Which AI detectors detect both AI text and AI images? The 15 cards that list both, with the study record behind each modality
Fifteen of the 72 AI detectors in the index list both text and images among the content they check, by the modalities field of every card read from the vendor pages on 8 and 9 November 2026 in the data file, and not one of the fifteen has an indexed study on both modalities: 6 have a text record (GPTZero with 16 records down to QuillBot and Winston AI with 2), 3 have an image record (Hive with 2, AI or Not and Illuminarty with 1 each), and 6 have neither. The best-documented image result among them is Hive's in Ha and Passananti 2024, 98.03% (accuracy) with a 0% false positive rate on 280 human artworks and 350 AI images (arXiv 2402.03214, read 9 November 2026), and the same paper recorded false positive rates of 24.47% for AI or Not's Optic and 67.40% for Illuminarty on the same artworks. No text detector among the six with text records has had its image check scored by anyone in the index.
The modalities field records what a vendor says its product accepts, in the vendor's words, with the page and the date; a qualifier in brackets is the card's note on scope, such as GPTZero's AI Vision having no published specifications or SynthID's text check existing only for watermarked text. The table lists every card whose field names both text and image, then the study records the index holds for each modality, so a reader can see that "detects text and images" and "has been tested on text and images" are different statements about every one of the fifteen. This page is one of the questions answered from the index; the modalities view slices all 72 cards by content type, the ranking of image, voice, video and code detectors by independent studies sorts the non-text records, and the AI detector register carries every card.
| Detector | Modalities listed | Text study records | Image study records | Source | Read date |
|---|---|---|---|---|---|
| GPTZero | text, image (AI Vision, specs not published) | 16 | 0 | gptzero.me | 8 November 2026 |
| Copyleaks | text, code, image, video | 9 | 0 | copyleaks.com | 8 November 2026 |
| ZeroGPT | text, image, video | 7 | 0 | zerogpt.com | 8 November 2026 |
| Pangram | text, image, video | 4 | 0 | pangram.com | 8 November 2026 |
| QuillBot | text, image | 2 | 0 | quillbot.com | 8 November 2026 |
| Winston AI | text, image, handwriting (OCR) | 2 | 0 | gowinston.ai | 8 November 2026 |
| Hive | text, image, video, voice (audio), music | 0 | 2: Ha and Passananti 2024 (98.03% (accuracy), 0% false positives on 280 artworks, 3.17% false negatives on 350 AI images); RAID image 2025 (0.60 to 0.44 (accuracy) on 50 adversarial pairs) | docs.thehive.ai | 8 November 2026 |
| AI or Not | image, video, voice (audio), music, text | 0 | 1: Ha and Passananti 2024 (24.47% false positive rate on 280 human artworks, tested as Optic) | aiornot.com | 9 November 2026 |
| Illuminarty | image, text (proof of concept, vendor's wording) | 0 | 1: Ha and Passananti 2024 (67.40% false positive rate on 280 human artworks) | illuminarty.ai | 9 November 2026 |
| Phrasly | text, image | 0 | 0 | phrasly.ai | 8 November 2026 |
| Proofademic | text, image (API now; browser tool coming soon), voice (API) | 0 | 0 | proofademic.ai | 8 November 2026 |
| Walter Writes | text, image (API), voice (API) | 0 | 0 | walterwrites.ai | 8 November 2026 |
| Reality Defender | video, image, voice (audio), text | 0 | 0 | realitydefender.com | 9 November 2026 |
| SynthID | image, video, audio, text (watermarked at generation; no public text check at the read date) | 0 | 0 | synthid.com | 9 November 2026 |
| TruthScan | text, image, PDF documents, video (deepfake), voice (audio) | 0 | 0 | truthscan.com | 9 November 2026 |
Modalities as the vendor page lists them on the date shown; study counts are the card's records by modality at index version 2026.11.13, with the image results as each paper prints them (Ha and Passananti 2024 record; RAID image 2025 record).
Which detectors list both text and images?
Fifteen, in three groups. Six are text detectors that added an image check: GPTZero's AI Vision, whose card notes that no specifications are published (gptzero.me, read 8 November 2026); Copyleaks, which also lists code and video and accepts JPG, PNG, WEBP, BMP, HEIC and HEIF files (copyleaks.com, read 8 November 2026); ZeroGPT and Pangram, which list video as well, Pangram taking JPG, PNG and WebP images of at least 512 by 512 pixels up to 30 MB (zerogpt.com, pangram.com, read 8 November 2026); QuillBot and Phrasly, which list text and image and nothing else (quillbot.com, phrasly.ai, read 8 November 2026); and Winston AI, which adds handwriting through OCR (gowinston.ai, read 8 November 2026). Four are media detectors that added text: Hive, AI or Not, Reality Defender and TruthScan, each of which lists video and voice as well, TruthScan adding PDF documents such as receipts and invoices (docs.thehive.ai, read 8 November 2026; aiornot.com, realitydefender.com, truthscan.com, read 9 November 2026). The remaining five carry a qualifier on one side: Proofademic and Walter Writes offer images through an API rather than the web tool (proofademic.ai, walterwrites.ai, read 8 November 2026), SynthID checks text only where a Google model applied its watermark and offers no public text check (synthid.com, read 9 November 2026), and Illuminarty's own wording calls its text detector a proof of concept (illuminarty.ai, read 9 November 2026).
