How many AI detectors exist? The index count and every study's count
72 AI detectors have a card in the index at version 2026.11.13, by the count in the index data file compiled on 9 November 2026 (the AI detector index), and no published census counts more: the largest study in the index, Walters 2023, tested 16 detectors (DOI 10.1515/opis-2022-0158, read 8 November 2026), and Weber-Wulff 2023 tested 14 (DOI 10.1007/s40979-023-00146-z, read 8 November 2026). Nobody has published a complete count of AI detectors, because the category has no register and tools open and close without notice: OpenAI's own classifier was withdrawn in July 2023 (openai.com, read 8 November 2026), and the hosted demo of the GPT-2 Output Detector is paused.
The honest answer is three counts, each with its base. The index counts the detectors it has read on the same fourteen fields: 72, of which 20 detect text, 12 image, 5 video, 4 voice, 3 code and 1 music, as the vendors state (content type matrix). The 35 indexed studies each counted the tools their authors could reach on their dates, from 1 to 16. And the editors' register lists 52 further detectors to be read and carded through 2027, which would bring the index to 74. The table lists every count in the index with what it counted and when. This page is one of the questions answered from the index; its siblings ask which AI detectors share an owner and whether a legit AI detector exists.
| Source | Detectors counted | What was counted | Date of the count | Record |
|---|---|---|---|---|
| AI Detector Index, version 2026.11.13 | 72 | Detectors read on 14 fields, each with a card | 9 November 2026 | the index |
| Walters 2023 | 16 | Tools run on 126 student and generated essays | June to July 2023 | Walters 2023 record |
| Weber-Wulff et al. 2023 | 14 | Tools run on 54 documents, 12 public tools plus Turnitin and PlagiarismCheck | March to April 2023 | Weber-Wulff 2023 record |
| RAID (Dugan et al. 2024) | 12 | Detectors scored on the shared benchmark | 2024 | RAID 2024 record |
| NIST AI 700-1 (2025) | 11 teams, 2 baselines | Anonymised discriminator teams in the GenAI text-to-text pilot, 348 submissions | April 2024 to 2025 | NIST AI 700-1 2025 record |
| Orenstrakh et al. 2023 | 8 | Tools run on 164 computing coursework submissions | April 2023 | Orenstrakh 2023 record |
| Liang et al. 2023 | 7 | Tools run on 91 TOEFL and 88 US eighth-grade essays | March 2023 | Liang 2023 record |
| Perkins et al. 2024 | 7 | Tools run on 10 human and 104 adversarial AI samples | 2024 | Perkins 2024 record |
| Elkhatat et al. 2023 | 5 | Tools run on human and AI paragraphs | 2023 | Elkhatat 2023 record |
| Chaka 2023 | 5 | Free tools run on 21 chatbot responses and 30 translations | March to April 2023 | Chaka 2023 record |
| Krishna et al. 2023 | 5 | Detectors run on 3,000 paraphrased passages, including GPTZero and the OpenAI classifier | 2023 | Krishna 2023 record |
| Ha and Passananti et al. 2024 | 5 | Image detectors run on 280 artworks and 350 AI images | late 2023 to June 2024 | Ha and Passananti 2024 record |
| Sadasivan et al. 2023 | 4 classes | Watermark, zero-shot, trained and retrieval detectors under recursive paraphrase | 2023, revised 2025 | Sadasivan 2023 record |
| Jabarian and Imas 2025 | 4 | Commercial detectors and a RoBERTa baseline run on 1,992 human and 7,968 AI passages | 2025 | Jabarian and Imas 2025 record |
| Van Vlasselaer et al. 2026 | 4 | Tools run on 160 academic papers | May 2025 | Van Vlasselaer 2026 record |
| Gao et al. 2023 | 1 | The GPT-2 Output Detector on 100 medical abstracts | December 2022 | Gao 2023 record |
| OpenAI classifier statement 2023 | 1 | OpenAI's own classifier, released and withdrawn | January to July 2023 | OpenAI classifier 2023 record |
| Turnitin statements 2023 to 2026 | 1 | Turnitin's own detector, in the vendor's statements | 2023 to 2026 | Turnitin statements record |
How many AI detectors are in the index, and how were they chosen?
