Sapling vs GPTZero AI detector: two text detectors with three shared studies, on the same 14 fields

AIDI-C-15

Sapling and GPTZero share three study records, the most of any pair in the index that includes Sapling: Walters 2023 ran both on the same 126 documents and classified 81% correctly with GPTZero's Pro plan against 65% with Sapling, flagging 2% of 42 human papers with GPTZero and 0% with Sapling, where Sapling returned "uncertain" on 31% of the human papers (DOI 10.1515/opis-2022-0158, Table 4, read 8 November 2026); Kar 2024 read one ChatGPT 3.5 article at 100% AI on Sapling and 97% on GPTZero, and after a QuillBot paraphrase Sapling held 100 while GPTZero fell to 29 (DOI 10.1177/02537176241247934, Table 1, read 9 November 2026); Liang 2023 ran both among seven detectors on 91 TOEFL essays and printed the per-detector false positive rates only as a chart (arXiv 2304.02819, Fig. 1a, read 8 November 2026). GPTZero carries 16 study records to Sapling's 3, GPTZero's FAQ states a 99.9% (accuracy) figure and Sapling's detector page states no single number (gptzero.me/faq, sapling.ai, read 8 November 2026). The two cards are read side by side below on the same 14 fields, with the vendor's wording and the page it came from.

Updated Fields checked 8 November 2026

Short answer

Sapling states 16 languages, a free check of the first 2,000 characters with no cap on checks, Pro at $25 a month or $12 a month billed annually, and an API from $0.005 per 1,000 characters with a $5 minimum and a zero_retention flag; GPTZero states 5 languages, a free plan of 10,000 words a month, premium plans starting at $12.99 a month by the vendor's blog, a documented API with no price on the card, and a policy under which API documents are not stored while dashboard inputs are (sapling.ai/pricing, sapling.ai/privacy, gptzero.me/news/best-ai-detectors, gptzero.me/privacy-policy.html, read 8 November 2026). On human text in the one shared study that printed both rates, Sapling flagged 0 of 42 papers and GPTZero 2% of them, Walters 2023 (Walters 2023); GPTZero's wider record runs from 50% false accusations in 18 human documents in Weber-Wulff 2023 to 0 of 495 pre-2022 human passages in Epoch AI 2026 (Weber-Wulff 2023; Epoch AI 2026). No detector score should be the only evidence in a misconduct, hiring or publishing decision.

How this comparison is built

Sapling vs GPTZero AI detector: two text detectors with three shared studies, on the same 14 fields: register plate with the page title and its register ID
The comparison plate: register ID and title, drawn from the index data.

How do Sapling and GPTZero compare on the 14 fields?

