Index update, Q4 2026: twelve cards added, five fields added, four corrections filed
The AI detector index grew from 16 cards at version 2026.11 on 8 November 2026 to 72 cards at version 2026.11.13 on 9 November 2026, with 12 cards added in two waves of six, 5 fields added to every card's rank data, 9 study records added to the studies index and 4 lines filed in the corrections log, by the index data file and the log, both checked on 9 November 2026. Every addition below is dated and named by slug, so a reader who quoted a count from the index in its first week can see exactly what has moved since.
This is the first quarterly index update and it covers the launch quarter, from the first build on 8 November 2026 to the third content wave on 9 November 2026. It lists cards by version, fields by name, corrections by log line, and the rankings whose membership or order changed as a result, in the same fields the cards use. The studies digest for the same quarter covers the nine records added; the 2026 false positive evidence review sets every false positive figure in one table. 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 | 8 November 2026 | quillbot, grammarly, scribbr, winston-ai, pangram, gptzero, copyleaks, phrasly, sapling, justdone, originality-ai, zerogpt, proofademic, walter-writes, hive, turnitin (AIDI-D-0001 to 0016) | 16 |
| 2026.11.2 | 9 November 2026 | gltr, aidetectplus, gpt-2-output-detector, codequiry, wasitai, reality-defender (AIDI-D-0017 to 0022) | 22 |
| 2026.11.3 | 9 November 2026 | None; 8 study records and 5 rank fields added | 22 |
| 2026.11.4 | 9 November 2026 | writer-com, undetectable-ai-detector, ai-or-not, grammarly-authorship, turnitin-clarity, synthid (AIDI-D-0023 to 0028) | 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?
The first wave of 9 November 2026 added six text and code tools, two of them open-source research releases: GLTR and the GPT-2 Output Detector, both from 2019, joined the independently tested ranking at once because the 2023 and 2024 papers had already tested them (open-source view). AIDetectPlus is a text detector with a per-sentence report, Codequiry a source-code detector, WasItAI an image-only checker and Reality Defender a deepfake platform for video, image and voice.
The third wave, later the same day, added six cards of a different shape. Writer's AI Content Detector was withdrawn on 22 December 2025 (archived detector page, read 9 November 2026) and is carded because four 2023 studies cite it (which detectors were shut down?). Undetectable AI's detector is one surface of a humanizer vendor and states 50+ languages, the highest count in the index (undetectable.ai, read 9 November 2026) (languages ranking). AI or Not detects image, video, voice, music and text and carries the image result that Ha and Passananti 2024 recorded under its earlier name, Optic (arXiv 2402.03214, read 9 November 2026). Grammarly Authorship and Turnitin Clarity record a writing process and return no score (grammarly.com/authorship; Turnitin Clarity, read 9 November 2026), and SynthID checks for a Google watermark (synthid.com, read 9 November 2026); all three have empty fields where a classifier carries a rate, and is an AI detector itself an AI? explains the method difference.
Which fields were added to every card?
Five, all on 9 November 2026 at version 2026.11.3, each written by the wave 2 merge into the rank_fields block of every card with the same keys and a null form where no indexed figure qualifies: published_fnr (lowest published false negative rate, with study, base, threshold and content type), robustness (highest share of paraphrased or edited AI text still detected, with the attack), retention_days (days the policy says submissions are kept, with the clause), modality_records (count of records on image, voice, video, code or music) and claim_gap (vendor accuracy claim minus the best study accuracy, in points). Each field feeds one ranking: false negative rate, adversarial robustness, data retention, studies by content type and claim vs evidence gap. The metrics table in the data file gained a direction for each, and the methodology page gained a row per ranking with its field, direction, qualification and tie rule.
The third wave filled those fields for the six new cards and applied two rules the first wave had set. AI or Not's 24.47% false positive rate on 280 artworks is an image result, so its published_fpr value is null, as Hive's is, and the figure travels on the false negative and content-type rankings instead; Writer's attacked figures come from Weber-Wulff 2023 Table 5, which the robustness ranking takes from no card, so its robustness value is null and the card's prose carries the counts.
Which rankings changed membership?
Nine of the ten, by the computed order at version 2026.11.13. Independently tested went from 8 of 16 at launch to 14 of 28, with Writer (4 records) and AI or Not (1) the last to join. False positive rate went from 8 to 12 ranked, Writer joining at 5.6%, with Undetectable AI's vendor 0% added to the unranked vendor table. Free tiers went from 3 to 4, AI or Not at 1,000,000 words a month taking second place. Languages went from 12 to 13, Undetectable AI's 50+ taking first place from Copyleaks and Originality.ai's 30+. Data retention went from 5 to 9, with AI or Not and Undetectable AI at 0 days, SynthID at 1 day and Grammarly Authorship at 3,650 days, the longest stated period in the index. Studies by content type went from 14 to 16 and gained a second card with a record, AI or Not. False negative rate went from 10 to 12 and claim vs evidence gap from 7 to 8, AI or Not joining both. Adversarial robustness stayed at 8 and input limits, a hand-read table, was not re-ranked. Every change is dated in the revision log at the foot of the ranking it touched.