72 cards, every one read on the vendor's own pages between 8 and 9 November 2026 and recorded on the same fourteen fields: sixteen on 8 November 2026 and six more (GLTR, AIDetectPlus, the GPT-2 Output Detector, Codequiry, WasItAI and Reality Defender) on 9 November 2026, per the corrections log. A detector enters when its vendor page states what it detects and how it is sold, and a card is never built from a review of the tool. Twelve of the 72 have at least one study record and ten have none, as listed on the ranking by independent testing compiled 9 November 2026; the count of cards says how many tools the index has read, not how many exist.
The index data file is the only count on this site. The hub tables, the home board and this page read the same number from it at build time, so the figure on this page changes on the day a card is added and the revision log records the change (methodology).
How many detectors did each study count?
Walters 2023 ran 16 tools over 126 essays: Content at Scale, ContentDetector.ai, Copyleaks, Crossplag, Grammica, GPT Radar, GPTZero, IvyPanda, OpenAI, Originality.ai, Sapling, Scribbr, SEO.ai, Turnitin, Writer and ZeroGPT, as the record lists them (DOI 10.1515/opis-2022-0158, read 8 November 2026); seven of the sixteen have a card. Weber-Wulff et al. 2023 ran 14, twelve publicly available tools plus Turnitin and PlagiarismCheck, from Check For AI to Writeful GPT Detector (DOI 10.1007/s40979-023-00146-z, read 8 November 2026); five of the fourteen have a card. RAID scored 12 detectors (DOI 10.18653/v1/2024.acl-long.674, read 8 November 2026), Orenstrakh et al. 2023 ran 8 (arXiv 2307.07411, read 9 November 2026), and Liang 2023 and Perkins 2024 ran 7 each.
The newer studies counted fewer tools and more text. Jabarian and Imas 2025 ran four detectors over 9,960 passages (NBER w34223, read 9 November 2026); Van Vlasselaer 2026 ran four over 160 papers of 4,000 words or more (DOI 10.1007/s40979-026-00226-w, read 9 November 2026); Gao 2023 ran one detector over 100 abstracts (DOI 10.1038/s41746-023-00819-6, read 9 November 2026). The NIST pilot is the one count that is not a list of products: 172 entities registered, 11 discriminator teams submitted and the teams are anonymised, so no product name can be matched (NIST AI 700-1, read 9 November 2026). Across all 35 records, 12 index cards were tested at least once and 10 were tested by nobody.
How many AI detectors will the index hold, and which are planned?
Three tools on the record have been withdrawn or paused by their makers, listed with the vendor statements on which AI detectors were shut down. The editors' register lists 52 further detectors to be read and carded through 2027, for a total of 74, and each addition is logged in the corrections log: the next six are Google's SynthID detector, Turnitin Clarity, Undetectable AI's detector, Grammarly Authorship, Writer.com's detector and AI or Not, then Compilatio, Surfer and Quetext among ten more. A planned card is not a count of a detector that exists in the index; it is a tool the editors have seen named on a vendor page or in a study and have not yet read on the fourteen fields. Tools named in the studies without a card include Check For AI, Compilatio, Content at Scale, Crossplag, DetectGPT, GPTKit, PlagiarismCheck, Writer, Grammica, IvyPanda, Optic AI or Not and Illuminarty, each listed on its study record with "no card".
The 2027 census, written 10 November 2026, extends this count: how many of the 72 cards are withdrawn, unreachable or redirected (12), how many have a study record (35), how many are open source (8), and what the study index holds by year and kind.
