Open in the beta
AI Photo Detector
Find out if an image was generated by AI, and why.
How it works1 credit per analysis
Before you trust, verify.
Real or AI? The free AI detector that explains why.
Check a photo, a video, a voice message or a text — and documents, code and websites too. We read its provenance, run open-source detectors and explain the verdict in plain words, so you can decide before you share, pay or reply.
- First test free, no account
- A verdict explained in plain words
- Files deleted after 30 days
Drop a photo or a screenshot
JPG, PNG, WebP or HEIC · up to 20 MB
No account needed to start. Your result unlocks when you create a free account.
Fakes now reach you where you trust the most.
Pick a situation to see what we check — then test your own file for free.
AI-generated
M
Miranda
online
0:23
“Hello, I’m Miranda. Is this really my voice or an AI-generated clone? Even my parents can’t tell the difference.”
Synthetic voice made with ElevenLabs.
What AnyFact checks
- Each few seconds of audio scored for synthetic speech
- The message transcribed and checked for scam patterns: urgency, money, secrecy
- A clear answer before you send anything
Real sample files with a known origin: the voice and the two photos are AI-generated, the video is real. The bank call and the cover letter are illustrative examples.
Not a mysterious percentage. A verdict with its reasons.
Step 1
Provenance first
Content credentials (C2PA) and metadata often say how a file was made. When they name an AI generator, that settles it.
Step 2
Open-source detectors
Otherwise, models built for each format score the content — pixel by pixel, second by second, passage by passage.
Step 3
A verdict you can explain
Probably human, uncertain or probably AI — with the reasons, the limits of the method and a PDF report. Claude, a language model, writes the explanation and only weighs in on borderline cases.
AI Photo Detector · sample report
Probably humanHuman18 / 100AI
No AI generator is named in the file, the detector scores it well within the human range, and the light, skin texture and background blur are consistent with a camera photo. As always, this is an indication, not proof.
- • No content credentials or AI generator in the metadata
- • Classifier score in the human range
- • Light direction and shadows are consistent
PDF report This result is wrong
One place for every kind of content.
All 7 detectors are open in beta, on every plan.
Open in the beta
AI Text Detector
See which passages of a text were probably written by AI.
How it works1 credit per analysis
Open in the beta
AI Voice Detector
Check whether a voice message was cloned or synthesised.
How it works2 credits per analysis
Open in the beta
AI Video & Deepfake Detector
Spot AI-generated or face-swapped videos, moment by moment.
How it works5 credits per analysis
Open in the beta
AI Document Detector
Analyse the text, images and metadata of a PDF, Word or PowerPoint file.
How it works3 credits per analysis
Open in the beta
AI Code Detector
Estimate whether source code was written by an AI assistant.
How it works1 credit per analysis
Open in the beta
AI Website Checker
Measure how much of a website’s text and images is AI-generated.
How it works10 credits per analysis
Open in the beta
Business Check
Is this business legit? Is this SaaS reliable? A trust score from public signals, each with its source.
How it works3 credits per analysis
Built to be trusted, including when it is unsure.
No detector is perfect. So we show our work, say when we do not know, and let you flag a wrong result.
What you usually get
What AnyFact gives you
A single percentage
Three clear levels: probably human, uncertain, probably AI
No reason given
The signals behind the verdict, explained in plain words
Only the pixels are looked at
Content credentials and metadata read first
Presented as certain
The limits of the method shown on every result
Nothing to keep
A dated PDF report with a unique reference
Unclear what happens to your files
Files deleted after 30 days
Methods and models we rely on:C2PAEXIF / XMPCLIP classifiersFast-DetectGPTQwen2.5wav2vec2 anti-spoofingWhisperClaude
One AI detector, every kind of content
Pick the AI detector that fits what you need to check
Every detector reads provenance first, then runs an open-source model and explains the verdict: probably human, uncertain or probably AI.
