What is the minimum word count for an AI detector? Every hard minimum and every recommended floor in the index, in words

Ten of the 72 cards, from 30 to 300 words, by the vendor pages read 8 and 9 November 2026 in the data file. Two of the ten phrase the floor as a condition of accuracy rather than a refusal: Phrasly's 30 words are "for accurate results" and JustDone's 350 characters are "for reliable results", and both sit in the ranking because the vendor states them as the minimum for a scan (phrasly.ai, justdone.com, read 8 November 2026). The other eight are refusals in the vendor's words: Turnitin does not process a submission under 300 words of prose (guides.turnitin.com, read 8 November 2026), Turnitin Clarity returns an AI writing result only on a pasted block of 300 or more words (guides.turnitin.com, read 9 November 2026), Smodin's input box requires 1,000 characters, about 182 words (smodin.io, read 9 November 2026), and Surfer runs no scan under 100 words (docs.surferseo.com, read 9 November 2026); Originality.ai, QuillBot, Copyleaks and AI or Not state 100 words, 80 words, 255 characters and 250 characters on the pages linked in the table.

Turnitin's floor has a dated history on the record. The minimum was 150 words at the April 2023 launch and rose to 300 on 24 May 2023, announced in the post of 23 May 2023 that also found documents scored under 20% carried "a higher incidence of false positives" (Turnitin claims record; turnitin.com, read 8 November 2026). The same FAQ says that in documents of only a few hundred words "the prediction will be mostly all or nothing", which is the vendor's own statement that its floor is where the score starts, not where it becomes graded.

Five, and the advice runs from about 45 to 200 words. Sapling says its web detector is "more accurate after about 50 words", recommends at least 300 characters per API request and flags any result on under 500 characters low_confidence (sapling.ai, read 8 November 2026). ZeroGPT's pricing page gives "at least 150 to 200 words" as guidance (zerogpt.com, read 8 November 2026). TruthScan and Undetectable AI, two products of the same company, each recommend at least 200 words on their text API for best accuracy (truthscan.com; help.undetectable.ai, read 9 November 2026). BrandWell's archived page of 9 August 2026 carried a tip to try more than 250 characters (moxby.com, read 9 November 2026). Each of the five returns a score on shorter text, so a reader can paste a tweet and get a number; the vendor's own page says the number is less reliable than the same tool's number on a longer text, and none of the five states by how much.

One indexed study scored texts under 50 words, and the floors in the table were set by vendors, not by that study. Jabarian and Imas 2025 held out 160 human and 160 AI stubs under 50 words per model, 8% of its corpus, and recorded a false positive rate of at most 0.0250 for Pangram and 0.0333 for GPTZero on them, against 0.0000 and 0.0071 on the full 1,992 human passages; Originality.ai was excluded from the stub test because its API returned a length filter error on some stubs, which is the 100-word minimum in the table acting on a study (NBER w34223, Tables 3 and B.1, read 9 November 2026; Jabarian and Imas 2025 record). Neither Pangram nor GPTZero states a minimum on the pages read, and both still scored the stubs.

Every other record worked above the highest vendor floor. Weber-Wulff 2023 wrote its human documents at about 10,000 characters and set a 2,000-character minimum, about 360 words, for its ChatGPT documents (arXiv 2306.15666, read 8 November 2026; Weber-Wulff 2023 record); Perkins 2024 started from 15 AI texts of about 500 words (arXiv 2403.19148, read 8 November 2026; Perkins 2024 record); Epoch AI 2026 scored passages of roughly 500 words (epoch.ai, read 9 November 2026); Kar 2024 used one ChatGPT article of under 500 words (DOI 10.1177/02537176241247934, read 9 November 2026); Van Vlasselaer 2026 used papers of at least 4,000 words (DOI 10.1007/s40979-026-00226-w, read 9 November 2026). Chaka 2023 ran 21 whole chatbot responses and 30 translations and prints no word counts (DOI 10.37074/jalt.2023.6.2.12, read 9 November 2026). The ranking by independent testing counts these records per card and says nothing about length.

A minimum word count is where a vendor lets the check run, not where the vendor has shown the check to be right.

Cards that state no floor

A minimum counts prose: Turnitin's 300 words are "words of prose text" with lists, headers, references, tables and code removed first, and the bullet points and code question, added 9 November 2026, lists the 10 cards that state such an exclusion.

All read 8 November 2026 unless a later date is shown; the input limits source of every card is linked in the tables.

• Turnitin AI writing detection FAQs: https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs
• Turnitin, AI writing detection update from the Chief Product Officer (23 May 2023): https://www.turnitin.com/blog/ai-writing-detection-update-from-turnitins-chief-product-officer
• Turnitin Clarity instructor FAQs: https://guides.turnitin.com/hc/en-us/articles/37198804513165-FAQs-for-Instructors-using-Turnitin-Clarity (read 9 November 2026)
• Phrasly AI detector: https://phrasly.ai/ai-detector
• Copyleaks AI detector: https://copyleaks.com/ai-content-detector
• AI or Not API reference: https://docs.aiornot.com/api-reference/reports-by-modality/image (read 9 November 2026)
• JustDone AI detector: https://justdone.com/ai-detector
• QuillBot AI detector: https://quillbot.com/ai-content-detector
• Originality.ai pricing: https://originality.ai/pricing
• Surfer help centre, Humanizer and AI detector: https://docs.surferseo.com/en/articles/9396434-surfer-s-humanizer-and-ai-detector (read 9 November 2026)
• Smodin AI detector: https://smodin.io/ai-content-detector (read 9 November 2026)
• Sapling AI detector: https://sapling.ai/ai-content-detector
• ZeroGPT pricing: https://www.zerogpt.com/pricing
• TruthScan text detection API documentation: https://truthscan.com/truthscan-ai-text-detection-api-documentation (read 9 November 2026)
• Undetectable AI detector API article: https://help.undetectable.ai/en/article/detector-api-1cf74il/ (read 9 November 2026)
• Moxby AI Content Detector Mod (BrandWell): https://moxby.com/marketplace/ai-content-detector/ (read 9 November 2026)
• OpenAI, New AI classifier for indicating AI-written text (31 January 2023, withdrawal note 20 July 2023): https://openai.com/index/new-ai-classifier-for-indicating-ai-written-text/
• Jabarian and Imas 2025, NBER w34223: https://www.nber.org/papers/w34223 (read 9 November 2026)
• Weber-Wulff et al. 2023: https://arxiv.org/pdf/2306.15666
• Perkins et al. 2024: https://arxiv.org/pdf/2403.19148
• Epoch AI 2026, AI detectors rarely flag human writing: https://epoch.ai/data-insights/ai-detectors-false-negatives (read 9 November 2026)
• Kar et al. 2024: https://doi.org/10.1177/02537176241247934 (read 9 November 2026)
• Van Vlasselaer et al. 2026: https://doi.org/10.1007/s40979-026-00226-w (read 9 November 2026)
• Chaka 2023: https://doi.org/10.37074/jalt.2023.6.2.12 (read 9 November 2026)

Revision log: 2026-11-09: page created from the input limits and minimum input fields of 42 cards at index version 2026.11.7.; 2026-11-09: eighth-wave pages linked