Sapling vs GPTZero: the 14 fields side by side
FieldSaplingGPTZero
1. Name, vendor, official domainSapling AI Detector, Sapling Intelligence, Inc., sapling.aiGPTZero, GPTZero LLC, gptzero.me
2. Modalities detectedtexttext, image (AI Vision, specs not published)
3. Generators coveredGPT-5, ChatGPT, the OpenAI GPT family, Claude 4.5, Claude 5.5, Gemini 2.5, Qwen3, DeepSeek-V3, Kimi, Llama, MistralClaude, ChatGPT, GPT-3, GPT-4, GPT-5, GPT-6, Gemini, Llama, Deepseek, Grok and "AI services based on those models"
4. Languages supported16, all named: English, Spanish, Portuguese, French, German, Italian, Dutch, Japanese, Polish, Chinese, Arabic, Russian, Swedish, Korean, Turkish, Romanian5: English, German, Portuguese, French, Spanish
5. Input limitsWeb: free check covers the first 2,000 characters; Pro and Enterprise up to 100,000 characters per check; files as PDF or Word with selectable text, 5 per batch on Free and 10 on Pro; API up to 200,000 characters per request, results under 500 characters flagged low_confidenceWidget: 10,000 characters per scan without an account; a free account takes texts up to 150,000 characters or a batch of up to 250 files; file types and size not published
6. Free tierFree plan at $0 a month: 2,000 characters per check, 5 files per batch, number of checks not capped; a 1-month Pro trial on registration; API keys get 50,000 characters a day and 250,000 a monthFree plan covering 10,000 words a month (vendor blog, 25 October 2026); 10,000 characters per scan without login
7. Paid plansPro $25 a month, or $12 a month billed annually (100,000 characters per check, 10 files per batch); Enterprise from 10 seats at $15 a seat a month, with SSO, SCIM and self-hostingPremium plans "starting at $12.99/month" by the vendor's blog; the individual plan cards on gptzero.me/pricing did not render on 8 November 2026; Professional Team Plan 500,000 words per member a month, minimum 2 seats, price not rendered
8. APIsapling.ai/docs/api/detector: from $0.005 per 1,000 characters with volume discounts, one-time $5 minimum; subscribed keys default to 10 million characters a daydocs.gptzero.me: documented; no API price recorded on the card
9. IntegrationsChrome extension for ChatGPT, Word add-in, browser extensions (Chrome, Firefox, Edge, Opera, Brave), Google Docs, Gmail, Outlook, 73 integrations including Zendesk, Salesforce, Slack, Notion and WordPress, MCP serverChrome extension (Gmail, Google Docs, Classroom, social media), Google Docs, Google Classroom, Canvas, Zapier, API
10. Vendor accuracy and false positive claims"Accuracy varies with the dataset, text length, and decision threshold"; no single number; false positives "fell by 60%" on internal held-out evaluations after the 17 September 2026 model upgrade, no absolute rate"An accuracy rate of 99.9%" (accuracy) "validated on RAID, detecting 95.7% of AI texts while only incorrectly predicting 1% of human texts as AI"; mixed documents 96.5%; the 4o launch post of 18 October 2026 states a false positive rate of 1 in 10,402 on student writing
11. Independent studies3 records: Liang 2023, Walters 2023, Kar 202416 records: Weber-Wulff 2023, Liang 2023, Walters 2023, Elkhatat 2023, Perkins 2024, RAID 2024, Chaka 2023, Krishna 2023, Orenstrakh 2023, Jabarian and Imas 2025, Van Vlasselaer 2026, Pegoraro 2023, Kar 2024, Alshammari and Rao 2025, DAMAGE 2025, Epoch AI 2026
12. Data retention and privacy statementPrivacy policy: text entered "is stored by us and used only to improve the Service"; API results cached on the text for 3 days; a zero_retention flag scores a text without a copy being kept; servers on AWS us-west; security page states SOC Type 2 and GDPRPrivacy policy (20 May 2026): "We do not store or collect the documents passed into any calls to our API. However, we do store inputs from calls made from our dashboard", used in aggregate; FERPA, GDPR and CCPA named; personal data deleted on request
13. Launch year and ownershipNot stated; Sapling Intelligence, Inc. (California), developed by former researchers at Berkeley, Stanford, Meta and Google; no parent stated2023; GPTZero LLC; announced on 23 June 2026 that it plans to join Superhuman; the footer lists Superhuman and Grammarly
14. Rankings the card appears inIndependently tested, false positive rate, languages, non-English evidence, input limits, file types, API pricing, platform integrations, generators covered, data retentionIndependently tested, false positive rate, false negative rate, adversarial robustness, studies by content type, claim vs evidence gap, free tiers, languages, non-English evidence, input limits, platform integrations, LMS coverage, generators covered, launch year

Which studies tested Sapling or GPTZero?

Three studies tested both, and all three are in the table first. Walters 2023 ran 16 detectors on 42 human first-year composition papers, 42 GPT-3.5 papers and 42 GPT-4 papers between 25 June and 12 July 2023, and printed each tool's correct, incorrect and uncertain shares per class (DOI 10.1515/opis-2022-0158, Table 4, read 8 November 2026). Kar 2024 ran ten free detectors on one ChatGPT 3.5 article of under 500 words and three paraphrases of it (DOI 10.1177/02537176241247934, Table 1, read 9 November 2026). Liang 2023 ran seven detectors on 91 TOEFL essays and 88 US eighth-grade essays, reported an average false positive rate of 61.22% on the TOEFL set across the seven, and printed the per-detector rates as a bar chart only (arXiv 2304.02819, Fig. 1a, read 8 November 2026). GPTZero's other 13 records run from Weber-Wulff 2023, 54 documents in March 2023 (arXiv 2306.15666, Tables 5, 9 and 10, read 9 November 2026), to Epoch AI 2026, 495 pre-2022 human passages and 297 AI passages per prompt on model 2026-05-11-base (epoch.ai, read 9 November 2026). Sapling has no record the two do not share.

