FAQ

No. Texts are analysed on your computer. The app connects to the internet only to download model files, to check for updates and, if you enter a key, to check that key (the key and a random device label are sent, nothing else).

No detector is perfect. On held-out human texts the false-flag rate is about 0–2% for general writing, and 5.5% for TOEFL exam essays in sensitive mode (about 1% in precise mode, measured on version 1.3). Writers who are not native English speakers are affected more. Never use a result as the sole basis for decisions about people.

Version 1.3.1 uses new models merged from three trained variants. On a public humanizer test it flags 81% of rewritten texts (version 1.1: 70%, version 1.0: 31%). Humanizers change every month, so the models are updated regularly and the app downloads the updates.

English is the main language, but Russian and Polish now run on the same models. On our held-out test sets (350 AI and 350 human texts per language) the verdict “AI” catches about 80% (Polish) and 78% (Russian) of AI texts at 0.9% and 0.6% false alarms (version 1.1: 39% and 57%). Texts from newer models and rewritten texts are caught less often (Gemini-rewritten Polish: 48%, Russian: 50%). The sets are small, expect ±5 points. We have not measured Russian or Polish against other detectors, so we make no comparison claim. Other languages are not supported. Exact figures are in the results section.

Windows 10/11, 64-bit. Loupe: CPU, 4 GB RAM minimum (8 GB preferred), about 0.3 GB models and 2 GB free disk. Assay: 8 GB RAM minimum (16 GB preferred), about 2.9 GB models and 6 GB free disk; DirectX 12 GPU with 8 GB VRAM recommended for FP32, 4 GB unvalidated. CPU fallback is slower. Timings depend on text length, hardware and markup; old FP16 timings are not current FP32 guarantees. See Which version? and Trust.

Yes — for personal non-commercial use: your own texts, study, hobby, non-commercial research. No registration, no key, no limits. A paid key is only the permission for commercial use (paid work, clients, companies, schools and universities, products): it hides the licence reminder; the app itself works the same without it. See pricing.

The technical report lists data, tests, limitations and caveats. The code is on GitHub.

Open the Licence window and press Remove key — this frees the slot — then enter the key on the new computer. Copying the app or its licence file to another computer does not carry the licence over. If the old computer is gone, write to support and the slot is freed.

Payments are handled by Polar (Merchant of Record). Personal and Team keys: if Linda does not work for you, write to support within 14 days of purchase and the licence is refunded in full through Polar, no questions asked. The app is free for personal use, so you can try everything before you buy.

Yes. Batch check takes any number of PDF, DOCX, TXT or MD files one after another, with CSV export for your records. Document comparison shows two versions side by side (shared passages of 8 or more identical words in a row, and how the highlighting changed) or one summary table of overlaps across many documents. Everything still runs on your PC; nothing is uploaded. Checking single sentences is about 70–75% accurate — weaker than checking the whole text.

After the first start (model download) the analysis runs offline, and the free version needs no key at all. A paid key is activated once (online) and then works fully offline too: it is the permission for commercial use and hides the licence reminder — the app works the same without it. History of checks is kept only on your computer and can be switched off.

If a school or university buys a licence for centralised checking on the computers of the organisation, the Organization licence is required (one site, 200 devices, yearly, invoice with VAT details). If you are a teacher or tutor checking papers personally on your own computer, the free version is enough (or inexpensive Personal if you need official removal of reminders for work). Never use a result as the sole basis for decisions about people. Free commercial evaluation lasts 14 calendar days from first launch, without a paid key, registration or automatic billing. This is a licence condition; the app does not enforce a timer. Continued commercial use then needs a paid licence. Personal non-commercial use stays free without a time limit.

It is the confidence in the verdict shown next to it (“AI”, “human” or “uncertain”), not the share of the text written by AI: 88% next to “AI” means the models are fairly sure the text is machine-written. The share of the text that looks AI-written is shown separately (the colour map, “AI share” and the authorship bar). A high percentage is a reason to look closer, not proof; see “If the score is high”. Very short texts (under about 150 words) give less reliable results, and the app warns about it.

A lighter and faster set of models with the same app. It works on any computer without a graphics card and checks a long text in seconds. It is somewhat less accurate than Linda Assay, mostly on texts rewritten by “humanizers”. See the comparison.

A simple or old computer: Linda Loupe. A gaming computer or a recent graphics card: Linda Assay. Not sure: Linda Loupe, you can add Linda Assay later.

Yes to all. The installer asks what to put on the computer; later, in Settings, you can tick or untick either version. The app works with only Linda Loupe, only Linda Assay, or both (it then picks the suitable one and lets you switch).