AI detectors ranked by maximum input per check
Detecting-AI.com states the largest input per check that can be sorted in the index, 1,000,000 characters per scan on every paid plan, about 182,000 words at 5.5 characters a word, per its home page and pricing page on detecting-ai.com, read 10 November 2026. AI or Not states the largest API figure, 500,000 characters of text per API request, about 91,000 words at 5.5 characters a word, per its API documentation on docs.aiornot.com, read 9 November 2026. Winston AI states the largest web-tool figure, 200,000 characters per scan on its Essential, Advanced and Elite plans, about 36,000 words, per the pricing table on gowinston.ai, read 8 November 2026; Sapling states the same 200,000 characters for one API request but 100,000 for a web check, so Winston AI sits above Sapling on the web-tool tie rule. Twenty-eight of the 72 detectors state a numeric ceiling that can be ranked; six more state an unlimited tier and sit in a second table. With the eleventh wave on 10 November 2026 SEO.ai joined at 20,000 characters per check (the widget's enforced cut; the page's meta description says 25,000), IvyPanda at 4,500 characters per check and AI Writing Check at 400 words per check from its archived counter; GPT Radar, Grammica, RADAR, Ghostbuster, Hiya and McAfee state no ceiling and are omitted. On 10 November 2026 the tenth wave added four text ceilings: Passed.ai at 5,000 words per Academic Honesty Report and StealthWriter at 5,000 words per input on every paid plan, both 27,500 characters at 5.5 characters a word and placed after Writer's equal figure on the study-count tie rule, Wordtune at 5,000 characters per check from its withdrawn widget, and Monica at 250 words, the smallest ceiling in the index. On 10 November 2026 the ninth wave of cards was recounted: Detecting-AI.com ranks first at 1,000,000 characters per scan on every paid plan, about 182,000 words, per its home page (detecting-ai.com, read 10 November 2026), and AI or Not's 500,000 characters per API request is now second. The ranking was recounted on 9 November 2026 for the sixteen cards added in the third and fourth waves, seven of which state a ceiling.
Updated Limits checked 8 November 2026 for the first sixteen cards and 9 November 2026 for the twenty-two added that day
Short answer
Detecting-AI.com states 1,000,000 characters per scan on every paid plan, about 182,000 words, read 10 November 2026; AI or Not states 500,000 characters of text per API request, about 91,000 words at 5.5 characters a word, read 9 November 2026; Winston AI and Sapling both state 200,000 characters, about 36,000 words; Turnitin states 30,000 words, and TruthScan and Undetectable AI each state 30,000 words per API request, read 9 November 2026; Proofademic 25,000 words; GPTZero 150,000 characters; JustDone 15,000 words; Smodin 50,000 characters; Writer 5,000 words on the archived page of a tool sunset on 22 December 2025; Walter Writes 2,000 words; BrandWell 10,000 characters inside the Moxby Mod its page now resolves to; Quetext 1,000 words per free scan; AIDetectPlus 3,000 characters in its home page box, about 500 words. ZeroGPT, QuillBot, Copyleaks and Phrasly each state a tier with no limit, so they cannot be sorted and appear in the second table with the vendor's wording. Character figures are converted to words only as "about N words at 5.5 characters a word". Sources read 8 November 2026. No detector score should be the only evidence in a misconduct or hiring decision.
Which AI detectors accept the most text in one check?
| Rank | Detector | Largest stated input per check | About N words | Plan carrying the limit | Best for | Main limitation |
|---|---|---|---|---|---|---|
| 1 |  Detecting-AI.com | 1,000,000 characters per scan (every paid plan) | about 182,000 | Writer, Pro and Team plans; the free tier is 15,000 characters per check | Whole manuscripts or books checked in one paid scan | 15,000 characters per free check; the API takes 160,000 characters per request; no study record |
| 2 |  AI or Not | 500,000 characters per request (text API) | about 91,000 | API, every plan | Developers scoring very long text through the API | Minimum 250 characters; the web tool states no text maximum; the one study record scored images |
| 3 |  Decopy | 200,000 characters per check | about 36,000 | Web tool, basic detection without an account | A long document checked in the browser without an account | Metered in quotas and credits (Free: 10 a month); a daily usage notice states no figure; no study record |
| 4 |  Winston AI | 200,000 characters per scan | about 36,000 | Essential, Advanced, Elite | Long reports in one web scan | Free row of the table is blank |
| 5 |  Sapling | 200,000 characters per API request | about 36,000 | API | Developers scoring long documents | Web check 100,000 characters (Pro); free check 2,000 |
| 6 |  Turnitin | 30,000 words | 30,000 | Institutional licence | Theses and long essays inside an LMS | 300-word minimum; English only for paraphraser detection |
| 7 |  TruthScan | 30,000 words per request (text API) | 30,000 | API, every plan, 1 credit per word | Teams scoring long text through an API that also checks images and PDFs | At least 200 words recommended; the web box states no maximum; no price per text word published |
| 8 |  Undetectable AI | 30,000 words per request (API) | 30,000 | API | Developers who want a 1 to 100 score with eight third-party estimates | Web tool 10,000 words per check; at least 200 words recommended; no study record |
| 9 |  GPTKit | 153,600 characters per request (Institution plan) | about 28,000 | Unlimited (Institution), $20 a month; 30,720 on Team, 10,240 on Personal | Teams pooling credits across up to 20 members | 2048 characters per request as a guest or on the Free Plan; 2 study records point in opposite directions |
