Do AI detectors store the text you paste? Every retention clause in the index, quoted and dated
Most do. Of the 72 AI detectors in the index, 7 state that pasted text or uploaded media is not stored at all, 6 state a retention period, from SynthID's 24 hours to Grammarly Authorship's 10 years, 20 keep submissions for as long as the account is active, "as long as necessary" or inside a report without naming a period, 2 (QuillBot and Grammarly) publish no clause about pasted text that the index could read, and 3 run on the reader's own machine so no vendor policy applies, by the policies, documentation and product pages read on 8 and 9 November 2026 and recorded in each card's retention field. The data retention ranking sorts the thirteen cards that state a period or state no storage; this page quotes every card's clause in one table so a reader can see what the other twenty-five say or leave unsaid. The fourth wave of cards on 9 November 2026 added two no-storage statements (Sightengine and Smodin), two periods (TruthScan's 30 days and Compilatio's 5 years) and five policies with no period.
Every clause below is quoted from the page the card cites, with the date the index read it. "Not stored" means the vendor's words say submissions are deleted after analysis or never kept; a period in days is the vendor's number; "no period" means the policy keeps content but names no end date. Where a vendor states one thing for its API and another for its web tool, both are quoted, because a manager pasting into a dashboard is covered by the dashboard's clause. This page is one of the questions answered from the index; its siblings ask which detectors were shut down and which detectors share an owner, and the AI detector register shows every card on the same fields.
| Detector | Stored? | Retention clause, as published | Source, read date |
|---|---|---|---|
| AI or Not | Not stored | "Privacy first: uploaded media are processed and deleted immediately after inference"; free text tool: "No data stored" | docs.aiornot.com, 9 November 2026 |
| Scribbr | Not stored | "we do not store or share your data. Submissions will never be added to our content database" | scribbr.com, 8 November 2026 |
| Phrasly | Not stored | "processed securely and deleted immediately after analysis. We don't store your writing"; the terms grant a perpetual licence over Contributions | phrasly.ai, 8 November 2026 |
| Undetectable AI | Not stored | "Your text is never stored or used to train models"; the privacy policy of 18 March 2023 does not mention submitted text | undetectable.ai, 9 November 2026 |
| WasItAI | Not stored | "We do not store submitted images beyond the duration required to complete the analysis request"; log data kept 90 days | wasitai.com, 9 November 2026 |
| Sightengine | Not stored | "Your content is deleted once it has been processed, unless you ask us to keep it or the law requires us to"; demo files kept as logs "deleted automatically after a limited period"; Feedback API content kept to improve models | sightengine.com, 9 November 2026 |
| Smodin | Not stored | "Your text is never stored or shared" and "We also never store your files" (detector page and FAQ); the terms grant a "perpetual, irrevocable license" over User Content and the privacy policy does not mention submitted text | smodin.io, 9 November 2026 |
| SynthID | 1 day | Uploaded files "are immediately deleted" once results return; a digital signature "is mapped to your user identifier and permanently deleted after 24 hours" | synthid.com, 9 November 2026 |
| Sapling | 3 days (API) | "AI detection results are cached on the entire text for 3 days"; zero_retention: true scores without a copy; web tool content "is stored by us and used only to improve the Service" | sapling.ai, 8 November 2026 |
| TruthScan | 30 days | "Content Submitted for Analysis: Retained for 30 days by default, then deleted, unless retention is required for service improvement or otherwise agreed upon in a customer's service agreement"; Zero Data Retention on Business and above, applied to image and PDF submissions | truthscan.com, 9 November 2026 |
| Hive | 14 days | "Our default policy is to retain customer data for 14 days. We can increase or decrease our retention time upon request" | docs.thehive.ai, 8 November 2026 |
| Compilatio | 5 years | Account data, including documents submitted, stored for "5 years from the end of the contractual relationship" on servers in Chavanod, France; submitted documents "are not used for commercial purposes or reused outside the terms of the agreement with your institution" | support.compilatio.net, 9 November 2026 |
| Grammarly Authorship | 10 years | Report data "remains on our servers for up to 10 years"; shared reports stay "until you delete your account"; deletable per document | support.grammarly.com, 9 November 2026 |
| GPTZero | No period (dashboard); API not stored | "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" | gptzero.me, 8 November 2026 |
| Copyleaks | No period; API not stored | Data may be used "including by training Copyleaks' models", opt-out via support; API content "is not stored, shared, or used to train public AI models" | copyleaks.com, 8 November 2026 |
