The AI detector that doesn't guess.
VERITAS Engine runs multi-layer forensic analysis across images, text, audio and video — cross-checking visual artifacts, embedded metadata and AI signatures through our proprietary detection engine. A transparent verdict you can stand behind.
Free to start · No credit card · Nothing stored, ever
Rewrites stay yours. We never store or train on your uploads.
4
Detection modalities
7+
AI generators identified
0
Files stored, ever
100%
Transparent verdicts
Forensic depth. Honest results.
Three independent analysis layers check against each other. When they agree, you can put your name on it.
Zero-storage, zero-compromise
Every file is analyzed entirely in RAM and discarded the instant your result is returned. No disk writes. No retention. No training on your data — ever. Privacy isn't a setting here, it's the architecture.
Evidence you can act on
Detect AI-written essays, generated images and synthetic voiceovers with a transparent confidence score and a visual forensic breakdown — not just a yes or no.
Publish without the flag
Rewrite AI-assisted drafts into natural prose that bypasses Turnitin, GPTZero and Originality.ai — or cloak AI renders so they pass image detectors.
The VERITAS Vision Engine
Our proprietary detection core cross-references visual artifact patterns, frequency-domain anomalies and embedded provenance metadata — giving you a layered verdict, not a coin flip.
Why VERITAS Engine?
Detection tools are only as good as their honesty. We built ours to be transparent about what it knows — and what it doesn't.
Built for accuracy, not marketing
Every verdict shows a confidence score and the reasoning behind it — no black-box outputs. If the result is inconclusive, we say so.
Continuously updated models
Powered by the AI or Not API, which continuously trains on the latest generators — Sora, Veo, FLUX, GPT-4o images and beyond.
Designed for professionals
Used by journalists, educators, trust-and-safety teams and content researchers who need evidence they can act on — not just a yes or no.
Everything the AI era demands — in one place
Forensic detection across images, text, audio and video. Generator attribution. Deepfake analysis. Humanization. All powered by the VERITAS Engine.
AI image detector
Identify images from Midjourney, DALL·E, Stable Diffusion, FLUX, Grok and Gemini Imagen — with generator attribution, EXIF metadata analysis, and a forensic face-manipulation heatmap for deepfake scans.
AI text detector
Detect ChatGPT, Claude and Gemini output with sentence-level probability scoring. Each passage gets its own confidence score so you know exactly where the AI fingerprint is.
Deepfake & AI voice detection
Analyze MP3 and WAV files for vocoder traces, splice points and synthetic envelope patterns that reveal AI-cloned or generated voices.
Deepfake & AI video detector
Analyze video files for AI-synthesized footage or face-swap deepfakes. Returns separate confidence scores for the video track, voice, and music — powered by the AI or Not v2 engine.
ChatGPT text humanizer
Rewrite AI-generated drafts into natural prose that passes Turnitin, GPTZero and Originality.ai — structurally varied, not just paraphrased.
Bypass AI image detection
Apply imperceptible adversarial noise to AI renders so they pass standard image detectors, while preserving visual quality.
Generator attribution
Don't just get AI or not — know which tool made it. VERITAS identifies the specific AI generator behind an image with a confidence score.
The humanizer suite — coming soon
Two tools in active development that complete the VERITAS picture: detect AI, then — if you need to — make it disappear.
The Ghost
Text Humanizer
Paste robotic AI prose and watch it rewrite live — character by character — into natural, confident writing that slips past every major AI detector.
- Live character-by-character rewrite stream
- Breaks LLM perplexity & rhythm fingerprints
- Bypasses Turnitin, GPTZero & Originality.ai
- Preserves your original meaning & tone
- 150 words free · unlimited on Pro
- Zero storage — in-memory only
The Cloak
Image Humanizer
Drop an AI render and The Cloak layers in imperceptible adversarial noise and organic grain — defeating the artifact patterns that image detectors rely on.
- Adversarial perturbation layer
- Organic film-grain blending
- Defeats frequency-domain classifiers
- Imperceptible to the human eye
- PNG, JPEG & WebP support
- Zero storage — processed in memory
Frequently asked questions
How the VERITAS Vision Engine works, what it detects, and how your data stays private.
How does the VERITAS image detector work?
VERITAS runs a three-layer forensic pipeline on every image. First, the VERITAS Vision Engine — a purpose-built commercial detection model — analyzes the image for AI-generation signatures and assigns a confidence score. Second, we cross-check embedded metadata: camera EXIF data (make, model, focal length, shutter speed) is strong evidence of a real photograph, while AI generator tags confirm a synthetic origin. Third, specialist corroborating models validate the verdict. All three layers are weighted to produce a final score that reflects the full picture, not just one model's opinion.
Which AI image generators can VERITAS detect?
VERITAS currently identifies images from Midjourney, DALL·E (including GPT-4o image generation), Stable Diffusion, FLUX, Adobe Firefly, Grok Aurora and Google Gemini Imagen. When a generator is identified with sufficient confidence it is named explicitly in the result — so you know not just that an image is AI-generated, but exactly which tool produced it.
How accurate is the AI image detector?
In balanced testing across authentic photographs and AI-generated images, VERITAS achieves over 99% accuracy. Unlike single-model detectors that can be fooled by re-saves or style transfers, the metadata cross-check catches cases where visual analysis alone is uncertain. When evidence is genuinely ambiguous, VERITAS returns an 'Inconclusive' verdict rather than forcing a confident wrong answer.
Can I check if an image is AI generated for free?
Yes. Every account starts with 3 free scans per month — no credit card required. A basic scan (1 credit) gives you the AI probability score, generator attribution and a provenance summary. The forensic heatmap (+1 credit) and deepfake face-swap check (+1 credit) are available to any logged-in user.
What is the forensic heatmap?
The forensic heatmap is an interactive X-Ray slider that overlays colour-coded hotspots on your image, marking the regions that most strongly triggered the AI detector. For deepfake scans it shows bounding boxes around manipulated facial areas. Dragging the handle lets you compare the original against the forensic overlay — giving you visual evidence you can share, not just a percentage.
Do you store uploaded images or text?
Never. Every upload is held only in volatile server memory for the seconds it takes to analyze, then immediately discarded. We don't write to disk, don't log your content, and don't train any model on your data. Privacy is the architecture, not an afterthought.
Run your first scan free.
No guessing. No black box. VERITAS shows you exactly what it found and why — in under 3 seconds, with nothing stored.