| 10 |  PlagiarismCheck | 97 pages per check, 1 page being 275 words (26,675 words) | 26,675 | Web tool, every plan; pages are the vendor's unit | Long papers submitted through an LMS | Stated in pages, converted here by the vendor's own 275-word page; rank 13 of 14 in Weber-Wulff 2023 |
| 11 |  Proofademic | 25,000 words per request | 25,000 | Professional | Long academic documents | 3,500 words on Essential |
| 12 |  GPTZero | 150,000 characters | about 27,000 | Free account | Batches of up to 250 files | 10,000 characters without an account |
| 13 |  JustDone | 15,000 words in one scan | 15,000 | Premium | Mid-length papers | 350-character minimum |
| 14 |  Smodin | 50,000 characters per text or document | about 9,100 | Web tool, free without an account | A long article checked free with no account | 1,000-character minimum; the API bills 0.5 words per input word |
| 15 |  AI Detector Pro | 40,000 characters per paid scan | about 7,300 | Basic, Pro and Unlimited; the FAQ calls it about 6,000 to 7,000 words | A chapter checked in one paid scan | The free version's document size is limited with no figure stated; 3 free scans a month; no study record |
| 16 |  Writer | 5,000 words per check (archived page, 9 December 2025) | 5,000 | Free web tool, before sunset | Reading the limit of a tool four 2023 studies tested | The tool was sunset on 22 December 2025 and the page redirects to writer.com |
| 17 |  Passed.ai | 5,000 words per Academic Honesty Report | 5,000 | Pay-as-you-go reports from $1.49 (archived pricing page, 17 April 2026); the in-app tool states no per-check limit | A single essay audited as a shareable report | The service is closing and registration is closed; no study record; the engine was Originality.ai's |
| 18 |  StealthWriter | 5,000 words per input (every paid plan) | 5,000 | Starter and above; 1,000 words per input on Free; 500 words per scan on the no-account demo | A chapter checked on a paid plan beside a humanizer | The free input stops at 1,000 words; no accuracy figure; no study record |
| 19 |  SEO.ai | 20,000 characters per check (widget) | about 3,600 | The free tool, no login; the page's meta description says 25,000 characters, the widget enforces 20,000 | A long article checked free without an account | The one record printed 12% of 42 human papers called AI; the 98.4% claim has no dataset |
| 20 |  Walter Writes | 2,000 words per request | 2,000 | Elite and Teams | Short documents | 750 words per request on Starter |
| 21 |  BrandWell | 10,000 characters per check (Moxby Mod) | about 1,800 | Moxby Free plan, in the Moxby browser extension | A page of text checked inside a browser extension | The BrandWell page itself redirects to the Mod since 29 August 2026; the archived free scan took 2,500 characters |
| 22 |  Paperpal | 1,200 words per free scan | 1,200 | Free, 5 scans a day; paid plans state a monthly allowance (Prime 25,000 words) and no per-scan ceiling | A manuscript section checked free | Paid plans meter words per month rather than per scan; .docx uploads up to 25 MB; no study record |
| 23 |  Quetext | 1,000 words per free scan | 1,000 | Free | A short essay checked free with per-sentence AI probability | Paid plans meter words per month (100,000 and 200,000) rather than per check |
| 24 |  Wordtune | 5,000 characters per check (archived widget) | about 910 | The free web widget, withdrawn in 2025; the launch blog post's 250,000 characters contradicts it | Readers tracing what the withdrawn widget accepted | The page returned 404 from 26 July 2025; the ceiling sits in the shipped JavaScript; one study record on one text |
| 25 |  IvyPanda | 4,500 characters per check | about 820 | The free GPT Essay Checker, no sign-in prompt | A short essay checked free with a colour-coded result | A beta with no figure; 23% of 126 documents uncertain in Walters 2023 |
| 26 |  AIDetectPlus | 3,000-character box on the home page widget | about 500 | Home page widget, no account | A short passage checked without an account | No per-document maximum published for the signed-in detectors |
| 27 |  AI Writing Check | 400 words per check (archived counter) | 400 | The free nonprofit tool, withdrawn in August 2023; longer responses had to be split into sections | Readers tracing what the withdrawn tool accepted | Withdrawn since August 2023; the one record prints an averaged rate across seven tools |
| 28 |  Monica | 250 words per check (web tool paste box) | 250 | The free web tool inside the Monica assistant, metered at 1 Advanced Credit per 5 words | A short passage checked inside an assistant already installed | The smallest ceiling in the index; the detector calls GPTZero, Copyleaks and ZeroGPT; no study record |
Stated as unlimited
| Detector | Vendor wording on the unlimited tier | Largest numeric figure also stated | Source, read 8 November 2026 |
|---|---|---|---|
| ZeroGPT | ENTERPRISE unlimited | 350,000 characters per detection (EXPERT); 500,000 characters (API VIP) | zerogpt.com/pricing |
| QuillBot | Premium: no word limit for detection scans | 1,200 words per scan (Free) | quillbot.com |
| Copyleaks | Web-Based Platform: maximum 2,000 pages (no character max) | 25,000 characters (browser extension and free scan) | copyleaks.com |
| Phrasly | No character limits on detection (web tool) | 15,000 characters per request (Business API) | phrasly.ai |
| Isgen | Ultimate: unlimited AI Detector submission length | 60,000 characters per submission (Academic); 5,000 characters per scan free | isgen.ai/pricing (read 9 November 2026) |
| Content Detector AI | Archived page (11 May 2026): "unlimited word limit" and "Check Counts Unlimited" | 10,000 characters per check in the Moxby Mod the address now redirects to | web.archive.org (read 9 November 2026) |
One of five rankings in the rankings hub, beside free tiers and independently tested detectors. The home board answers which AI detector takes the longest text in one check?