Two cards sit outside the fifteen on a reading of the field: Compilatio lists text and handwriting (OCR on handwritten documents), which scans writing rather than judging whether a picture was generated (compilatio.net, read 9 November 2026), and Sightengine, Sensity, WasItAI and Resemble Detect list images with no text at all.
Which of the fifteen have been tested on text, and which on images?
Six on text, three on images, none on both. The text records belong to the tools that were text detectors first: GPTZero has 16, from Weber-Wulff 2023's 9 false accusations in 18 human documents to Epoch AI 2026's 0 of 495 (Weber-Wulff 2023 record; Epoch AI 2026 record); Copyleaks 9, ZeroGPT 7, Pangram 4, QuillBot 2 and Winston AI 2, all counted on the ranking by independent testing. Not one of those records scored the same tool's image check; every image figure in the index belongs to a media-first product.
The image records come from two papers. Ha and Passananti 2024 ran 280 human artworks and 350 AI-generated images through Hive, Optic (now AI or Not) and Illuminarty and recorded 98.03% (accuracy) for Hive with 0% false positives and 3.17% false negatives, a 24.47% false positive rate for Optic and 67.40% for Illuminarty (arXiv 2402.03214, read 9 November 2026; Ha and Passananti 2024 record). RAID image 2025 ran 50 real and 50 adversarial AI images through Hive's API and recorded a fall from 0.60 to 0.44 (accuracy) on the AI images, a result its authors call preliminary (arXiv 2506.03988, read 9 November 2026; RAID image 2025 record). Hive's text check, AI or Not's text check and Illuminarty's proof-of-concept text detector appear in no indexed record.
Does one score cover both a text and an image?
No card states that it does, and the fields say the opposite in two ways. Every vendor in the table describes separate checks with separate inputs: Copyleaks meters text in credits of 250 words and images at 1 credit each (copyleaks.com, read 8 November 2026), Pangram publishes image and video file limits beside a text API at 5 queries a second (pangram.com, read 8 November 2026), and Hive scores audio per 10-second chunk and text in 2,048-character chunks through different endpoints (docs.thehive.ai, read 8 November 2026). And the two vendors with a published image false positive rate state it apart from text: TruthScan's pricing page gives an image false positive rate under 1% with no base and a separate 99%+ figure for text (truthscan.com, read 9 November 2026), and Resemble Detect, which lists no text at all, gives 4% on 1,000 image files (resemble.ai, read 9 November 2026). The modalities view marks each content type per card for the same reason.
A vendor that lists text and images has two detectors under one name, and the study record of one says nothing about the other.
Cards that list text only, images only, or neither
| Detector | Modalities listed | Study records | Source | Read date |
|---|---|---|---|---|
| Grammarly | text | 0 | grammarly.com | 8 November 2026 |
| Scribbr | text | 1 | scribbr.com | 8 November 2026 |
| Sapling | text | 3 | sapling.ai | 8 November 2026 |
| JustDone | text | 0 | justdone.com | 8 November 2026 |
| Originality.ai | text, code | 7 | originality.ai | 8 November 2026 |
| Turnitin | text | 5 | guides.turnitin.com | 8 November 2026 |
| GLTR | text | 4 | gltr.io | 9 November 2026 |
| AIDetectPlus | text | 0 | aidetectplus.com | 9 November 2026 |
| GPT-2 Output Detector | text | 12 | huggingface.co | 9 November 2026 |
| Codequiry | code | 0 | codequiry.com | 9 November 2026 |
| WasItAI | image | 0 | wasitai.com | 9 November 2026 |
| Writer | text | 5 | web.archive.org | 9 November 2026 |
| Undetectable AI | text | 1 | undetectable.ai | 9 November 2026 |
| Grammarly Authorship | text | 0 | grammarly.com | 9 November 2026 |
| Turnitin Clarity | text | 0 | turnitin.com | 9 November 2026 |
| Smodin | text | 0 | smodin.io | 9 November 2026 |
| BrandWell | text | 3 | web.archive.org | 9 November 2026 |
| Surfer | text | 0 | surfer.ai | 9 November 2026 |
| Compilatio | text, handwriting (OCR on handwritten documents) | 1 | compilatio.net | 9 November 2026 |
| Quetext | text | 0 | quetext.com | 9 November 2026 |
| DetectGPT | text | 4 | arxiv.org | 9 November 2026 |
| Sightengine | image, video, deepfake (separate model), voice (audio speech), music | 1 (image: RAID image 2025) | sightengine.com | 9 November 2026 |
| Deepware | video (face manipulation only) | 0 | deepware.ai | 9 November 2026 |
| Sensity | video, image, voice (audio), file forensics | 0 | sensity.ai | 9 November 2026 |
| Codeleaks | code | 0 | copyleaks.com | 9 November 2026 |
| Pindrop Pulse | voice (audio), video (Pulse for Meetings) | 0 | pindrop.com | 9 November 2026 |
| Resemble Detect | voice (audio), image, video, music | 0 | resemble.ai | 9 November 2026 |
Modalities as each vendor page lists them, read on the date shown; study records are the card's count at index version 2026.11.13.