| Originality.ai | No period | Input text and results "may constitute scan history"; "Unless you opt out, we may use your scan history and scan results to help train and improve our models" | originality.ai, 8 November 2026 |
| Pangram | No period (account lifetime) | "This content is stored for as long as your account is active" and can be deleted from the history tab; zero data retention for the enterprise API | pangram.com, 8 November 2026 |
| Winston AI | No period | "We collect and store any text or images you submit"; deleting a report or the account deletes the content | gowinston.ai, 8 November 2026 |
| ZeroGPT | No period | Content "is retained only for the period reasonably necessary to provide and operate the applicable Service"; not used to train models | zerogpt.com, 8 November 2026 |
| Turnitin | No period (in the report) | The AI result is "retained as part of the similarity report"; the institution chooses whether papers enter the repository | guides.turnitin.com, 8 November 2026 |
| Turnitin Clarity | No period | Writing Process Data (keystrokes, deletions, interactions) is Educational Data retained "for the period necessary to enable the continued use of the Services", then deleted | guides.turnitin.com, 9 November 2026 |
| JustDone | No period | Personal data stored "for as long as it is reasonably necessary"; "Your text is never sold or published" | justdone.com, 8 November 2026 |
| Proofademic | No period | Submitted text "not used for model training or resale"; personal data kept "only as long as necessary" | proofademic.ai, 8 November 2026 |
| Walter Writes | No period | Information retained "for as long as necessary to provide you with the Services"; text processed on self-hosted models | walterwrites.ai, 8 November 2026 |
| AIDetectPlus | No period | Information "may be stored, collected, processed and used in countries other than the Republic of India"; deletion on request | aidetectplus.com, 9 November 2026 |
| Codequiry | No period (indefinite) | Code submitted without the Rapid add-on is added to a matching corpus and "retained indefinitely including after account closure" | codequiry.com, 9 November 2026 |
| Reality Defender | No period | "We will not retain your personal information for longer than is reasonably necessary" | realitydefender.com, 9 November 2026 |
| Writer | No period | "We will only retain your information for as long as we need it to fulfil the purposes for which we collected it"; inputs never used to train models; the detector was sunset on 22 December 2025 | writer.com, 9 November 2026 |
| BrandWell | No period | "We retain information for as long as reasonably necessary for the purposes described in this Policy"; the Moxby listing says submitted text "is sent through the Moxby gateway to the analysis service" | brandwell.ai, 9 November 2026 |
| Surfer | No period | Customer data processed "for the period necessary for the performance of the Contract" and up to 30 days after it ends for backups; nothing stated for text pasted into the free tool | surfer.ai, 9 November 2026 |
| Quetext | No period | FAQ: "Your scanned content is not stored, saved, or used to train any models"; privacy policy section 6.1: User Content may be included in "AI Systems training, fine-tuning, evaluation, and quality-assurance workflows" with an email opt-out; personal data kept "only as long as necessary" | quetext.com, 9 November 2026 |
| Deepware | No period | An uploaded sample "is immediately shared for review by the Service's Partners" under a worldwide, irrevocable licence; "do not send it" if you do not want it shared; personal data kept "only for as long as is necessary" | deepware.ai, 9 November 2026 |
| Sensity | No period | Personal information stored "for as long as you use our Services or as necessary"; biometric data (selfie videos, identification documents) deleted "within three years of your last interaction"; no period for submitted media; submissions deletable through the API | sensity.ai, 9 November 2026 |
| QuillBot | No clause read | The QuillBot privacy page did not load on 8 or 9 November 2026, so the card's retention field is empty and no clause is quoted | quillbot.com, 8 November 2026 |
| Grammarly | No clause read | The AI detector page carries no retention clause for pasted text, so the card's retention field is empty and no clause is quoted; the Authorship card quotes the Grammarly support centre separately | grammarly.com, 8 November 2026 |
| GLTR | Runs locally | Open-source code under the Apache License 2.0; no vendor policy applies to text pasted into a local server | github.com, 9 November 2026 |
| GPT-2 Output Detector | Runs locally | MIT-licensed weights run on the reader's machine; the hosted demo is paused | github.com, 9 November 2026 |
| DetectGPT | Runs locally | MIT-licensed code runs the paper's scripts on the reader's machine; the hosted demo returned 404 on 9 November 2026 | github.com, 9 November 2026 |
Clauses as each vendor publishes them on the page linked in the row; the retention field of every card carries the fuller quotation.
Which AI detectors say they do not store your text?