SourceSaplingGPTZero
Walters 2023, 126 documents overall (Table 4)65% correct, 7% incorrect, 28% uncertain81% correct, 4% incorrect, 15% uncertain (Pro plan)
Walters 2023, 42 GPT-3.5 and 42 GPT-4 papers93% correct on GPT-3.5, 33% on GPT-498% correct on GPT-3.5, 57% on GPT-4
Walters 2023, 42 human papers69% correct, 0 incorrect, 31% uncertain88% correct, 2% incorrect
Kar 2024, one ChatGPT 3.5 article and three paraphrases (Table 1)100% AI on the original; 100 after the QuillBot, Grammarly and ChatGPT paraphrases97% AI on the original; 29 after QuillBot, 38 after Grammarly, 97 after ChatGPT
Liang 2023, 91 TOEFL essays and 88 US eighth-grade essays, accessed 15 March 2023One of seven detectors; the paper reports a 61.22% average false positive rate across the seven on the TOEFL essays and per-detector rates only in Fig. 1aOne of seven detectors; the same average; 18 of 91 TOEFL essays flagged by all seven
Weber-Wulff 2023, 54 documents (Tables 5, 9 and 10)Not tested29 of 54 correct (54%, rank 8 of 14); 9 false accusations in 18 human documents (50%, the highest); 3 of 36 AI documents undetected (8.3%, the lowest)
Elkhatat 2023, 15 paragraphs per class and 5 human controls (Table 4)Not tested14 of 15 GPT-3.5 and 4 of 15 GPT-4 paragraphs flagged; 1 of 5 human controls flagged
Perkins 2024, 104 AI and 10 human samples (Tables 7, 8 and 10)Not tested26.3% baseline mean accuracy, the lowest of seven; 26.4% on unmanipulated AI text falling to 16.7%; 1 of 10 human samples falsely accused; 58 of 104 AI samples undetected
RAID 2024, human-written documents (Tables 4, 5 and 6)Not testedFalse positive rate 0.03% at threshold 0.25 and 0.00% at 0.5 and above; 98.8% accuracy at a fixed 5% FPR on open-source chat models with greedy decoding, 9.4% on non-chat models with sampling; lost 0.3% under the homoglyph attack
Chaka 2023, 21 English chatbot responses and 30 machine-translated onesNot tested8 of 21 English responses classified as AI, 10 as human, 3 indecisive (rank 3 of 5); all 30 translated responses classified as human
Krishna 2023, 3,000 Wikipedia prompts at a fixed 1% FPR (Table 1)Not tested13.9% of GPT-2 XL, 8.7% of OPT-13B and 7.1% of GPT-3.5 text detected, falling to 1.2%, 1.0% and 1.8% after one DIPPER paraphrase
Orenstrakh 2023, 114 human and 20 ChatGPT submissions (Tables 6, 8 and 9)Not tested54.39% on human and 45.00% on ChatGPT submissions, rank 8 of 8; 52 false positives in 114; 70% to 20% on 10 QuillBot-paraphrased texts
Jabarian and Imas 2025, 1,992 human and 1,992 AI passages per model (Tables 3, 4, 6 and B.4)Not testedFalse positive rate 0.0071 at every fixed threshold; false negative rate 0.0020 to 0.0296 before and 0.2552 to 0.7730 after StealthGPT rewrites
Van Vlasselaer 2026, 40 papers per class (Tables 4 and 5)Not tested70% of 40 fully AI-generated papers scored under 20%; 0 of 40 hybrid and 1 of 40 humanised papers within range; all 40 human papers correct
Pegoraro 2023, ChatGPT and human Q&A benchmark (Table I)Not testedTrue positive rate 27.3%, true negative rate 93.5%
Alshammari and Rao 2025, 49 answers per set (Tables III and V)Not testedMean AI score 100% on DeepSeek-V3 answers; mean human score 98.2% on 49 human answers; 94.1% on DeepThink text, 52% after a humanized paraphrase
DAMAGE 2025, AI-written student essays (Tables 3 and 7)Not testedTrue positive rate 99.73% and 60.04% after humanizing at a fixed 5% FPR; 95.60% and 34.53% at the default threshold; 1.47% default-threshold false positive rate on human essays
Epoch AI 2026, model 2026-05-11-base at its default verdictNot tested0 of 495 pre-2022 human passages flagged; 2 of 297 plain-prompt AI passages missed (0.67%); 32 of 297 (10.77%) missed when written in a named author's style
Vendor statement, sapling.ai (read 8 November 2026)No single accuracy figure; false positives fell by 60% on internal held-out evaluations after the 17 September 2026 upgrade
Vendor statement, gptzero.me (read 8 November 2026)99.9% (accuracy) and 1% of human texts flagged, "validated on RAID" (FAQ); 1 in 10,402 false positive rate on student writing (4o launch post, 18 October 2026)

