Writefull GPT Detector

The detector was never sold. The archived FAQ called it a "Completely free ChatGPT detector" with the unquantified daily quota above, so the free tier field records a free tool with a per-day period and no figure, and the free words per month and free checks per day rank fields stay empty. Writefull's live subscription page renders its prices by script and the server HTML prints only "Loading Pricing..." (my.writefull.com/plans, read 10 November 2026); the plans belong to the Word and Overleaf writing products, so the card carries no plan table.

On retention, the archived detector page made a direct statement under "Safe & secure": "Your data is encrypted and not stored" (archived detector page, 31 December 2024, read 10 November 2026). The live Privacy Notice, last updated May 13, 2025, states: "We only keep your personal data for as long as it is necessary for the purposes for which it was collected, after which it will be destroyed, erased or anonymised", and it names no period for submitted text (privacy notice, read 10 November 2026). The card does not rank the archived statement: the tool is discontinued and the live notice governs, the reading the index gave Writer's archived page, so the retention field is empty and both statements are quoted here and in the second table of the data retention ranking (ruling of 10 November 2026). Ownership comes from the same notice, which says "Writefull is operated by ThinqLab Ltd, with registered offices at 6 Briset Street, London, EC1M 5NR, UK", and from the Terms of Use effective May 13, 2025, which give the notice address as "Writefull c/o Digital Science", name Digital Science UK Ltd as an affiliate and apply the laws of England (terms of use, read 10 November 2026).

Writefull published no accuracy percentage. The archived FAQ answered the threshold question directly: "The higher the percentage, the more likely, but there is no absolute threshold. The more creative the language used in a GPT-plagiarized text, the more difficult it is to recognize the 'voice' of GPT-3, GPT-4, or ChatGPT in it. And a GPT-text's language creativity depends on many things, like the parameters set and the prompt entered" (archived detector page, 31 December 2024, read 10 November 2026). No false positive rate accompanied it, so the claim gap field is empty and the figures on this card come from the studies below.

The discontinuation notice is the vendor's fullest caveat. The live route reads: "AI-generated text is now almost indistinguishable from human writing. New models produce more varied and complex text, with better fine-tuning through prompts. This makes it harder to tell whether a text was written by AI or a human. Research by OpenAI and other third parties, as well as our own analysis, has shown that AI detection tools have high error rates. They often misidentify human-written text as AI-generated and vice versa", and then: "students who speak English as a second language or use a more formulaic or concise writing style are more likely to be incorrectly flagged as using AI. All AI detectors face the same challenge. As AI improves, detection methods become less accurate, making them unreliable for assessing academic integrity. Because of this, we've decided to take our GPT Detector offline. Academic integrity is important to us, and we don't want to offer a tool that no longer works as intended" (x.writefull.com/gpt-detector, read 10 November 2026). The card carries both passages as the vendor caveat.