Which AI detectors should you check first?
- Detecting-AI.com, rank 1Best for: Whole manuscripts or books checked in one paid scan. Main limitation: 15,000 characters per free check; the API takes 160,000 characters per request; no study record.
- AI or Not, rank 2Best for: Developers scoring very long text through the API. Main limitation: Minimum 250 characters; the web tool states no text maximum; the one study record scored images.
- Decopy, rank 3Best for: A long document checked in the browser without an account. Main limitation: Metered in quotas and credits (Free: 10 a month); a daily usage notice states no figure; no study record.
- Winston AI, rank 4Best for: Long reports in one web scan. Main limitation: Free row of the table is blank.
- Sapling, rank 5Best for: Developers scoring long documents. Main limitation: Web check 100,000 characters (Pro); free check 2,000.
What does each ranked limit mean in practice?
AI or Not
- Verdict
- Fits developers who send very long text through an API.
- Value
- 500,000 characters of text per API request, from a 250-character minimum (about 64 words), per the text endpoint documentation (docs.aiornot.com, read 9 November 2026).
- Best for
- Teams scoring whole documents or transcripts in one call.
- Main limitation
- The web tool states no text maximum, only that text can be pasted or uploaded as DOC, DOCX or TXT; the one study record, Ha and Passananti 2024, scored images, not text (arXiv 2402.03214, read 9 November 2026).
Winston AI
- Verdict
- Fits the longest single web scan in the index.
- Value
- 200,000 characters per scan on Essential, Advanced and Elite, about 36,000 words at 5.5 characters a word (gowinston.ai/pricing, read 8 November 2026).
- Best for
- Editors checking a full report or dissertation chapter at once.
- Main limitation
- The Free row of the characters-per-scan line is blank, and Essential lists at $18 per month billed monthly with a promo price of $9 shown (gowinston.ai/pricing, checked 8 November 2026).
Sapling
- Verdict
- Fits developers who score long documents through the API.
- Value
- 200,000 characters per API request, with at least 300 characters recommended; Pro and Enterprise web checks take 100,000 characters (sapling.ai/docs, sapling.ai, read 8 November 2026).
- Best for
- Teams wiring detection into their own systems.
- Main limitation
- The free web check covers the first 2,000 characters only; Pro is $25 per month billed monthly or $12 per month billed annually (sapling.ai/pricing, checked 8 November 2026).
Turnitin
- Verdict
- Fits institutions submitting long student work through an LMS.
- Value
- Up to 30,000 words per English submission, with a 300-word minimum of prose and a file under 100 MB (guides.turnitin.com, read 8 November 2026).
- Best for
- Universities with a Turnitin Originality licence.
- Main limitation
- No individual access and no published prices; lists, tables, references and code are excluded from scoring (guides.turnitin.com, read 8 November 2026).
TruthScan
- Verdict
- Fits teams that score long text through an API alongside image and PDF checks.
- Value
- Request input under 30,000 words per text API call, billed at 1 credit per word, with at least 200 words recommended for best accuracy (truthscan.com, read 9 November 2026).
- Best for
- Operations teams already paying for TruthScan's image and PDF results.
- Main limitation
- The web detector page states no maximum for its Basic AI Detection Analysis box, and the pricing page counts results in images and PDF pages with no price per text word (truthscan.com, read 9 November 2026); no study record.
Undetectable AI
- Verdict
- Fits developers who want a large API ceiling and a free 10,000-word web check.
