Index update, Q2 2027: nine cards added, nine fields added, one card held, three log lines

Nine text detectors, all at version 2026.11.10 on 10 November 2026, researched from vendor pages read on 9 and 10 November 2026. Four arrived with study records already in the index: Content Detector AI, which Walters 2023, Kar 2024 and Alshammari and Rao 2025 had tested as a card-less tool and which now redirects to a Moxby browser Mod; GPTKit, the only detector of eight at 0 false positives in 114 human submissions in Orenstrakh 2023 and the lowest post-manipulation accuracy of seven in Perkins 2024; PlagiarismCheck, 13th of 14 in Weber-Wulff 2023; and Fast-DetectGPT, the ICLR 2024 open-source method RAID 2024 ran with its repository defaults (Walters 2023; Orenstrakh 2023; Weber-Wulff 2023; RAID 2024). Five arrived with none: Decopy, Isgen, AI Detector Pro, Paperpal and Detecting-AI.com, each a vendor product with its own plans, limits and claims and no indexed outside result (decopy.ai, isgen.ai, aidetector.pro, paperpal.com, detecting-ai.com, read 9 and 10 November 2026).

One card was researched and held. Imagga (register row ADI-144, AIDI-D-0052) was planned as an AI image detection card, and the research of 10 November 2026 found that the vendor publishes no AI-generated image detector on any live page, documentation section or Wayback Machine capture: its products are image tagging, categorisation, adult content detection, face recognition and related computer vision services (imagga.com, docs.imagga.com, read 10 November 2026). The card copy is kept in the research folder, the register ID stays reserved, no page is published and the register row reads "held: no AI detector published", pending a decision to strike the row or keep it; the index does not publish a detector card for a product that does not exist.

Nine, all on 9 November 2026, each written by a merge script into the rank_fields block of every card with the same keys and a null form where no figure qualifies. At version 2026.11.8, five fields that each feed one ranking: lms_count (named LMS platforms of nine), free_per_day (the vendor's per-day free allowance, in checks), generator_count (distinct generator models or families the vendor names), nonenglish_records (study records that print a result on non-English or non-native text) and code_language_count (programming languages named for code detection). At version 2026.11.9, four flags that the month 8 views filter on and no ranking sorts: no_account (the vendor states a check runs with no sign-up), bulk (a bulk upload, batch API, ZIP or CSV route), compliance (named security or privacy standards from a vendor page) and deployment (on-premise, self-hosted, private cloud or offline modes and a named data region). The methodology page gained a row per ranking, and each field carries the source URL and read date of the card field it was read from.

No study-fed field moved by rule in the period, because no record was added; the nine new cards brought their own values. Three data rulings on those values are recorded in the merge script and the page map: PlagiarismCheck's and Isgen's retention fields are null because each vendor publishes two statements that conflict (a product page saying submissions are not retained, a policy or pricing page stating a period), the reading the index gave Sensity and Pindrop; PlagiarismCheck's LMS count is 6 rather than the 7 the vendor names, because Populi falls outside the nine-platform list the ranking counts; and GPTKit's claim gap pairs its 93% with the 87.50% mixed-set accuracy of Orenstrakh 2023, with the 18-point alternative against the 75.00% ChatGPT-only figure noted on the card. Content Detector AI and Detecting-AI.com carry no false positive rate: no record prints a human-text count for the first, and the second's "under 1%" is a bound with no base.

Five rankings were added at version 2026.11.8, one per field: LMS coverage (11 of 42 at creation, Compilatio first at 7 of 9), free checks per day (4 of 42, Sightengine first at 500), generators covered (36 of 42, Resemble Detect first at 62 named), non-English evidence (14 of 42, GPTZero first with 5 records) and code languages (3 of 42, Codequiry first at 12). At version 2026.11.10 the nine cards joined fifteen rankings and changed the leader of one: input limits now puts Detecting-AI.com first at 1,000,000 characters per scan on every paid plan, ahead of AI or Not's 500,000 per API request, and lists 21 of 72 with Decopy, GPTKit, PlagiarismCheck, AI Detector Pro and Paperpal also joining and Isgen and Content Detector AI in the unlimited table. Independently tested went from 21 of 42 to 35 of 72 (Content Detector AI 3, GPTKit 2, Fast-DetectGPT 1, PlagiarismCheck 1). False positive rate went from 15 to 18 (GPTKit and PlagiarismCheck at 0, Fast-DetectGPT at 23.2%). Languages went from 21 to 27, Isgen second at 80. Data retention went from 13 to 17 (three zeros and AI Detector Pro at 90 days). Adversarial robustness went from 11 to 13 (Fast-DetectGPT 71.8%, GPTKit 30%). False negative rate went from 16 to 17 (PlagiarismCheck 52.8%). Claim vs evidence gap went from 9 to 11, PlagiarismCheck at plus 60.23 points the widest gap in the index. Platform integrations went from 38 to 46, file types from 20 to 27, API pricing from 10 to 12, LMS coverage from 11 to 13, free checks per day from 4 to 6, generators covered from 36 to 44 and non-English evidence from 14 to 16. Free tiers, minimum input, studies by content type and code languages were unchanged. The order of the cards already ranked did not move in any ranking; the new cards inserted among them. Every change is dated in the revision log at the foot of the ranking it touched.