The sibling pairing is voice with video: 9 cards list both, 6 list one, and none has a record on either, on the voice and video question, added 9 November 2026.
Questions about detectors for text and images
Is there one AI detector for text and images?
Fifteen of the 72 cards list both, per the vendor pages read 8 and 9 November 2026, and each runs them as separate checks; GPTZero, Copyleaks, ZeroGPT, Pangram, QuillBot, Winston AI, Phrasly, Hive, AI or Not, TruthScan and Reality Defender list both without a qualifier, and Proofademic, Walter Writes, SynthID and Illuminarty list one side through an API, a watermark or a proof of concept.
Which text-and-image detector has the best image test result?
Hive, with 98.03% (accuracy), a 0% false positive rate on 280 human artworks and a 3.17% false negative rate on 350 AI images in Ha and Passananti 2024 (arXiv 2402.03214, read 9 November 2026); the same tool fell from 0.60 to 0.44 (accuracy) on 50 adversarial pairs in RAID image 2025 (arXiv 2506.03988, read 9 November 2026).
Has GPTZero's or Copyleaks' image detection been tested by a study?
No: GPTZero's 16 records and Copyleaks' 9 records in the index all score text, and no indexed paper has run either tool's image check, per the cards at index version 2026.11.13. GPTZero's card notes that AI Vision's specifications are not published (gptzero.me, read 8 November 2026).
Does any detector score text and images in one result?
No card states it: Copyleaks meters text at 250 words a credit and images at 1 credit each (copyleaks.com, read 8 November 2026), and Hive scores text in 2,048-character chunks and media through separate endpoints (docs.thehive.ai, read 8 November 2026).
Which detectors check images but not text?
Four cards list images with no text: WasItAI, Sightengine, Sensity and Resemble Detect, per their pages read 9 November 2026; Sightengine has 1 image record, RAID image 2025, in which it fell from 0.66 to 0.00 (accuracy) on 50 adversarial AI images (arXiv 2506.03988, read 9 November 2026).
Sources
All read 9 November 2026 unless an earlier date is shown; the modalities source of every card is linked in the tables.
- Ha and Passananti et al. 2024, arXiv 2402.03214: https://arxiv.org/abs/2402.03214
- Eddoubi et al. 2025, RAID image benchmark, arXiv 2506.03988: https://arxiv.org/abs/2506.03988
- GPTZero: https://gptzero.me/ (read 8 November 2026)
- Copyleaks AI detector: https://copyleaks.com/ai-content-detector (read 8 November 2026)
- ZeroGPT: https://www.zerogpt.com/ (read 8 November 2026)
- Pangram: https://www.pangram.com/ (read 8 November 2026)
- QuillBot AI detector: https://quillbot.com/ai-content-detector (read 8 November 2026)
- Winston AI pricing: https://gowinston.ai/pricing/ (read 8 November 2026)
- Hive AI-generated content detection documentation: https://docs.thehive.ai/docs/ai-generated-content-detection (read 8 November 2026)
- Hive supported file formats: https://docs.thehive.ai/docs/supported-file-formats (read 8 November 2026)
- AI or Not pricing: https://www.aiornot.com/pricing
- Illuminarty: https://illuminarty.ai/en/
- Phrasly AI detector: https://phrasly.ai/ai-detector (read 8 November 2026)
- Proofademic: https://proofademic.ai/ (read 8 November 2026)
- Walter Writes API: https://walterwrites.ai/api/ (read 8 November 2026)
- Reality Defender FAQ: https://www.realitydefender.com/faq
- SynthID: https://synthid.com/
- TruthScan: https://truthscan.com/
- TruthScan pricing: https://truthscan.com/pricing
- Resemble AI benchmarks: https://www.resemble.ai/benchmarks
- Compilatio Magister: https://www.compilatio.net/en/magister
Revision log: 2026-11-09: page created from the modalities and modality records fields of 42 cards at index version 2026.11.7.; 2026-11-09: eighth-wave pages linked
Related pages: Which AI detectors have a free API tier? The ten vendors whose free allowance covers API calls, and the twenty that sell an API with no free tier, Which AI detectors detect AI-paraphrased text? The 16 vendors that claim it, the 5 that report it as a label or setting of its own, and what 11 studies measured, 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.