A clear signal while you browse

The problem we solve

Feeds, ads, and documents mix real photos with synthetic media, and human writing with machine text. Most people want a quick, readable answer on whether media are real, not a research project. DotCheck gives you a clear score you can use right away, as you scroll, or when you upload a file.

Three ways to check

Chrome extension. Colored dots on images and video frames, underlines on supported-language text. More green leans human; more red leans AI. Hover for a percent. The rest of the page stays readable.

Check on the web. Score an image or video, then share the result as a short clip built from DotCheck’s real score.

Check. Score media as a guest, or sign in for text, documents, usage, Billing, Pro API keys, Premium history, Share clips, and Pro Confirm when you ask. One account. Same plans everywhere. To cancel a paid plan: Check → Account → Cancel subscription (opens Stripe Billing; access ends immediately).

What makes this different

Most AI checkers pull you out of the feed: upload a file, wait, come back. DotCheck scores media and supported-language text on the page as you browse. Everyday scoring uses the same in-house models on Free, Premium, and Pro. Confirm is a second opinion only when you ask on Pro, never a silent background vendor call.

What we care about

Scores while you browse

Other tools ask you to upload and wait. DotCheck scores images, video frames, and supported-language text on the page as you scroll. When the same content appears again, you can get the score you already have.

Published measurements

We publish how we measure on Docs, with model cards and open weights on Hugging Face under Apache-2.0. Leviathan’s shared memory stays with DotCheck. You can check the numbers; we do not dress the product in unverified accuracy claims.

One stack for everyone

Free, Premium, and Pro use the same in-house models. Paying unlocks History, Shares, documents, and optional Confirm. Paying does not unlock a secretly better detector.

How scoring works

Everyday scoring runs through the Leviathan system on our servers. Named models score images, text, video frames, and audio. When we’ve seen the content before, shared memory can return the score you already have. Free, Premium, and Pro share that path. No plan hides a better model behind a menu.

Vermeer@12

Vermeer scores how AI-like an image looks. You get a clear percent on stills in your feed and on files you upload. Light and surface made a painter’s reputation; here that careful looking becomes a practical signal, without asking you to become an expert.

Valla@9

Valla scores how AI-like text reads in supported languages. Use it as a signal about language, not a verdict on whether the claim is true. A scholar once exposed forged writing by taking the words seriously; Valla keeps that spirit and stops short of fact-checking the claim.

Muybridge@2

Muybridge scores video from a few frames across the clip, so the picture can carry an AI score as you browse. You do not need to watch the whole file first. Early motion studies looked at life frame by frame; Muybridge borrows that idea and stays fast enough for real scrolling.

Helmholtz@2

Helmholtz scores how AI-like audio sounds: speech, music, and general sound. On a video with a soundtrack, it is the ear beside Muybridge, so either channel can catch a synthetic clip. Upload audio on Check or the extension popup when you want a score for sound alone.

Covenant

When a clip has sound, Covenant brings Muybridge and Helmholtz together into one score. Silent clips stay on frames. Same path on every plan.

We publish measured results on Docs. A score is a signal, not a verdict.

DotCheck AI index

The home page shows a live DotCheck AI index: the average AI score across media we newly scored in the last 30 days. Each unique image, video, or text unit counts once when first scored. Cache repeats do not move the index. It is a signal about what people check, not proof that any one item is real or fake.

On Pro, when you already have a DotCheck score and want a second opinion, Confirm is there: up to 300 times a month, only when you ask. Media confirms use Sightengine (France); text confirms use Winston AI (Canada). Free and Premium never send your checks to those services in the background.

Privacy by design

When you check something, we use that media or text to score it, nothing more. We are not building a diary of everything you browse, and we do not collect full page HTML as a product. Limits follow your signed-in account when you are logged in, or a salted IP hash when you are not. Idle Free accounts are not kept forever. Routine checks run on DotCheck’s own models (not a paid detector for every look); heavier work stays on our servers; uploads and Share renders are temporary. Everyday analysis runs in the Netherlands; the product is controlled from the Czech Republic. We sell a service, not user data. Details and your rights: Privacy Policy.

Plans

Free to start. Premium for History, Shares, and documents. Pro when you want Confirm and API keys. See full plan details.

Straight talk

Use DotCheck as a signal, not a verdict. A score estimates how AI-like an image, video frame, audio, or text appears. It is not proof of authenticity, and it does not replace your judgment. Text scoring supports English, Spanish, Portuguese, French, Italian, German, and Dutch. Other languages are not scored yet. We publish measured results as PDFs on our technical docs page, and we do not decorate this site with unverified “99% accurate” claims.