TextSight.ai
Detect AI content sentence by sentence, with an authenticity score
About TextSight.ai
TextSight.ai does AI detection at sentence granularity rather than returning a single verdict on a document, which is a meaningful difference. Most real text is mixed: a human draft with AI-expanded sections, or an AI draft a person has partly rewritten. A document-level percentage tells you almost nothing about that, whereas a per-sentence breakdown shows you where the machine writing actually is. The product also states that its v4 model is calibrated for low false positives, which is the failure mode that matters most, since wrongly flagging human writing does real harm.
Coverage spans the major models explicitly, naming ChatGPT, Claude, Gemini and Llama 3, and the tool reports more than a million texts analysed across 50 or more languages. Output combines a 0 to 100 authenticity score with the per-sentence view. A one-click rewrite of the flagged sentences is included, and it is worth being straightforward that this is the same detection-evasion capability that dedicated humanisers offer, whatever the framing. Twenty free AI writing tools are published alongside the detector. Input accepts pasted text, a URL to scan, or an uploaded file, up to 5,000 characters per check, and it is free to try with no card and no signup.
Paste text, supply a URL to scan, or upload a file, up to 5,000 characters per check. Sample texts are provided covering AI, mixed and human writing, which is a good way to calibrate your own expectations of the tool before trusting it on real material. Run the check and the result gives an overall AI probability and authenticity score from 0 to 100, alongside a sentence-by-sentence breakdown showing which specific passages read as machine-written. From there, the one-click rewrite reworks the flagged sentences. The whole detection flow is free with no card and no account.
- •Sentence-Level Detection - Shows which specific passages read as AI rather than returning one document-level number
- •0-100 Authenticity Score - A single headline figure alongside the detailed breakdown
- •Calibrated for Low False Positives - The v4 model explicitly tuned against wrongly flagging human writing
- •Named Model Coverage - ChatGPT, Claude, Gemini and Llama 3 identified specifically
- •50+ Languages - Detection well beyond English
- •Three Input Modes - Paste text, scan a URL or upload a file
- •Calibration Samples - Provided AI, mixed and human samples for judging the tool before trusting it
- •One-Click Rewrite - Reworks flagged sentences, which is a detection-evasion capability however it is framed
For editors and content managers who need to know whether submitted work was AI-written, and where the sentence-level view matters because mixed authorship is the normal case rather than the exception. It suits educators reviewing student work, with the important caveat that no detector is reliable enough to be the sole basis for an academic misconduct finding, which is why the low false positive calibration is the specification worth caring about. Writers can check their own work before submitting. The rewrite feature serves people wanting to reduce AI tells, whether for genuine style reasons or for evasion.
Pricing
$0 - $79.99/mo
- Free$0/mo
- Starter$9.99/mo
- Pro$19.99/mo
- Business$39.99/mo
- Enterprise$79.99/mo
From the vendor pricing page, 2026-09-16