- Value
- Less than 30,000 words per API request, with at least 200 words recommended; the web tool takes up to 10,000 words per check with no signup (help.undetectable.ai, undetectable.ai, read 9 November 2026).
- Best for
- Writers checking a long draft free, and teams wiring the API.
- Main limitation
- The same company sells a humanizer, and no study in the index tested the detector.
Proofademic
- Verdict
- Fits academic users with long documents on the top plan.
- Value
- 25,000 words per request on Professional, 8,000 on Premium and 3,500 on Essential (proofademic.ai, read 8 November 2026).
- Best for
- Graduate students and supervisors checking whole chapters.
- Main limitation
- Professional costs $45 per month billed monthly or $25 per month billed at $300 annually; the free tier is a one-time 1,000-word scan (proofademic.ai/pricing, checked 8 November 2026).
GPTZero
- Verdict
- Fits readers who want a large limit without paying.
- Value
- Texts up to 150,000 characters, about 27,000 words at 5.5 characters a word, or a batch of up to 250 files, with a free account (gptzero.me/faq, read 8 November 2026).
- Best for
- Teachers checking a full set of essays in one batch.
- Main limitation
- The widget takes 10,000 characters per scan without an account; premium plans start at $12.99 per month per the vendor blog (gptzero.me/faq, gptzero.me, checked 8 November 2026).
JustDone
- Verdict
- Fits mid-length papers on a paid plan.
- Value
- Up to 15,000 words in one scan on Premium, with a 350-character minimum for reliable results (justdone.com, read 8 November 2026).
- Best for
- Students checking a term paper in one pass.
- Main limitation
- Plan prices are absent from the detector page read on 8 November 2026, so the cost of the 15,000-word scan is unstated on the card (justdone.com).
Smodin
- Verdict
- Fits readers who want a long free check without an account.
- Value
- A ceiling of 50,000 characters per text or document, about 9,100 words at 5.5 characters a word, with a 1,000-character minimum in the input box (smodin.io, read 9 November 2026).
- Best for
- Bloggers and students checking a full article at once.
- Main limitation
- Sentence-level analysis needs a free account; the API bills 0.5 words per input word with a 120-requests-a-minute limit (docs.smodin.io, read 9 November 2026); no study record.
Writer
- Verdict
- Fits readers who want the ceiling of a tool that four 2023 studies tested.
- Value
- Up to 5,000 words per check, pasted or by URL, on the archived detector page of 9 December 2025 (web.archive.org, read 9 November 2026).
- Best for
- Readers reading the Weber-Wulff, Walters, Elkhatat and Chaka results in context.
- Main limitation
- The tool was sunset on 22 December 2025 and the page redirects to writer.com, so the limit describes a product no longer offered (Writer card).
Walter Writes
- Verdict
- Fits short documents checked often.
- Value
- 2,000 words per request on Elite and Teams, 1,500 on Pro and 750 on Starter (walterwrites.ai/pricing, read 8 November 2026).
- Best for
- Writers checking articles and cover letters.
- Main limitation
- The smallest ranked ceiling; Elite costs $47 per month billed monthly or $26 per month billed at $312 annually (walterwrites.ai/pricing, checked 8 November 2026).
BrandWell
- Verdict
- Fits a page of text checked inside a browser extension.
- Value
- Up to 10,000 characters per check in the AI Content Detector Mod for the Moxby browser extension, about 1,800 words at 5.5 characters a word; the archived BrandWell page offered a free scan of 2,500 characters (brandwell.ai, web.archive.org, read 9 November 2026).
- Best for
- Readers who already use the Moxby extension.
- Main limitation
- The BrandWell detector page has redirected to the Moxby listing since 29 August 2026, and the listing says the Mod "does not determine whether a person or AI wrote the text"; two 2023 records tested the tool as Content at Scale.
Quetext
- Verdict
- Fits a short essay checked free with per-sentence results.
- Value
- Up to 1,000 words per free scan on the detector page's 0 / 1,000 counter (quetext.com, read 9 November 2026).
- Best for
- Students checking one essay at a time without paying.
- Main limitation
- Paid plans meter the detector by words per month, 100,000 on Essential and 200,000 on Professional, rather than per check, and the API bills $0.10 per 1,000 words (quetext.com, read 9 November 2026); no study record.
AIDetectPlus
- Verdict
- Fits a short passage checked without an account.
- Value
- A 3,000-character box on the home page widget, about 550 words at 5.5 characters a word, free with no account (aidetectplus.com, read 9 November 2026).
- Best for
- Readers checking one paragraph or a short email before signing up.
- Main limitation
- The signed-in detectors state no per-document maximum, so the widget box is the only numeric ceiling on the card; credits are metered at 1 credit = 1 word, with 1,000 free on sign-up and 5,000 credits at $9.00 as shown (aidetectplus.com, checked 9 November 2026).
Detecting-AI.com
- Verdict
- Fits readers who want to check a whole manuscript in one paid scan.
