Hey. That's not human.

heynotai catches AI across text, images, audio, and video — essays, headshots, voice notes, deepfakes, and anything else you can read, write, or upload.

0/5000

Over 57%of the internet is AI-generated — don't be the one deceived.

Install for Chrome

Trusted by worldwide leading companies

VercelLinearNotionStripeFigmaGitHubOpenAIAnthropicReplitVercelLinearNotionStripeFigmaGitHubOpenAIAnthropicReplit
Live preview

One detector. Every kind of AI.

Switch surfaces below — text, image, audio, video, or the extension — and see exactly what the verdict looks like in context.

Spots AI in pixels — and tells you which region

Region localization · 0.4 MP/s
AI
portrait_03.png
Human
street_dx.jpg
Mixed
composite_2.webp
87%
3 images analyzed · 1 AI-generated, 1 human, 1 mixed
Region-level localization on portrait_03.png
1 flagged
How it works

Three steps. No model. No guesswork.

heynotai pipes your input through four independent detectors and reconciles their verdicts. You get one answer, not four conflicting opinions.

Step 01 — Capture

Drop in anything

Paste text, drag images, attach a voice memo, link a YouTube video. Or let the extension watch every page you visit.

drop_anything.* — paste · drag · link
Step 02 — Reconcile

Four models vote

Each surface runs its own ensemble — language model heads for text, vision transformers for images, spectral and prosody analysis for audio.

gpt-detectvit-forensictts-matchfacial-flow
Step 03 — Verdict

One honest answer

Likely human, likely AI, or uncertain — with a probability you can argue with. Never “yes/no”, never an emoji, never made up.

HumanAIMixed
Features

Built like a forensics lab.

Every part of the product is designed to give you a verdict you can stand behind in a meeting, a classroom, or a courtroom.

Browser extension

A colored trust border draws around any YouTube player in real time. Click for the full verdict.

Chrome MV3 · Edge · Brave

Works on any page

Text fields, image previews, embedded videos — heynotai watches the DOM and verifies in place.

1.4M sites compatible

Reports you can defend

Every verdict ships with the model breakdown, the flagged regions, and the confidence interval. Export as PDF.

Audit trail · API access

Privacy-first

Nothing leaves your device unless you ask. The browser extension runs locally; the API is opt-in.

On-device · zero retention

Honest about uncertainty

Three verdicts: likely human, likely AI, uncertain. Never a fake 100%, never a hallucinated reason.

Probabilistic by design

Eight detectors, one API

Text, image, audio, video, deepfake, voice clone, watermark, paraphrase. One auth, one bill.

REST + WebSocket
Customers

What teams say after a week.

We caught seventeen AI-written cover letters in a single morning. Two months ago we wouldn't have caught any. heynotai is now the first thing every reviewer opens.

PM
Priya Mehta
Head of Talent · Veritas HR

I read the regional flags on a student's essay before I read the essay itself. It's not a verdict on the student — it's a starting point for a conversation.

DO
Daniel Okafor
English Faculty · Falcon Edu

The trust border on YouTube has changed how my team reviews source video. We can tell at a glance whether to dig in or move on.

HR
Hana Reyes
Investigations Editor · Atlas Media
FAQ

Questions, honestly answered.

How accurate is heynotai?
Across our internal benchmark of 40 000 mixed samples, the four-model ensemble lands at 96.3% on text, 94.1% on image, 91.7% on audio, and 89.4% on video. We publish the confusion matrix and a probability for every verdict — never a false 100%.
Will it flag my own writing as AI?
Sometimes — particularly for highly-formal academic prose, where humans and language models converge in style. That's why every verdict comes with a probability and the underlying model votes. The product is a starting point for a conversation, not a binary judgment.
Does it work offline?
The browser extension runs on-device for text and image checks. Audio and video analysis use the API — opt-in only, with zero retention by default. You can self-host the API if you need an air-gapped deployment.
What about students and teachers?
We license heynotai for education at a flat per-seat rate, with regional flags rather than scores so teachers can see which paragraphs warrant a follow-up. We do not surface a 'cheating' verdict — that's a conversation for the classroom.
How do you handle false positives?
Every verdict can be appealed in-product; the appeal goes back into our re-training pipeline (with consent). Customers on the team plan get a private retrieval index so their own content style isn't flagged repeatedly.

Catch AI before it catches you.

Add the extension in under a minute. Free for the first 500 detections each month — no card, no model, no guesswork.