Nineteen question pages, ten views and two comparisons. The seventh wave added eleven questions, from the minimum word count to Canvas without Turnitin; the eighth added eight, from Google Docs links to deleting your data, five use-case views (law firms, grant reviewers, newsrooms, localisation teams, government) and three feature views (no account, bulk and batch, compliance). The ninth added two feature views, model changelogs (10 cards with a dated release note, model card or version post) and institution-only detectors (5 cards with no self-serve purchase), and two comparisons, Scribbr vs QuillBot (two Learneo brands with no shared record) and Reality Defender vs Sensity (two deepfake vendors with no record at all). The index now holds 193 pages.

Three lines in the corrections log since the Q1 update. The wave 7 line (version 2026.11.8) carried five corrections: every character-to-word conversion sitewide moved from 6 to 5.5 characters a word, one rate; Compilatio's image cell on the content type matrix cleared to follow its card; the false positive evidence review gained DAMAGE 2025, Pegoraro 2023 and the Ha and Passananti 2024 image rows; the Jabarian and Imas 2025 record gained its GPT-4.1 false negative row; and stale literal counts were corrected on the questions hub and three rankings. The wave 8 line (version 2026.11.9) carried the Walter Writes integrations correction from "Docs" to "Google Docs", the SynthID Nature DOI, the Weber-Wulff 2023 record's four missing Table 5 rows and the Table 9 and 10 counts for all 14 tools, the shut-down question's roster count (9 of 14 carded, not 6), the GPTZero FAQ figure set beside its leaderboard entry, Turnitin's launch and model dates, and the accuracy-claim count rule (a percentage figure, not any digit: 30 of 42, previously 37). The wave 9 line (version 2026.11.10), filed with this update, records the nine cards, the Imagga hold, the three data rulings above, the seven study records that gained a card row (with the Weber-Wulff roster count moving to 10 of 14 and the Walters count to 10 of 16), and the ContentDetector.ai redirect added to the shut-down question from the Wayback CDX record. No price, limit or study figure quoted on a 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 is the Q3 2027 index update, covering the tenth, eleventh and twelfth waves (21 cards, two fields, five views) and the two file type corrections of 10 November 2026.

• The index data file, versions 2026.11.7 to 2026.11.10 (read 10 November 2026)
• Decopy AI detector: https://decopy.ai/ai-detector/ (read 9 November 2026)
• Isgen AI detector: https://isgen.ai/ (read 9 November 2026)
• AI Detector Pro: https://aidetector.pro/ (read 10 November 2026)
• Paperpal AI detector: https://paperpal.com/tools/ai-detector (read 10 November 2026)
• Detecting-AI.com: https://detecting-ai.com/ (read 10 November 2026)
• GPTKit: https://gptkit.ai/ (read 9 November 2026)
• PlagiarismCheck AI detector: https://plagiarismcheck.org/ai-detector/ (read 9 November 2026)
• Fast-DetectGPT repository: https://github.com/baoguangsheng/fast-detect-gpt (read 10 November 2026)
• ContentDetector.ai, archived 11 May 2026: https://web.archive.org/web/20260511112354/https://contentdetector.ai/ (read 9 November 2026)
• Imagga (held): https://imagga.com/ and https://docs.imagga.com/ (read 10 November 2026)
• Walters 2023: https://www.degruyterbrill.com/document/doi/10.1515/opis-2022-0158/pdf (read 8 November 2026)
• Orenstrakh et al. 2023: https://arxiv.org/abs/2307.07411 (read 9 November 2026)
• Weber-Wulff et al. 2023: https://arxiv.org/pdf/2306.15666 (read 9 November 2026)
• Dugan et al. 2024, RAID: https://arxiv.org/pdf/2405.07940 (read 10 November 2026)

Revision log: 2026-11-10: update created at index version 2026.11.10; 2026-11-10: twelfth-wave pages linked