Seven, by their own words read 8 and 9 November 2026, the last two added with the fourth wave of cards: Sightengine ("Your content is deleted once it has been processed", sightengine.com), Smodin ("Your text is never stored or shared", smodin.io, whose terms grant a perpetual licence over User Content in the same breath), and the five indexed earlier: AI or Not ("uploaded media are processed and deleted immediately after inference", docs.aiornot.com), Scribbr ("we do not store or share your data", scribbr.com), Phrasly ("deleted immediately after analysis", phrasly.ai), Undetectable AI ("Your text is never stored or used to train models", undetectable.ai) and WasItAI, for images ("We do not store submitted images beyond the duration required", wasitai.com). Two of the five carry a second document that reads differently: Phrasly's terms, updated 12 October 2026, grant the company "an unrestricted, unlimited, irrevocable, perpetual license to use, copy, store and exploit your Contributions" (phrasly.ai/terms, read 8 November 2026), and Undetectable AI's privacy policy of 18 March 2023 does not mention submitted text at all (undetectable.ai/privacy, read 9 November 2026). The data retention ranking puts all five at 0 days and breaks the tie on study records.
Which AI detectors state how long they keep it?
Six, and the spread is wide. TruthScan keeps submitted content 30 days by default with a paid Zero Data Retention toggle (truthscan.com, read 9 November 2026), and Compilatio stores submitted documents for 5 years from the end of the institution's contract (support.compilatio.net, read 9 November 2026), both added with the fourth wave of cards. SynthID's Detector portal deletes uploaded files once results return and keeps a digital signature for 24 hours to enforce quotas (synthid.com, read 9 November 2026). Sapling caches an API text for 3 days unless the request sets zero_retention: true, while its web tool stores content with no period (sapling.ai/privacy, read 8 November 2026). Hive keeps customer data 14 days by default, adjustable on request (docs.thehive.ai, read 8 November 2026). Grammarly Authorship keeps a document's process record on its servers "for up to 10 years" once the report is opened, and a shared report "until you delete your account", with per-document deletion offered (support.grammarly.com, read 9 November 2026). Authorship is built to be kept: it is a record a student shows later, not a score, so the long period is the product rather than a side effect.
Which AI detectors train on what you paste?
Two say they may and name the opt-out. Copyleaks' privacy policy, revised 24 August 2026, allows use of data "including by training Copyleaks' models", with direct customers able to opt out through support, while its API page says API content is not used to train public models (copyleaks.com, read 8 November 2026). Originality.ai's policy, amended 17 June 2026, says "Unless you opt out, we may use your scan history and scan results to help train and improve our models" (originality.ai, read 8 November 2026). GPTZero stores dashboard inputs and uses them "in aggregate" to improve the service (gptzero.me, read 8 November 2026). Seven vendors state the opposite: ZeroGPT, Pangram, Phrasly, Proofademic, WasItAI, Undetectable AI and Writer each say submitted content is not used for training, on the pages linked in the table. Turnitin Clarity's administrator FAQ says its pre-trained models "do not use or retain any student or assignment data submitted to Turnitin for AI training", while anonymised assignment data is used to improve the AI assistant (guides.turnitin.com, read 9 November 2026).
An API clause does not cover the web tool, and a "not stored" line on a product page does not override the terms. GPTZero, Copyleaks and Sapling each state a stricter rule for the API than for the dashboard, and Phrasly's product page and terms pull in opposite directions, so the clause that applies is the one for the surface the reader uses.
Storage and removal are different fields: 17 cards state a route to delete submitted text or media and 10 state that the input is not kept past the check, on the deletion routes question, added 9 November 2026; the no-account view and the compliance view, added the same day, read these statements for checks made without an account and beside each vendor's named standards.
Questions about AI detectors and stored text
Do AI detectors keep the text you submit?
Most do, per the policies read 8 and 9 November 2026: 15 of the 72 cards keep submissions with no stated period, 4 state a period (24 hours to 10 years), 5 say content is not stored, 2 publish no clause the index could read, and 2 run on the reader's machine. The data retention ranking sorts the nine with a stated period or a no-storage line.
Which AI detector keeps data the longest?
Codequiry's policy, updated 22 October 2026, keeps code submitted without the Rapid add-on "indefinitely including after account closure" (codequiry.com, read 9 November 2026); among stated periods, Grammarly Authorship's "up to 10 years" is the longest (support.grammarly.com, read 9 November 2026).
Does Turnitin store student papers?