- Value
- 1,000,000 characters per scan on every paid plan, about 182,000 words at 5.5 characters a word, against 15,000 characters per check on the free tier; the API and MCP server take 160,000 characters per request (detecting-ai.com, detecting-ai.com/pricing, read 10 November 2026).
- Best for
- A book-length document checked once rather than in pieces.
- Main limitation
- The free check is 15,000 characters; no study in the index tested Detecting-AI.com, and the vendor's 99% (accuracy) and under-1% false positive claims carry no base.
Decopy
- Verdict
- Fits readers who want a long document checked in the browser without an account.
- Value
- 200,000 characters per check, about 36,000 words: the counter reads 0 / 200000 characters and the page states support for up to 200,000 characters per check (decopy.ai/ai-detector, read 9 November 2026); the web tool sits above Sapling's equal API figure on the tie rule.
- Best for
- A report checked once, free, with no sign-up.
- Main limitation
- Usage is metered in quotas and credits (the Free plan lists 10 of each a month) and a notice reading "You have reached the usage limit for today" states no daily figure; no study in the index tested Decopy.
GPTKit
- Verdict
- Fits teams on the Institution plan who send long texts to the API.
- Value
- 153,600 characters per request on Unlimited (Institution), about 28,000 words, with 30,720 on Team and 10,240 on Personal; a guest or Free Plan request is 2048 characters (gptkit.ai, gptkit.ai/plans, read 9 November 2026).
- Best for
- Institutions pooling credits across up to 20 members.
- Main limitation
- The free limit is 2048 characters, about 370 words; the two records print 0 of 114 false positives (Orenstrakh 2023) and 4.5% accuracy on manipulated AI text (Perkins 2024).
PlagiarismCheck
- Verdict
- Fits institutions submitting long papers through an LMS.
- Value
- 97 pages per check, a page being 275 words by the vendor's Terms, so 26,675 words; the API accepts plain text or a doc, docx, pdf, txt, odt or rtf file (plagiarismcheck.org/ai-detector, read 9 November 2026).
- Best for
- Theses and long essays inside Canvas, Moodle, Blackboard, Brightspace, Schoology or Google Classroom.
- Main limitation
- The limit is stated in the vendor's page unit, converted here by its own definition; the one record, Weber-Wulff 2023, ranked the detector 13th of 14.
AI Detector Pro
- Verdict
- Fits subscribers checking a chapter at a time.
- Value
- 40,000 characters per paid scan, which the FAQ calls "about 6000-7000 words" (about 7,300 at 5.5 characters a word); the API accepts units of up to 4,000 characters (aidetector.pro/pages/faq, read 10 November 2026).
- Best for
- Paid users scanning long documents inside the web app.
- Main limitation
- The free version states a limited document size with no figure and 3 scans a month; no study in the index tested AI Detector Pro.
Paperpal
- Verdict
- Fits researchers checking one section of a manuscript free.
- Value
- 1,200 words per free scan, 5 scans a day, per the placeholder and the comparison table on the detector page; paid plans state a monthly allowance (Prime 25,000 words, Pro unlimited) and no per-scan ceiling (paperpal.com/tools/ai-detector, read 10 November 2026).
- Best for
- A free abstract-and-introduction check before a paid plan.
- Main limitation
- Paid plans meter words per month rather than per scan, so no larger per-check figure exists to rank; no study in the index tested Paperpal.
Passed.ai
- Verdict
- Fits existing customers who audit one essay at a time as a report.
- Value
- 5,000 words per Academic Honesty Report ("Up to 5,000 words/report"), the one per-check figure the archived pricing page prints; the in-app tool scans pasted text with no stated ceiling and the subscription meters 1 credit per 100 words (web.archive.org, pricing, 17 April 2026, read 10 November 2026).
- Best for
- A shareable report on a single student essay from an account with a term still to run.
- Main limitation
- Every page served a closing notice on 10 November 2026 and registration is closed; no study in the index tested Passed.ai, whose detector was Originality.ai's engine.
StealthWriter
- Verdict
- Fits paid subscribers checking a chapter at a time beside the vendor's humanizer.
- Value
- 5,000 words per input on every paid plan (Starter to Enterprise), 1,000 words per input on the Free plan and 500 words per scan on the no-account demo (stealthwriter.ai/pricing, stealthwriter.ai/ai-detector, read 10 November 2026).
- Best for
- A long passage checked on a paid plan with sentence-level feedback.
- Main limitation
- The free ceilings are 1,000 and 500 words; the vendor publishes no accuracy figure and no study in the index tested the detector.
Wordtune
- Verdict
- Fits readers who want to know what the withdrawn widget accepted.
- Value
- 5,000 characters per check, about 910 words: the paste box placeholder read "Paste your text here, up to 5,000 characters" and the validation message "The maximum text length is 5000 characters" (archived widget bundle, 26 November 2024, read 10 November 2026); the launch blog post's "up to 250,000 characters" contradicts the widget and is recorded as a contradiction.
