Index update, Q3 2027: twenty-one cards added, two fields added, two rows held, three log lines
The AI detector index grew from 51 cards at version 2026.11.10 to 72 cards at version 2026.11.13, with 21 cards added in two waves on 10 November 2026 (12 at version 2026.11.11 and 9 at 2026.11.12), 2 fields added to every card's rank data (a launch year metric at 2026.11.12 and a team field at 2026.11.13), 1 ranking added that sorts on the first of them, 5 views added, 2 researched rows held unpublished, and 3 lines filed in the corrections log, by the index data file and the log, both checked on 10 November 2026. Five study records were added in the same period, taking the studies index to 35; the Q3 2027 studies digest covers them. Every addition below is dated and named by slug, so a reader who quoted a count from the Q2 2027 index update can see what has moved since.
Checked 10 November 2026 for the quarter ending 30 September 2027; the next scheduled recheck of every card is 8 December 2026.
This is the fourth quarterly index update and the last of the twelve-month content plan. It covers the tenth, eleventh and twelfth content waves, versions 2026.11.11 to 2026.11.13, and lists cards by version, fields by name, rankings and views by slug, held rows by reason and corrections by log line. The 2027 census, written the same day, counts the whole index rather than the quarter. This page sits in the updates hub, and the AI detector register shows every card the versions below added.
| Version | Date | Cards added, by slug | Count after |
|---|---|---|---|
| 2026.11.11 | 10 November 2026 | binoculars, monica, trinka, stealthwriter-detector, wordtune-detector, check-for-ai, writefull-gpt-detector, passed-ai, elevenlabs-ai-speech-classifier, intel-fakecatcher, deepfake-o-meter, ircam-amplify (AIDI-D-0053 to 0060 and 0062 to 0065); AIDI-D-0061 held; 2 comparisons and 1 view added | 63 |
| 2026.11.12 | 10 November 2026 | gpt-radar, grammica, ivypanda, seo-ai, aiwritingcheck, radar, ghostbuster, hiya-deepfake-voice-detector, mcafee-deepfake-detector (AIDI-D-0066 to 0074); 5 study records, 1 question and 1 ranking added | 72 |
| 2026.11.13 | 10 November 2026 | None; the team field written to every card, two file type counts corrected, 4 questions, 2 comparisons, 4 views and 3 updates added | 72 |
Versions and slugs as recorded in the index data file at each build; the register ID of every card is in its front matter and on its plate.
Which cards were added, and what kind are they?
Twenty-one, in two waves on 10 November 2026. The tenth wave's twelve were eight text detectors and four media detectors: Binoculars, the open ICML 2024 method with two benchmark records (github.com/ahans30/Binoculars, read 10 November 2026); Monica, Trinka and StealthWriter, three live vendor tools; Wordtune, Check For AI, Writefull and Passed.ai, four tools withdrawn, discontinued or closing, carded from archived pages so that the records that tested them have somewhere to point; and the ElevenLabs AI Speech Classifier, Intel FakeCatcher (a status record with no access route), the University at Buffalo's Deepfake-O-Meter and Ircam Amplify, the index's first music-only card (ircamamplify.com, read 10 November 2026; Binoculars card; Ircam Amplify card; vendor and archive pages read 9 and 10 November 2026). The eleventh wave's nine were the four Walters 2023 tools that had no card (GPT Radar, Grammica, IvyPanda, SEO.ai), AI Writing Check from Liang 2023's roster, the open methods RADAR and Ghostbuster, and two voice tools, Hiya's Chrome extension and McAfee's on-device detector (Walters 2023 record; Liang 2023 record). Ten of the 21 arrived with a study record already in the index (Binoculars, Writefull, Wordtune, Check For AI, RADAR, GPT Radar, Grammica, IvyPanda, SEO.ai, AI Writing Check) and eleven with none.
Two rows were researched and held. Optic (ADI-153, AIDI-D-0061) was AI or Not's earlier name, and the AI or Not card with the Ha and Passananti 2024 record already carries the tool, so a second card would duplicate one product; the register ID stays reserved and the row reads "held: duplicate of ai-or-not". The Imagga hold of the ninth wave stands. Both wait on a decision to strike or keep the row, as the 2027 census lists.
Which fields were added to every card?
Two. At version 2026.11.12 the eleventh wave added launch_year to the metrics table, lower first, so that the launch year ranking (AIDI-R-20) could sort on a card field that already existed: 29 of 72 cards state a year, from Codequiry's 2017 to Turnitin Clarity's 2025, and the index enters no year from an archive's first capture. At version 2026.11.13 the twelfth wave wrote a team field to every card, the flag shape the wave 8 fields use: seats_stated is true when the plans field states a seat count, a per-seat price or a team-member count, and sso when any field names SSO, SAML or SCIM; 17 cards state seats, 6 state SSO and 20 state at least one, and the team seats and SSO view reads the field. No ranking sorts on it.
One vocabulary value was added rather than a field: McAfee's "on-device (Windows AI PC NPU)" joined the deployment field's vocabulary at 2026.11.12, the first on-device value in the index, and the government view and the offline question carry it. The data rulings of the two waves are recorded in the merge scripts and the page map: among them, the GPT family counts once on every card's generator count (Writer recounted from 2 to 1 for parity), a Hugging Face Space is not a product API (RADAR), a demo's refusal message is not a stated minimum (Ghostbuster, GPT Radar), and a withdrawn tool carries no no-account flag (Grammica, AI Writing Check, the Writer treatment).
Which rankings and views changed?