Results as each study reports them: DOI 10.1515/opis-2022-0158, DOI 10.1177/02537176241247934, arXiv 2304.02819, arXiv 2306.15666, DOI 10.1007/s40979-023-00140-5, arXiv 2403.19148, arXiv 2405.07940, DOI 10.37074/jalt.2023.6.2.12, arXiv 2303.13408, arXiv 2307.07411, NBER w34223, DOI 10.1007/s40979-026-00226-w, arXiv 2304.01487, arXiv 2507.17944, arXiv 2501.03437, epoch.ai, read 8 and 9 November 2026. "Not tested" means the paper did not run that tool. The false positive rate ranking sorts Sapling on its Walters 2023 zero and GPTZero on its RAID 2024 zero at threshold 0.5, the lowest printed rate per card, and lists GPTZero's eight other human-text results beside it.

What do the two cost over 12 months?

PlanMonthly as statedAnnual as statedInclusions as statedChecked on
Sapling Free$0AI detector checks up to 2,000 characters per check, 5 files per batch, number of checks not capped; 1-month Pro trial on registration8 November 2026
Sapling Pro (monthly)$25 per month$300 at twelve monthly paymentsChecks up to 100,000 characters, 10 files per batch, advanced writing assistance, email and ticket support8 November 2026
Sapling Pro (annual)$12 per month, billed annually$144 at the stated monthly rate times twelveSame as Pro (monthly)8 November 2026
Sapling Enterprise"Starts at 10 seats, $15/seat/month"$1,800 at ten seats, the stated rate times twelve100,000 characters per check, team controls, SSO and SCIM, self-hosted option, dedicated support8 November 2026
Sapling APIFrom $0.005 per 1,000 characters, volume discounts to $0.00375; one-time $5 minimumMeteredSubscribed keys default to 10 million characters a day and 25 million a month; new keys get 50,000 characters a day free8 November 2026
GPTZero Free$010,000 words a month (vendor blog, 25 October 2026); 10,000 characters per scan without login8 November 2026
GPTZero Premium (individual plans)"Premium plans starting at $12.99/month" (vendor blog; billing period not stated)Not computed: the pricing page's plan cards did not render and the billing period is not statedNot readable on 8 November 20268 November 2026
GPTZero Professional Team PlanNot rendered"Annual (Save 45%)" toggle shown, figure not rendered500,000 words per member a month; minimum 2 seats8 November 2026
GPTZero APINot recorded on the cardDocumented at docs.gptzero.me8 November 2026

Prices in USD as shown on sapling.ai/pricing and as readable from gptzero.me/pricing and gptzero.me/news/best-ai-detectors, checked 8 November 2026. Sapling's annual figures are the index's multiplication of the stated monthly rate by twelve, because the page prints no yearly total; GPTZero's are not computed, because the one readable individual figure is a blog's starting price with no billing period. Through the API, Sapling's $0.005 per 1,000 characters is $0.0275 per 1,000 words at 5.5 characters a word, the figure the API pricing ranking carries, and GPTZero sits in that ranking's unpriced table because no API price is recorded on its card.