- Best for
- Readers tracing a 2023 or 2024 Wordtune percentage on a short text.
- Main limitation
- The page returned 404 on 10 November 2026 and in captures since 26 July 2025, so the ceiling is a record of a withdrawn tool (web.archive.org, 30 March 2025, read 10 November 2026).
Monica
- Verdict
- Fits readers who want a short passage checked inside an assistant they already run.
- Value
- 250 words per check: the paste box counts "0 / 250 words" and the detector meters "Every 5 words costs 1 Advanced Credits" (monica.im/tools/ai-content-detector, read 10 November 2026), the smallest stated ceiling in the index.
- Best for
- A paragraph checked from the browser extension or phone app.
- Main limitation
- 250 words is under a page; the page states the result comes from GPTZero, Copyleaks and ZeroGPT calls, two of them locked on the free plan; no study in the index tested Monica.
SEO.ai
- Verdict
- Fits readers who want a long free check without an account.
- Value
- 20,000 characters per check, about 3,600 words: the widget's counter reads "Max 20000 allowed" and cuts the input there (tools.seo.ai widget bundle, read 10 November 2026); the page's meta description says "Check up to 25.000 characters", a contradiction recorded on the card, and the enforced figure ranks (the Wordtune reading).
- Best for
- A long article pasted free with no login.
- Main limitation
- Walters 2023 printed 12% of 42 human papers called AI and 72% of 126 correct (DOI 10.1515/opis-2022-0158, Table 4, read 8 November 2026); the 98.4% claim has no dataset.
IvyPanda
- Verdict
- Fits students who want a short essay checked free with a colour map.
- Value
- 4,500 characters per check: the paste box counter reads "4500 characters left" (ivypanda.com/tools/gpt-essay-checker, read 10 November 2026).
- Best for
- A short essay checked free with no sign-in prompt.
- Main limitation
- The vendor calls the tool a beta and prints no figure; Walters 2023 returned 23% of 126 documents as uncertain (DOI 10.1515/opis-2022-0158, Table 4, read 8 November 2026).
AI Writing Check
- Verdict
- Fits readers who want to know what the withdrawn nonprofit tool accepted.
- Value
- 400 words per check: the counters read "100 Minimum words required" and "400 Max words", and the page said "To check responses longer than 400 words, you must separate the text into individual sections" (web.archive.org, 17 July 2023, read 10 November 2026).
- Best for
- Readers tracing a 2023 verdict on a short student response.
- Main limitation
- The notice has read "AI Writing Check is no longer available" since August 2023 (aiwritingcheck.org, read 10 November 2026); the one record prints an averaged rate.
How we ranked
- Field sorted
- Largest words or characters per check the vendor states for any plan, characters shown beside a conversion at 5.5 characters a word, the one rate every page of the index uses since 9 November 2026
- Direction
- Higher first
- Tie rule
- Web tool before API, then a figure stated in words before one converted from characters
- Qualification
- A stated numeric maximum per check; an unlimited tier is listed apart
- Qualified
- 28 of 72 detectors
Field sorted: the input limits field on each card, read on 8 November 2026 for the first sixteen cards and 9 November 2026 for the twenty-two added that day, from the vendor page recorded on the card. From each field the largest words-or-characters-per-check figure the vendor states for any plan was taken; character figures were converted to words only as "about N words at 5.5 characters a word", and the stated figure is shown beside the conversion. Direction: more ranks higher. Tie rule: first the figure for the web tool rather than the API, then a figure stated in words over one converted from characters, then product name. Qualification rule: the vendor must state a numeric maximum per check; a card whose largest stated tier is "unlimited" or "no limit" cannot be placed on a number line and is listed in the second table with the vendor's words. 8 of 22 qualified for the ranked table on the morning of 9 November 2026; 15 of 72 qualify after the third and fourth waves of cards were recounted that evening (AI or Not, TruthScan, Undetectable AI, Smodin, Writer, BrandWell and Quetext added); 21 of 51 qualified after the ninth wave (Detecting-AI.com, Decopy, GPTKit, PlagiarismCheck, AI Detector Pro and Paperpal added on 10 November 2026) and 25 of 72 after the tenth (Passed.ai, StealthWriter, Wordtune and Monica added later the same day); 6 sit in the unlimited table (Isgen and Content Detector AI added); 32 are omitted, eight of the tenth wave among them (Binoculars truncates at 512 tokens in code, Trinka meters credits per 5,000 words with no per-check ceiling, Check For AI and Writefull published none, and ElevenLabs, FakeCatcher, Deepfake-O-Meter and Ircam Amplify take media rather than text), and Fast-DetectGPT because its demo states limits with no figures; of the earlier 19 because the input limits field is empty (Grammarly, Scribbr) or states no maximum per text check in words or characters (Pangram states file counts and media sizes, Originality.ai states a 100-word minimum, Surfer a 100-word minimum and "no limits" on how often the free tool is used, Hive states 2,048-character processing chunks, GLTR, the GPT-2 Output Detector and DetectGPT publish no character or word limit for pasted text, Codequiry meters source code per submission with no size limit published, Compilatio meters Studium in credits of 250 words with no document size limit published, WasItAI states image size and pixel limits, Reality Defender states file sizes in megabytes and kilobytes, including .txt up to 900KB, Grammarly Authorship and Turnitin Clarity record writing rather than taking an upload, SynthID, Sightengine, Deepware and Sensity state media sizes and durations, and Turnitin Clarity's 300-word figure is a minimum). Full rules: methodology page; no record in the studies index measures input limits, and the question page on free AI detectors with no limit reads the same field.
