TruthScan AI detector

The text API documentation sets the operating limits: at least 200 words are "recommended for best accuracy", the request input must be "less than 30,000 words", and text is "billed at 1 credit per word" (API documentation, read 9 November 2026). The endpoint returns a score from 1 to 100 where "any score under 50 is considered definitely human. 50-60 is possible AI. Over 60 is definite AI", a label of Human, AI or Paraphrase, and approximate scores attributed to GPTZero, OpenAI, Writer, CrossPlag, CopyLeaks, Sapling, ContentAtScale and ZeroGPT, which the documentation calls "approximate and not as accurate as the main result score". Images take JPG, PNG, JPEG, TIF and WEBP up to 10 MB per file, and "one result = one image or one page of a PDF" (truthscan.com/pricing, read 9 November 2026).

The pricing page prices results, not words. Figures below are as displayed on 9 November 2026 with the monthly toggle, annual prices as the page shows them; the Free plan is "free forever, no trial period", and every plan includes API access and the Chrome extension with unlimited seats (truthscan.com/pricing, read 9 November 2026). No price per text word is published; the API documentation's 1 credit per word is the only text rate.

One figure on the page disagrees with the table: the pricing FAQ says "40,000 results on Business is $399 total", while the Business card shows $333 a month (truthscan.com/pricing, read 9 November 2026). The card carries the plan card's figure and quotes the FAQ's beside it.

Retention is stated in days. The privacy policy reads: "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" (truthscan.com/privacy, last updated September 19, 2025, read 9 November 2026). The pricing FAQ confirms that "images are retained by default" and that Zero Data Retention "is available on Business and above", under which "submissions and customers' data are immediately discarded once detection is complete and never used for training"; the terms add that the toggle "currently applies to image and PDF submissions", with video or audio on request (truthscan.com/terms, last updated August 27, 2026, read 9 November 2026). The same terms grant TruthScan a licence over User Content "solely as necessary to provide, operate, secure, and support the Services" and state that "User Content and your use of the Services is not your confidential information". The card's retention rank field takes the 30-day default.

The detector page's stat strip reads "99%+ Detection Accuracy", and the FAQ below it narrows the sentence: "TruthScan's AI text detector achieves 99%+ accuracy (for custom tuned enterprise models) in identifying AI-generated content. Our detection models are trained on millions of text samples and are regularly validated against real-world datasets" (truthscan.com/ai-detector, read 9 November 2026). No sample size, test date or dataset accompanies the text figure. For images the pricing page states an accuracy rate of 99.3%, described as "average detection accuracy across 92 image generators and 250,000 real images", with per-category figures from 98.1% for invoices and cards to 99.9% for chats, a "False positive rate <1%" and "95%+" on every generator tested (truthscan.com/pricing, read 9 November 2026). The image false positive figure has no base and describes images, so the card's text-only false positive field stays empty and the figure travels here.

On paraphrase the page claims coverage: "Detect AI-generated content even after it's been paraphrased, rewritten, or humanized. Our sophisticated algorithms analyze underlying linguistic patterns and statistical markers to identify AI origin, regardless of surface-level modifications" (truthscan.com/ai-detector, read 9 November 2026), and the API's Paraphrase label is the product form of that claim. The vendor's caveats are operational: the 200-word recommendation and the third-party scores being "approximate" in the API documentation, and in the image FAQ, "Heavy degradation still costs you. A screenshot of a screenshot, or a thumbnail recompressed to a fraction of its original size, discards pixel-level signal that no detector can recover" (truthscan.com/pricing, read 9 November 2026). The pricing FAQ also says "You set the thresholds, so most teams route borderline scores to review rather than auto-deny".