One of the comparisons in the compare hub. The cards are Sapling and GPTZero, and the home board answers which AI detectors exist and who has tested them? A developer choosing between the two per-character and per-word APIs, or an instructor weighing the detector with the long outside record against the one that charges nothing on the web for a 2,000-character check, is the reader this pair is written for; the API pricing ranking converts Sapling's rate to the index unit and the free API tiers question lists both free allowances.

Which card fits which reader?

  • Choose Sapling ifthe check runs from code and the bill must be predictable: a published rate from $0.005 per 1,000 characters with a $5 minimum and 250,000 free characters a month on a new key, against GPTZero's documented API with no price on its card (sapling.ai/pricing, docs.gptzero.me, read 8 November 2026).
  • Choose GPTZero ifyou want the longer outside record: 16 study records from March 2023 to July 2026 against Sapling's 3, including 0 of 495 pre-2022 human passages flagged in Epoch AI 2026 and a 0.0071 false positive rate on 1,992 passages in Jabarian and Imas 2025 (Epoch AI 2026; Jabarian and Imas 2025).
  • Choose Sapling ifthe text has been paraphrased by QuillBot: Sapling held 100 on all three paraphrases in Kar 2024 where GPTZero fell from 97% to 29 after the QuillBot version (DOI 10.1177/02537176241247934, Table 1, read 9 November 2026), a single-text result that the DAMAGE 2025 figure for GPTZero, 60.04% after humanizing at a fixed 5% FPR, measures on a larger set GPTZero alone was in (DAMAGE 2025).
  • Choose GPTZero ifthe check must run inside Google Classroom or Canvas: both are named on its page, where Sapling names Google Docs, Gmail, Outlook and Word but no LMS (gptzero.me, sapling.ai/integrations, read 8 November 2026).
  • Choose Sapling ifyou check text in Italian, Dutch, Japanese, Polish, Chinese, Arabic, Russian, Swedish, Korean, Turkish or Romanian: Sapling names 16 languages and GPTZero 5 (sapling.ai, gptzero.me, read 8 November 2026).
  • Choose Sapling ifthe submitted text must not be kept: the API's zero_retention flag scores without a copy and teams can disable retention on every request, where GPTZero stores dashboard inputs and does not store API documents (sapling.ai/privacy, gptzero.me/privacy-policy.html, read 8 November 2026).
  • Choose GPTZero ifyou want a free allowance counted in words rather than characters per check: 10,000 words a month against Sapling's 2,000 characters per check with uncapped checks (gptzero.me/news/best-ai-detectors, sapling.ai/pricing, read 8 November 2026).
  • IfChoose neither on the vendor's accuracy figure alone: Sapling publishes none and says accuracy "varies with the dataset, text length, and decision threshold", and GPTZero's 99.9% (accuracy) sits beside study results from 26.3% in Perkins 2024 to 98.8% in RAID 2024 (sapling.ai, gptzero.me/faq, read 8 November 2026; Perkins 2024; RAID 2024).

How does the evidence compare, tool by tool?

Sapling

Register ID
AIDI-D-0010
Study records
3 in the index: Liang 2023, Walters 2023, Kar 2024
Shared with GPTZero
3: Liang 2023, Walters 2023, Kar 2024
Vendor accuracy claim
Accuracy varies with the dataset, text length, and decision threshold. Measure false positives and missed detections on examples from your workflow. (No single accuracy number published.)
Free tier
Free plan $0 / month: AI detector checks up to 2,000 characters per check, 5 files per detector batch; count of checks per period not capped. Free 1-month trial of Sapling Pro on registration (no credit card). API: new keys get 50,000 characters/day and 250,000 characters/month.
Languages
16 stated
Read the Sapling card