Why numbers differ between sources
Vendors state limits in different units on different pages. GPTZero states 10,000 characters per scan on its FAQ for the widget and 150,000 with a free account (gptzero.me/faq), while its blog states a free plan of 10,000 words per month (gptzero.me); a per-scan ceiling and a monthly allowance are separate facts. ZeroGPT states characters per detection on the plan cards and words per month in the compare table (zerogpt.com/pricing). Sapling states 100,000 characters for a web check and 200,000 per API request (sapling.ai, sapling.ai/docs). All read 8 November 2026; the Updated line is the date the page last changed.
A limit is a count of words or characters, and several vendors also exclude kinds of text from the count: Turnitin's model eliminates lists, headers, references, tables and code before scoring, as the bullet points and code question, added 9 November 2026, records for 10 cards. The grant reviewers view and the bulk and batch view, added the same day, read these limits for a proposal and for a batch.
The ceilings above apply to text the detector can read; a scanned PDF carries page images, and the scanned PDF question records that no card in the index states OCR of one, with Turnitin's FAQ of 3 November 2026 saying AI detection focuses on typed text (read 10 November 2026).
Questions about AI detector input limits
Which AI detector accepts the most text in one check?
Detecting-AI.com, at 1,000,000 characters per scan on every paid plan, about 182,000 words at 5.5 characters a word (detecting-ai.com, read 10 November 2026); AI or Not states the largest API figure, 500,000 characters of text per request, about 91,000 words (docs.aiornot.com, read 9 November 2026). Among free web tools, Decopy states 200,000 characters per check with no account (decopy.ai/ai-detector, read 9 November 2026), the same figure as Winston AI, at 200,000 characters per scan on its paid plans, about 36,000 words (gowinston.ai/pricing, read 8 November 2026); Sapling states the same 200,000 characters for one API request (sapling.ai/docs).
Is there an AI detector with no word limit?
Four cards state a tier with no limit on 8 November 2026, and Surfer's help centre, read 9 November 2026, says the free detector can be used "as often as you need" with no limits on use, which is a frequency statement rather than a size ceiling (surfer.ai): ZeroGPT ENTERPRISE "unlimited" (zerogpt.com/pricing), QuillBot Premium "no word limit for detection scans" (quillbot.com), Copyleaks "no character max" on the web platform (copyleaks.com) and Phrasly "no character limits on detection" (phrasly.ai). Only Phrasly's is a free tier. GLTR and the GPT-2 Output Detector, read 9 November 2026, publish no limit at all for pasted text, which is a missing figure rather than a stated unlimited tier (gltr.io, huggingface.co).
How many words is 200,000 characters?
About 36,000 words at 5.5 characters a word, the one conversion this ranking uses and the rate every page of the index uses since 9 November 2026; the page converted at six characters a word (about thirty-three thousand) before that date. The figure is an estimate; Winston AI and Sapling state 200,000 characters, and neither states a word count (gowinston.ai/pricing, sapling.ai/docs, read 8 November 2026).
Why is ZeroGPT in the unlimited table and not ranked first?
Because ZeroGPT's largest stated tier, ENTERPRISE, is "unlimited", which has no position on a number line; its largest numeric figures, 350,000 characters per detection on EXPERT and 500,000 on the API VIP plan, are shown in the second table (zerogpt.com/pricing, read 8 November 2026).
What is the largest free check?
GPTZero takes 150,000 characters with a free account, about 27,000 words at 5.5 characters a word (gptzero.me/faq, read 8 November 2026). Copyleaks states a free scan up to 25,000 characters and ZeroGPT 15,000 characters per detection on FREE (copyleaks.com, zerogpt.com/pricing).
Which cards are omitted, and why?