AI image detectorWas this photo generated? Content credentials, then a detector.AI text detectorPassage-by-passage check of any text of 300+ characters.ChatGPT detectorText from ChatGPT, Claude, Gemini or another assistant.AI essay checkerFor students checking their work and teachers reviewing it.Deepfake detectorFace swaps, generated clips, fake photos and cloned voices.AI video detectorFrame by frame, faces and soundtrack, on a timeline.AI voice detectorCloned or synthetic voice notes, plus scam patterns.AI art detectorIllustrations, digital paintings and contest entries.AI document detectorPDF, Word and PowerPoint, page by page.AI code detectorFunction by function, with an honest reliability warning.AI website checkerHow much of a site is AI-generated, and where.C2PA checkerRead an image’s content credentials and metadata.Is this website legit?Registries, legal notice, domain age and reputation, each with its source.Website scam checkerSafe Browsing, regulator warnings, domain age and a hidden operator.Is this company legit?The company behind a site or an offer, with the people named and sources.AI scam checkerThe company, then the deepfake, cloned voice or AI content it sent you.SaaS reliability checkerCustomers, docs, security claims and the company behind a SaaS.
Pricing
Start free.
10 credits every month.
Free
$0/month
10 credits a month
Refilled on the 1st of every month
- About 10 photo or text checks
- Business Check preview: trust score and warning signs
- No credit card required
Pro
$15/month
1,000 credits a month
Refilled on each monthly renewal
- About 1,000 photo or text checks
- Business Check preview: trust score and warning signs
- Cancel anytime from your account
Business
$49/month
Unlimited Business Checks and AI checks
Fair use: up to 5,000 credits a month
- Full Business Check reports: trust score, company, people and sources
- Manual review by our team within 48 hours when the AI cannot confirm everything
- About 1,666 Business Checks or 5,000 photo or text checks a month
- Cancel anytime from your account
Credits per analysis
Every plan includes every AI detector, PDF reports and a Business Check preview. The full Business Check report is part of the Business plan.
FAQ
Common questions
Is the result proof?
No. A detection result is an indication, not proof. Detectors make mistakes, and a well-edited AI content can look human (and the reverse). Use the verdict and its explanation as one input in your own judgment, never as the only basis for a sanction or an accusation.
How does it work?
Three steps. First we read the provenance: C2PA content credentials and metadata, which can settle the question on their own. Otherwise an open-source detection model scores the content. In the grey zone, Claude, a large language model, gives a second opinion, and it always writes the plain-language explanation. The verdict has three levels: probably human, uncertain or probably AI (and “mixed” for texts).
Which files can I analyse?
During the beta: images (JPG, PNG, WebP or HEIC, up to 20 MB), videos (MP4, MOV or WebM, up to 3 minutes), voice messages and recordings (MP3, WAV, M4A or WhatsApp notes, up to 5 minutes) and texts of at least 300 characters. Documents (PDF, Word or PowerPoint, up to 20 MB and 50 pages), source code (at least 30 lines) and websites (up to 20 pages) are open too.
What happens to my files?
Your files are used only to run your analysis and are deleted after 30 days. Your results stay in your account history, and you can download a PDF report of each one.
How much does it cost?
There is a free plan: you get 10 free credits every month, with no card. Pro costs $15 a month for 1,000 credits, and Business $49 a month for unlimited Business Checks and AI checks (fair use). Each analysis costs a number of credits that reflects its compute cost — 1 for a photo and 1 for a text.
Can it be wrong?
Yes. No AI detector is perfect, especially on short texts, heavily compressed images or edited content. Every result page has a “this result is wrong” button: your feedback helps us measure and improve the detectors.
What is an AI detector?
An AI detector estimates whether a piece of content was made by a generative AI tool. AnyFact covers seven kinds of content in one place: images, videos, voices, texts, documents, source code and websites. It reads provenance data first, then runs an open-source detector and explains the verdict.
Is AnyFact a free AI detector?
Yes, there is a free plan. Your first photo, video, voice or text test needs no account, and a free account gets 10 credits every month, with no credit card. Pro ($15 a month) gives 1,000 credits a month; Business ($49 a month) gives unlimited Business Checks and AI checks under a fair-use limit.
Can it detect ChatGPT, Claude or Gemini text?
The text detector does not look for one brand: it measures how predictable each passage is to an open-source language model, which works on text from ChatGPT, Claude, Gemini or other assistants. The result is given passage by passage, so a partly generated text shows up as mixed.
Can it detect deepfake videos and cloned voices?
Yes. The video detector scores one frame per second, checks clearly visible faces for face swaps and sends the soundtrack to the voice detector. The voice detector scores each segment of a recording and flags scam patterns in the transcript. Compressed clips are harder to judge, and the verdict says so.
Used AnyFact? Tell others what you think.
Good or bad, your honest review on Trustpilot helps people decide, and helps us improve.
A doubt about a file? Check it now.
Your first test needs no account. Then 10 free credits every month with a free account — no card.