GPTZero

Register ID
AIDI-D-0001
Study records
16 in the index: Weber-Wulff 2023, Liang 2023, Walters 2023, Elkhatat 2023, Perkins 2024, RAID 2024, Chaka 2023, Krishna 2023, Orenstrakh 2023, Jabarian and Imas 2025, Van Vlasselaer 2026, Pegoraro 2023, Kar 2024, Alshammari and Rao 2025, DAMAGE 2025, Epoch AI 2026
Shared with Sapling
3: Liang 2023, Walters 2023, Kar 2024
Vendor accuracy claim
An accuracy rate of 99.9% when spotting AI-generated text vs. human writing, validated on RAID, detecting 95.7% of AI texts while only incorrectly predicting 1% of human texts as AI; mixed documents with a 96.5% accuracy rate. (FAQ)
Free tier
Free plan covering 10,000 words per month (vendor blog, Oct 25, 2026); 10,000 characters per scan without login
Languages
5 stated
Read the GPTZero card

How this comparison is built

Cards compared
Sapling and GPTZero
Fields
The 14 fields of every card, in card order
Shared study records
3: Liang 2023, Walters 2023, Kar 2024
Fields checked
8 November 2026
Records in the studies index
35 records

Read the full methodology

Both columns are drawn from the same 14 fields in the index data file, read on 8 November 2026 from the vendor page recorded on each card, and from the studies index, where each result is copied as the paper reports the figure, with the base stated. Three records ran both tools, Walters 2023, Kar 2024 and Liang 2023, so the evidence table opens with their five rows and lists GPTZero's other thirteen records with "Not tested" in the Sapling column; Liang 2023 is kept although it prints no per-detector number, because both cards cite it, and the row says what the paper gives. Each vendor's own statement sits in a labelled row, and GPTZero's blog-sourced plan price is flagged in the cost table rather than silently carried. The rules, the recheck schedule (next pass 8 December 2026) and the correction route are on the methodology page; the claim vs evidence gap ranking pairs GPTZero's 99.9% with its 98.8% in RAID 2024 and records no gap for Sapling because Sapling states no claim figure.

Where each figure comes from

Values as each vendor states them on sapling.ai/ai-content-detector, sapling.ai/pricing, sapling.ai/privacy, sapling.ai/integrations, sapling.ai/security, gptzero.me, gptzero.me/faq, gptzero.me/pricing, gptzero.me/news/best-ai-detectors and gptzero.me/privacy-policy.html, read 8 November 2026 (the Sapling security page on 9 November 2026). GPTZero's individual plan prices are the one field on either card taken from a blog post rather than the pricing page, because the pricing page's plan cards did not render to the index's fetcher; the card says so, and the figure should be re-read in a browser before it is relied on.

Questions about Sapling and GPTZero

Did any study test Sapling and GPTZero on the same text?

Yes, three: Walters 2023 classified 81% of 126 documents correctly with GPTZero and 65% with Sapling, flagging 2% of 42 human papers with GPTZero and 0 with Sapling (DOI 10.1515/opis-2022-0158, Table 4); Kar 2024 read one ChatGPT article at 100% on Sapling and 97% on GPTZero (DOI 10.1177/02537176241247934, Table 1); Liang 2023 ran both among seven detectors on 91 TOEFL essays and printed per-detector rates only as a chart (arXiv 2304.02819, Fig. 1a), read 8 and 9 November 2026.

Which has the lower measured false positive rate?

Both carry a zero as their lowest printed rate: Sapling's 0 of 42 human papers in Walters 2023, with 31% returned uncertain (DOI 10.1515/opis-2022-0158, Table 4), and GPTZero's 0.00% at threshold 0.5 on RAID's human documents (arXiv 2405.07940, Table 4), read 8 November 2026. GPTZero's other human-text results run from 2% in Walters 2023 to 50% of 18 documents in Weber-Wulff 2023 and 52 of 114 in Orenstrakh 2023.

Which has more study records?

GPTZero, with 16 to Sapling's 3, by the study lists on the two cards read 10 November 2026 (GPTZero card, Sapling card); every Sapling record is shared with GPTZero.

Which is cheaper through the API?