Nineteen at index version 2026.11.13; the ten below were the omissions at 22 cards, and the recount of 9 November 2026 added nine more, listed in the method section above: Surfer (100-word minimum, no ceiling), DetectGPT (no product limits), Compilatio (credits of 250 words, no document size limit), Grammarly Authorship and Turnitin Clarity (process records), and SynthID, Sightengine, Deepware and Sensity (media sizes and durations). Grammarly and Scribbr state no input limit field at all. Eight state a field without a maximum per text check in words or characters: Pangram states up to 100 files at a time and media sizes (pangram.com), Originality.ai a 100-word minimum (originality.ai/pricing) and Hive text processing in 2,048-character chunks (docs.thehive.ai), all read 8 November 2026; GLTR publishes no character or word limit for its demo (gltr.io), the GPT-2 Output Detector no character limit for its local web UI (huggingface.co), Codequiry meters source code per submission with no file size limit published (codequiry.com), WasItAI states a maximum image size of 8MB and 10000x10000px (wasitai.com), and Reality Defender states file sizes by type, including .txt up to 900KB (docs.realitydefender.com), all read 9 November 2026.
Do minimum lengths matter as much as maximums?
For short texts, yes: Turnitin requires 300 words of prose, JustDone 350 characters, Sapling says results are more accurate after about 50 words, and Originality.ai sets a 100-word minimum, all read 8 November 2026; Surfer requires 100 words, Smodin 1,000 characters, AI or Not 250 characters, and TruthScan and Undetectable AI recommend 200 words, all read 9 November 2026 (surfer.ai, smodin.io, docs.aiornot.com, truthscan.com, help.undetectable.ai). The minimum input ranking, added 9 November 2026, sorts the ten stated requirements from Phrasly's 30 words up, and the question page on short texts tables every stated minimum (guides.turnitin.com, justdone.com, sapling.ai, originality.ai/pricing). A paragraph under those floors returns no result or a low-confidence one. The minimum word count question, added 9 November 2026, tables the ten hard minimums and five recommendations with the 5.5 characters a word conversions.
How often is this ranking updated?
The ranking was created on 9 November 2026 from limits checked on 8 November 2026, recounted the same evening for the sixteen cards added on 9 November 2026 in the third and fourth waves, and changes whenever a vendor's input limits wording changes on the page recorded on the card. The next scheduled recheck of every card is 8 December 2026 (methodology).
Sources and revisions
Sources
- Winston AI pricinggowinston.ai/pricing
- Sapling detector pagesapling.ai/ai-content-detector
- Sapling API docssapling.ai/docs/api/detector
- Sapling pricingsapling.ai/pricing
- Turnitin FAQsguides.turnitin.com/hc/en-us/articles/284...
- Proofademic detector pageproofademic.ai/ai-text-detector
- Proofademic pricingproofademic.ai/pricing
- GPTZero FAQgptzero.me/faq
- GPTZero blog, 25 October 2026gptzero.me/news/best-ai-detectors
- JustDonejustdone.com/ai-detector
- Walter Writes pricingwalterwrites.ai/pricing
- ZeroGPT pricingzerogpt.com/pricing
- QuillBotquillbot.com/ai-content-detector
- Copyleakscopyleaks.com/ai-content-detector
- Phraslyphrasly.ai/ai-detector
- Pangrampangram.com
- Originality.ai pricingoriginality.ai/pricing
- Hive file formatsdocs.thehive.ai/docs/supported-file-formats
- AIDetectPlus home page and detectoraidetectplus.com
- GLTRgltr.io
- GPT-2 Output Detector model cardhuggingface.co/openai-community/roberta-b...
- Codequiry AI code detectioncodequiry.com/detect-ai-written-code
- WasItAIwasitai.com
- Reality Defender SDK quickstartdocs.realitydefender.com/sdks/quickstart
- AI or Not text API documentationdocs.aiornot.com/api-reference/reports-by...
- TruthScan text API documentationtruthscan.com/truthscan-ai-text-detection...
- TruthScan pricingtruthscan.com/pricing
- Undetectable AI detector API articlehelp.undetectable.ai/en/article/detector-...
- Undetectable AI detectorundetectable.ai/ai-detector
- Smodin AI detectorsmodin.io/ai-content-detector
- Smodin API referencedocs.smodin.io/reference/v2-ai-detection-...
- Writer AI content detector, archived 9 December 2025web.archive.org/web/20251209115654/https:...
- BrandWell detector page (resolves to the Moxby Mod)brandwell.ai/ai-content-detector
- BrandWell detector page, archived 9 August 2026web.archive.org/web/20260809041054/https:...
- Quetext AI detector and pricingquetext.com/ai-content-detector
- Surfer AI detectorsurfer.ai/ai-content-detector
Revision log
- Ranking created
- 22-card recount, aidetectplus ranked eighth
- Third and fourth wave cards recounted, ai or not ranks first at 500,000 characters per api request, truthscan, undetectable ai, smodin, writer, brandwell and quetext join, 15 of 38 qualify
- Link to the minimum input ranking added
- Every character-to-word conversion moved from 6 to 5
- Minimum word count question linked
- Eighth-wave pages linked
- Nine wave 9 cards recounted
- Twelve wave 10 cards recounted
- Nine wave 11 cards recounted
- Twelfth-wave pages linked
). 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.: register plate](/img/independently-tested-480.webp)