One ranking was added, launch year, at 29 of 72. Of the nineteen existing rankings, sixteen gained members in the tenth wave and fourteen in the eleventh, with no existing order moved in any of them except Writer's generator recount: independently tested went from 25 to 35 (Binoculars, Writefull and RADAR at 2; Wordtune, Check For AI, GPT Radar, Grammica, IvyPanda, SEO.ai and AI Writing Check at 1; the GPT-2 Output Detector from 12 to 13 with Rashidi 2023), false positive rate from 18 to 26 (Binoculars at 0, Writefull and Check For AI at 5.6%, RADAR at 3.48%, SEO.ai at 12%), languages from 27 to 31, input limits from 21 to 28, data retention from 17 to 21, adversarial robustness from 13 to 15 (Binoculars 80.3 fifth, RADAR 67.3 seventh), false negative rate from 17 to 20, claim vs evidence gap from 11 to 12 (SEO.ai plus 26.4), platform integrations from 46 to 52, file types from 27 to 30, minimum input from 10 to 12, generators covered from 44 to 56, non-English evidence from 16 to 19, LMS coverage from 13 to 14 and free checks per day from 6 to 7; free tiers, API pricing and code languages were unchanged in membership. In the twelfth wave the file types ranking's counts for Resemble Detect (17 to 19) and Ircam Amplify (7 to 9) were corrected under the ranking's stated token rule (docs.resemble.ai/detect/create, read 9 November 2026; docs.ircamamplify.io, read 10 November 2026), which moved Ircam Amplify from 13th to 12th, the one order change of the quarter made by a correction rather than an addition.
Five views were added: the self-publishing view in the tenth wave (12 cards, the KDP Content Guidelines fetched live) and, in the twelfth, the customer support QA view (2 cards state all four fields), the team seats and SSO view (20), the handwriting and OCR view (3 state OCR, 0 state OCR of a scanned PDF) and the mobile apps view (5). Four questions and two comparisons were added in the twelfth wave and one question in the eleventh, from student plans to Pindrop vs Resemble Detect, and the index now holds 237 pages: the 49 of the launch set, the three roadmap hubs and 185 of the 189 register rows, with the four held.
Which corrections were filed?
Three lines in the corrections log since the Q2 update, one per wave. The wave 10 line (version 2026.11.11) records the twelve cards, the Optic hold, five data rulings (Binoculars' bulk flag off, Wordtune's bundle-only values and Writefull's archived retention statement left unranked, Deepfake-O-Meter's API tier value left unranked, Writer's generator recount), the five records that gained a card row, the four rows added to the shut-down question and three vendor figures written with the accuracy word bracketed for the claim gate. The wave 11 line (2026.11.12) records the nine cards and five records, the GPT-2 Output Detector's thirteenth record, the four records that gained a card row, the RADAR, Ghostbuster, GPT Radar, Grammica, AI Writing Check and McAfee rulings, and two register corrections (the pathology study's authorship and DOI; 17 studies in the Originality.ai post, not 14). The wave 12 line (2026.11.13) records the team field, the Turnitin typed-text sentence added to the card from the FAQ re-read of 10 November 2026 (guides.turnitin.com), and the two file type recounts. No price, limit or study figure quoted on a card has been corrected since launch; the two file type counts are the first field values changed by a correction, and the first scheduled recheck of prices and free tiers on all cards is 8 December 2026.
Questions about the Q3 2027 index update
How many AI detectors were added to the index since the Q2 2027 update?
Twenty-one, all on 10 November 2026: twelve at version 2026.11.11 (Binoculars to Ircam Amplify) and nine at 2026.11.12 (GPT Radar to McAfee Deepfake Detector), taking the index from 51 to 72 cards; two researched rows, Optic and Imagga, are held.
Which fields did every card gain?
A launch year metric at version 2026.11.12, which the launch year ranking sorts on (29 cards state a year), and a team field at 2026.11.13 with seats and SSO flags (20 cards state at least one), which the team seats and SSO view reads.
Were any published figures corrected in the quarter?
Two field values: Resemble Detect's upload format count (17 to 19) and Ircam Amplify's (7 to 9), recounted on 10 November 2026 under the file types ranking's stated rule after the wave 12A note found that the counting vocabulary had dropped aif, aiff, amr and 3gp; no price, limit or study figure changed, as the corrections log records.
Sources
- The index data file, versions 2026.11.10 to 2026.11.13 (read 10 November 2026)
- Binoculars repository: https://github.com/ahans30/Binoculars (read 10 November 2026)
- Ircam Amplify: https://www.ircamamplify.com/ (read 10 November 2026)
- RADAR model card: https://huggingface.co/TrustSafeAI/RADAR-Vicuna-7B (read 10 November 2026)
- McAfee Deepfake Detector: https://www.mcafee.com/ai/deepfake-detector/ (read 10 November 2026)
- Turnitin capabilities FAQ: https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs (read 10 November 2026)
- Resemble Detect API reference: https://docs.resemble.ai/detect/create (read 9 November 2026)
- Ircam Amplify audio input methods: https://docs.ircamamplify.io/api-docs/aimd-guide/audio-input-methods (read 10 November 2026)
- Walters 2023: https://www.degruyterbrill.com/document/doi/10.1515/opis-2022-0158/pdf (read 8 November 2026)
- Liang et al. 2023: https://arxiv.org/pdf/2304.02819v3 (read 8 November 2026)
Revision log: 2026-11-10: update created at index version 2026.11.13 (twelfth wave).
Related pages: Studies digest, Q3 2027: five records added, one headline figure and base each, two closed-access papers still held, 2026 review: every published AI detector false positive figure in the index, AI detectors ranked by launch year, with every change of hands and withdrawal the cards record, How accurate are AI detectors? Every study, indexed and Index updates: dated reviews of the AI detector data.