Sapling publishes its rate, from $0.005 per 1,000 characters with a $5 minimum, $0.0275 per 1,000 words at 5.5 characters a word; GPTZero documents an API and publishes no price that the card records (sapling.ai/pricing, docs.gptzero.me, read 8 November 2026), so the API pricing ranking ranks Sapling and lists GPTZero as unpriced.

Does either keep the text you submit?

Sapling's policy says entered text "is stored by us and used only to improve the Service", API results are cached for 3 days, and a zero_retention flag prevents a copy being kept; GPTZero's FAQ says API documents are not stored while dashboard inputs are, used in aggregate (sapling.ai/privacy, gptzero.me/privacy-policy.html, read 8 November 2026).

What happened to each after paraphrasing?

Sapling held 100 on all three paraphrases of one ChatGPT article in Kar 2024, and GPTZero fell from 97% to 29 after the QuillBot paraphrase and to 38 after the Grammarly one (DOI 10.1177/02537176241247934, Table 1, read 9 November 2026). GPTZero alone was measured on larger sets: 70% to 20% on 10 QuillBot-paraphrased texts in Orenstrakh 2023 and 99.73% to 60.04% after humanizing in DAMAGE 2025 (Orenstrakh 2023; DAMAGE 2025).

Which supports more languages?

Sapling, with 16 named to GPTZero's 5 (sapling.ai, gptzero.me, read 8 November 2026). The non-English evidence runs the other way: GPTZero has 5 records with non-English or non-native text, among them all 30 translated responses classified as human in Chaka 2023, and Sapling 1, the TOEFL essays of Liang 2023 (Chaka 2023; Liang 2023).

Sources and revisions

Sources

  1. Sapling AI detector (languages, generators, limits, claims)sapling.ai/ai-content-detector
  2. Sapling pricingsapling.ai/pricing
  3. Sapling privacy policysapling.ai/privacy
  4. Sapling API docssapling.ai/docs/api/detector
  5. Sapling integrationssapling.ai/integrations
  6. Sapling security (read 9 November 2026)sapling.ai/security
  7. GPTZero home pagegptzero.me
  8. GPTZero FAQ (claims, limits, generators)gptzero.me/faq
  9. GPTZero pricinggptzero.me/pricing
  10. GPTZero blog, free plan and starting pricegptzero.me/news/best-ai-detectors
  11. GPTZero privacy policygptzero.me/privacy-policy.html
  12. GPTZero 4o launch postgptzero.me/news/introducing-gptzero-4o
  13. GPTZero API docsdocs.gptzero.me
  14. Walters 2023degruyterbrill.com/document/doi/10.1515/o...
  15. Kar 2024 (read 9 November 2026)doi.org/10.1177/02537176241247934
  16. Liang 2023arxiv.org/pdf/2304.02819v3
  17. Weber-Wulff 2023 (read 9 November 2026)arxiv.org/pdf/2306.15666
  18. Elkhatat 2023link.springer.com/content/pdf/10.1007/s40...
  19. Perkins 2024 (read 9 November 2026)arxiv.org/pdf/2403.19148
  20. RAID 2024, Dugan et al.arxiv.org/pdf/2405.07940
  21. Chaka 2023 (read 9 November 2026)doi.org/10.37074/jalt.2023.6.2.12
  22. Krishna 2023 (read 9 November 2026)arxiv.org/abs/2303.13408
  23. Orenstrakh 2023 (read 9 November 2026)arxiv.org/abs/2307.07411
  24. Jabarian and Imas 2025 (read 9 November 2026)nber.org/papers/w34223
  25. Van Vlasselaer 2026 (read 9 November 2026)doi.org/10.1007/s40979-026-00226-w
  26. Pegoraro 2023 (read 9 November 2026)arxiv.org/abs/2304.01487
  27. Alshammari and Rao 2025 (read 9 November 2026)arxiv.org/abs/2507.17944
  28. DAMAGE 2025 (read 9 November 2026)arxiv.org/abs/2501.03437
  29. Epoch AI 2026 (read 9 November 2026)epoch.ai/data-insights/ai-detectors-false...

Revision log

  1. Comparison created at index version 2026

Corrections log

Related reading