Questions about counting AI detectors
How many AI detectors are there in total?
No complete count exists. The index holds 72 cards at version 2026.11.13, compiled 9 November 2026; the largest published test, Walters 2023, ran 16 tools (DOI 10.1515/opis-2022-0158, read 8 November 2026); the editors' register names 74 in all, 52 of them not yet carded.
How many AI detectors have been independently tested?
12 of the 72 cards have at least one study record, from GPTZero with 11 to Hive and Scribbr with 1, per the ranking by independent testing compiled 9 November 2026. The other 10 cards have been tested by no indexed paper.
How many AI image, voice or video detectors exist?
Of the 72 cards, 12 state image detection, 5 video, 4 voice and 1 music, as the vendors state on pages read 8 and 9 November 2026 (content type matrix); one of the twelve image detectors, Hive, has a published study (ranking by content type).
Which study tested the most AI detectors?
Walters 2023, with 16 tools on 126 essays (DOI 10.1515/opis-2022-0158, read 8 November 2026), followed by Weber-Wulff 2023 with 14 (DOI 10.1007/s40979-023-00146-z, read 8 November 2026) and RAID with 12. The NIST pilot received 348 discriminator submissions from 11 anonymised teams (NIST AI 700-1, read 9 November 2026).
Why does the number keep changing?
Because tools open, close and rename: OpenAI withdrew its classifier in July 2023 (openai.com, read 8 November 2026), the GPT-2 Output Detector's hosted demo is paused (Hugging Face, read 9 November 2026), Hive became Deploy on 26 October 2026 and GPTZero announced plans to join Superhuman on 23 June 2026, both read on the vendor pages 8 November 2026.
Sources
All read 8 November 2026 unless a later date is shown.
- Walters 2023: https://doi.org/10.1515/opis-2022-0158
- Weber-Wulff et al. 2023: https://doi.org/10.1007/s40979-023-00146-z
- Dugan et al. 2024 (RAID): https://doi.org/10.18653/v1/2024.acl-long.674
- Liang et al. 2023: https://doi.org/10.1016/j.patter.2023.100779
- Perkins et al. 2024: https://doi.org/10.1186/s41239-024-00487-w
- Elkhatat et al. 2023: https://doi.org/10.1007/s40979-023-00140-5
- Chaka 2023: https://doi.org/10.37074/jalt.2023.6.2.12 (read 9 November 2026)
- Orenstrakh et al. 2023: https://arxiv.org/abs/2307.07411 (read 9 November 2026)
- Krishna et al. 2023: https://arxiv.org/abs/2303.13408 (read 9 November 2026)
- Sadasivan et al. 2023: https://arxiv.org/abs/2303.11156 (read 9 November 2026)
- Gao et al. 2023: https://doi.org/10.1038/s41746-023-00819-6 (read 9 November 2026)
- Jabarian and Imas 2025: https://www.nber.org/papers/w34223 (read 9 November 2026)
- Van Vlasselaer et al. 2026: https://doi.org/10.1007/s40979-026-00226-w (read 9 November 2026)
- NIST AI 700-1: https://doi.org/10.6028/NIST.AI.700-1 (read 9 November 2026)
- Ha, Passananti et al. 2024: https://arxiv.org/abs/2402.03214 (read 9 November 2026)
- OpenAI, new AI classifier (withdrawn): https://openai.com/index/new-ai-classifier-for-indicating-ai-written-text/
- GPT-2 Output Detector model card: https://huggingface.co/openai-community/roberta-base-openai-detector (read 9 November 2026)
Revision log: 2026-11-09: page created at index version 2026.11.3, 22 cards and 17 records; 2026-11-10: twelfth-wave pages linked
Related pages: Which AI detectors share an owner? The ownership map from the vendor pages, Which AI detectors were shut down? Every withdrawal on the record, with the vendor's own statement, 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. and Questions about AI detectors, answered from the index.