brought the counts to 39 with an API, 14 naming a learning platform and 35 with a study record.

Coverage is wider than text. Thirty-four of the seventy-two cards detect AI text; twenty state image detection, thirteen state video, eleven state voice, four state code and four state music, as each vendor words it, and eight cards, Codequiry, Was It AI, Sightengine, Deepware, Sensity, Codeleaks, Pindrop Pulse and Resemble Detect, cover no text at all; the last three, with Illuminarty, joined on 9 November 2026 in the fifth wave of cards. The modalities slice lays that out as a matrix, and the API slice lists the 39 cards whose vendors document an API with a link to each set of docs.

The evidence behind the cards is uneven, and the hub does not hide that. GPTZero appears in sixteen of the thirty-five study records; the GPT-2 Output Detector in twelve; Copyleaks in nine; Originality.ai and ZeroGPT in seven each; Turnitin and Writer in five each; GLTR, DetectGPT and Pangram in four each; BrandWell and Sapling in three each; Winston AI, QuillBot and Hive in two each; AI or Not, Illuminarty, Compilatio, Scribbr, Undetectable AI and Sightengine in one (nine of those counts rose by one when Pegoraro 2023 was added on 9 November 2026, and eleven records added later that day moved the rest); the other 37 in none. Where a study tested a tool, the card quotes the result as the paper reports it, with its base. Weber-Wulff and colleagues, for example, tested fourteen detectors on 54 documents in 2023 and found every tool below 80% on their averaged accuracy measures (DOI 10.1007/s40979-023-00146-z), and Liang and colleagues reported an average false positive rate of 61.22% across seven detectors on 91 essays by non-native English writers (DOI 10.1016/j.patter.2023.100779). Both records sit in the studies index, and the ranking of independently tested detectors orders the cards by how many records name them.

A card is the place to check one tool; the rankings are the place to compare several on one field. Nineteen rankings are derived from the cards, from study record count and published false positive rate to named integration surfaces, upload formats and, since 9 November 2026, named LMS platforms and the generator models a vendor names. A card enters a ranking only when the field has a published value, so a tool with no study record is listed on its card as having none and is absent from the evidence rankings rather than placed last. The use-case views filter the same cards for business, school districts, agencies, recruiters and code. Since 9 November 2026 four more flags are compiled from the cards for the views: whether a check runs with no account (15 cards), whether a bulk or batch route is stated (15), which security standards the vendor names (10) and which deployments beyond the vendor's cloud and data regions are stated (18), read on the no-account, bulk and batch, compliance and government pages.

Nine cards joined on 10 November 2026 at index version 2026.11.10: Decopy, Content Detector AI, GPTKit, PlagiarismCheck, Isgen, AI Detector Pro, Fast-DetectGPT, Paperpal and Detecting-AI.com, all text detectors; Content Detector AI, GPTKit, Fast-DetectGPT and PlagiarismCheck arrived with records already in the index, and Detecting-AI.com took first place in the input limits ranking at 1,000,000 characters per paid scan. Two views read the cards on who may buy them and how the vendor dates its model: institution-only detectors and model changelogs. An Imagga card researched the same day is held, because the vendor publishes no AI-generated image detector.

Twelve more cards joined later the same day at index version 2026.11.11, the tenth wave: Binoculars (2 records, RAID 2024 and DAMAGE 2025), Monica, Trinka, StealthWriter, Wordtune (1 record, withdrawn), Check For AI (1 record, sunset 2023), Writefull (2 records, discontinued), Passed.ai (closing), the ElevenLabs AI Speech Classifier (voice, the vendor's own models only), Intel FakeCatcher (video, a status record with no access route), Deepfake-O-Meter (image, video and voice through 41 hosted research detectors) and Ircam Amplify (music, eleven named generators). Four withdrawn or closing tools keep a card because studies scored them or readers still meet their scores; the shut-down question carries each notice. A thirteenth researched row, Optic, is held as a duplicate: it was AI or Not's earlier name, and the AI or Not card already records it.

Nine more cards joined at index version 2026.11.12, the eleventh wave, later on 10 November 2026: four tools from Walters 2023's roster that had no card, GPT Radar (1 record, withdrawn in 2026), Grammica (1 record, removed in the January 2026 relaunch), IvyPanda (1 record, a free colour-coded checker) and SEO.ai (1 record, 12% of 42 human papers flagged); AI Writing Check (1 record, Liang 2023, withdrawn in August 2023); two open research methods, RADAR (2 records, RAID 2024 and DAMAGE 2025) and Ghostbuster; and two voice tools that listen to a playing video, the free Hiya Deepfake Voice Detector extension and the on-device McAfee Deepfake Detector. The same wave added the launch year ranking, which dates every card's launch and withdrawal, and the sentence-level highlighting question, which reads 17 cards' statements about marking sentences rather than returning one score.

Every card, whatever the vendor, is recorded on the same fourteen fields from the same kinds of sources and enters a ranking only on a figure the index took from a cited source. The editorial policy states that rule, and the corrections log records every change to a published figure.

No card joined in the twelfth wave on 10 November 2026; every card gained a team field (20 state a seat count or single sign-on, read in the team seats and SSO view), two file type counts were corrected (Resemble Detect 19, Ircam Amplify 9), and the 2027 census counts the 72 cards by status, modality, evidence and launch year.