Which corrections were filed?
Four lines in the corrections log as of 9 November 2026, all additions rather than corrections to a quoted figure: the creation line of 8 November 2026 (16 cards, 8 records); the six-card line of 9 November 2026 (AIDI-D-0017 to 0022, with the independent testing and false positive rankings at 10 of 22); the wave 1 line (version 2026.11.2: Chaka 2023, the input limits ranking, the admissions view, four questions, three comparisons and two hubs, with Copyleaks moving to third in the false positive ranking on the tie rule); and the wave 2 line (version 2026.11.3: eight records, five rankings, two questions, one update and the updates hub, with Pangram's false positive rate moving from the vendor figure to Jabarian and Imas 2025). A fifth line for version 2026.11.4 is filed with this update. No price, limit or study figure on any card has been corrected since launch; the first scheduled recheck of prices and free tiers on all cards is 8 December 2026, and any change it produces will be its own line.
The next update in the series, the Q1 2027 index update, written 9 November 2026, carries the fourth to sixth waves: fourteen more cards, four more fields and the four rankings that read them.
Questions about the Q4 2026 index update
How many AI detectors were added to the index in Q4 2026?
Twelve, all on 9 November 2026, in two waves of six: GLTR, AIDetectPlus, the GPT-2 Output Detector, Codequiry, WasItAI and Reality Defender at version 2026.11.2, then Writer, Undetectable AI, AI or Not, Grammarly Authorship, Turnitin Clarity and SynthID at version 2026.11.4, taking the index from 16 to 72 cards.
Which fields did every card gain?
Five rank fields on 9 November 2026: published false negative rate, adversarial robustness, retention days, modality records and claim gap, each with its study or policy source and a null form where no figure qualifies, each feeding one of the five rankings added at version 2026.11.3.
Were any published figures corrected in the quarter?
No. The four log lines record additions and the re-sorting they caused; no price, limit or study result quoted on a card has changed since 8 November 2026. The first scheduled recheck is 8 December 2026 (corrections log).
Sources
- The index data file, versions 2026.11 to 2026.11.4 (read 9 November 2026)
- Grammarly Authorship: https://www.grammarly.com/authorship (read 9 November 2026)
- Turnitin Clarity: https://www.turnitin.com/products/feedback-studio/clarity (read 9 November 2026)
- SynthID Detector: https://synthid.com/ (read 9 November 2026)
- Ha, Passananti et al. 2024: https://arxiv.org/abs/2402.03214 (read 9 November 2026)
- Weber-Wulff et al. 2023: https://arxiv.org/pdf/2306.15666 (read 9 November 2026)
- Writer AI Content Detector, archived 9 December 2025: http://web.archive.org/web/20251209115654/https://writer.com/ai-content-detector/ (read 9 November 2026)
- Undetectable AI detector: https://undetectable.ai/ai-detector (read 9 November 2026)
- Grammarly Authorship support article: https://support.grammarly.com/hc/en-us/articles/29548735595405 (read 9 November 2026)
Revision log: 2026-11-09: update created at index version 2026.11.4; 2026-11-09: link to the Q1 2027 index update added.
Related pages: Studies digest, Q4 2026: the nine evaluations added since launch, with one headline figure each, Index update, Q1 2027: fourteen cards added, four fields added, four rankings added, three log lines, AI detectors ranked by independent testing Two records added in the fifth wave on 9 November 2026, Pegoraro 2023 and Fleckenstein 2024, changed nine counts and none: Pegoraro 2023 tested ZeroGPT, GPTZero, Copyleaks, Originality.ai, Writer, Content at Scale (BrandWell), DetectGPT, the GPT-2 Output Detector and GLTR, lifting GPTZero to 12, the GPT-2 Output Detector to 10, Copyleaks to 7 and Originality.ai and ZeroGPT to 6; Fleckenstein 2024 tested 289 teachers and no software, so it counts for no card ([Fleckenstein 2024 record](/studies/fleckenstein-2024-teachers/)). Illuminarty, one of four cards added in the same wave, joined with its one image record from Ha and Passananti 2024, so 18 of 72 cards now qualify., How accurate are AI detectors? Every study, indexed and Index updates: dated reviews of the AI detector data.