The AI result is "retained as part of the similarity report" and the institution decides whether a paper enters Turnitin's repository (guides.turnitin.com, read 8 November 2026). Turnitin Clarity treats keystrokes and deletions as Educational Data kept "for the period necessary to enable the continued use of the Services" (Turnitin Services Privacy Policy, read 9 November 2026).
Is a "not stored" promise verified?
No. The index quotes each policy and the date it was read; no vendor's storage was inspected. Where a second document contradicts the promise, as Phrasly's terms do, both are quoted.
Which detectors can I run without sending text anywhere?
GLTR and the GPT-2 Output Detector, both open-source code run on the reader's own machine under the Apache 2.0 and MIT licences (GLTR, GPT-2 Output Detector, read 9 November 2026); the open-source view carries both repositories and SynthID Text. The offline question, added 9 November 2026, widens that answer to fourteen cards, adding DetectGPT, SynthID's text component and the on-premise or self-hosted tiers that Sensity, Hive, Resemble Detect, Sapling and others state.
Sources
All read 8 November 2026 unless a later date is shown.
- AI or Not API documentation: https://docs.aiornot.com/ (read 9 November 2026)
- Scribbr AI detector: https://www.scribbr.com/ai-detector/
- Phrasly AI detector: https://phrasly.ai/ai-detector
- Phrasly terms and privacy: https://phrasly.ai/terms/privacy
- Undetectable AI detector FAQ: https://undetectable.ai/ai-detector (read 9 November 2026)
- Undetectable AI privacy policy: https://undetectable.ai/privacy (read 9 November 2026)
- WasItAI privacy policy: https://wasitai.com/privacy (read 9 November 2026)
- SynthID Detector privacy notice: https://synthid.com/ (read 9 November 2026)
- Sapling privacy policy: https://sapling.ai/privacy
- Hive documentation FAQ: https://docs.thehive.ai/docs/frequently-asked-questions-faq
- Grammarly Authorship support article: https://support.grammarly.com/hc/en-us/articles/29548735595405 (read 9 November 2026)
- GPTZero privacy policy: https://gptzero.me/privacy-policy.html
- Copyleaks privacy policy: https://copyleaks.com/privacy-policy
- Originality.ai privacy policy: https://originality.ai/privacy-policy
- Pangram data privacy: https://www.pangram.com/data-privacy
- Winston AI privacy policy: https://gowinston.ai/privacy-policy/
- ZeroGPT privacy policy: https://www.zerogpt.com/privacy-policy
- Turnitin AI writing detection FAQs: https://guides.turnitin.com/hc/en-us/articles/28477544839821
- Turnitin Services Privacy Policy: https://guides.turnitin.com/hc/en-us/articles/27377195682317-Turnitin-Services-Privacy-Policy (read 9 November 2026)
- Turnitin Clarity administrator FAQs: https://guides.turnitin.com/hc/en-us/articles/37670935742989-FAQs-for-administrators-using-Turnitin-Clarity (read 9 November 2026)
- JustDone privacy policy: https://justdone.com/legal/privacy-policy
- Proofademic privacy policy: https://proofademic.ai/privacy-policy/
- Walter Writes privacy policy: https://walterwrites.ai/privacy-policy/
- AIDetectPlus privacy policy: https://aidetectplus.com/privacy-policy (read 9 November 2026)
- Codequiry privacy policy: https://codequiry.com/privacy-policy (read 9 November 2026)
- Reality Defender privacy policy: https://www.realitydefender.com/t-c/privacy-policy (read 9 November 2026)
- Writer privacy policy: https://writer.com/legal/privacy/ (read 9 November 2026)
- QuillBot AI content detector: https://quillbot.com/ai-content-detector
- Grammarly AI detector: https://www.grammarly.com/ai-detector
- GLTR repository: https://github.com/HendrikStrobelt/detecting-fake-text (read 9 November 2026)
- GPT-2 Output Detector README: https://github.com/openai/gpt-2-output-dataset/tree/master/detector (read 9 November 2026)
Revision log: 2026-11-09: page created from the retention field of 28 cards at index version 2026.11.4; 2026-11-09: ten wave 4 cards added to the table, 38 rows, Sensity's biometric clause quoted as biometric; 2026-11-09: offline question linked (seventh wave); 2026-11-09: eighth-wave pages linked
Related pages: Do AI detectors work on machine-translated text? What two studies measured, with bases, Which AI detectors claim a false positive rate under 1%, and who checked each claim?, AI detectors ranked by stated data retention and Questions about AI detectors, answered from the index.