AI SCAM CHECKER · THE AI DETECTOR FOR BUSINESS
AI Scam Checker: Check the Company and What It Sent You
An AI scam has two halves: a business or a person you cannot verify, and content made to convince you — a video call with a deepfaked boss, a cloned voice note, an AI photo of a product, a flawless message. AnyFact checks both: the Business Check for the website and the company behind it, and our AI detectors for the video, voice, photo or text they sent you.
Free preview · full report on the Business plan ($49 a month) · manual review within 48 hours
Business Check
Open in betaWhat you need
The website address of the business
Optionally add the registration number, the country and the company name.
At a glance
US$25.6M
Sent in 15 transfers by an Arup employee in Hong Kong after a video call with deepfaked colleagues; Arup confirmed in May 2024 that fake voices and images were used.
CNN and ITV, May 2024
0–100
A trust score with a level (high, medium, low, or not enough information) and every check behind it, each with its source.
AnyFact method
4 registries
France (Annuaire des entreprises), the UK (Companies House), US public companies (SEC EDGAR) and the worldwide LEI register (GLEIF).
Public registries
Free preview
Any free account can run a check (3 credits) and see how many checks were run and how many warning signs were found. The full report is on the Business plan ($49 a month).
AnyFact plans
Why an AI scam needs two checks
Generative AI made the convincing part cheap. A deepfake video call can show several fake colleagues at once, a cloned voice can be built from public recordings, and a scam shop can fill hundreds of product pages with AI photos and AI-written copy in an afternoon. Judging the content by eye or by ear no longer settles anything.
What AI does not create is a real company. A registry entry that matches the site, a domain that has existed for years, people named in public records, no regulator warning: the Business Check looks for each of these, with the source of each fact. A scam usually fails there first.
The content itself goes to our detectors: deepfake video frame by frame, voice segment by segment with a transcript checked for urgency and payment requests, photos for content credentials and AI-image signals, text passage by passage. Each result is an indication, not proof, with the reasons and the limits of the method.
Does this sound familiar?
- Your “CEO” asks for an urgent transfer on a video call and the image looks slightly off.
- A voice note from a relative or a manager asks for money, gift cards or a code.
- An investment platform shows a video of a celebrity endorsing it.
- A shop’s product photos look too perfect and the seller cannot send a real one.
- A supplier’s email reads perfectly but the company behind it is hard to find.
Here's what we check
Enter the website
Optionally add the registration number, the country and the company name.
Registries, site, reputation
Up to 12 pages are read; the company is looked up in public registries; domain, Safe Browsing and web reputation are checked.
A trust score with its reasons
A score from 0 to 100 — high, medium or low, or “not enough information” — with every check and its source.
How to read the result
The company was found and matches the site, its operators are identified, and nothing worrying came up. Not a guarantee: keep normal caution.
Some signals are reassuring, others are missing or unclear. The report lists which, so you know what to check yourself.
Several things could not be verified or point the wrong way. A reason to check further before paying or signing, not an accusation.
Too few checks could be run (site unreachable, nothing found). We say so rather than guess a score.
Every report lists the checks that were run, those that could not be, and the source of each fact.
What is an AI scam?
An AI scam is an ordinary fraud (a fake investment, a fake supplier, an urgent transfer) made more convincing with generated content: a cloned voice, a deepfake video, AI photos or AI-written messages. The FBI warned in December 2024 that criminals use generative AI to make fraud more believable at scale, in public service announcement I-120324-PSA.
The goal has not changed: money, codes or documents. That is why the company or the person asking is the first thing to check, before the content.
Which check for which AI scam?
| What happened | Check the business | Check the content |
|---|---|---|
| A video call where an executive asks for a transfer | Call back on a number you already have | A recording through the deepfake video detector |
| A voice note from a relative or a manager asking for money | Call the person back on their known number | The voice note through the AI voice detector |
| An investment platform with a celebrity video | Business Check: registry, regulator warning lists | The video through the deepfake video detector |
| A shop with flawless product photos | Business Check: domain age, company, reputation | A photo through the AI image detector |
| A supplier or recruiter email that reads perfectly | Company check: registry, people | The text through the AI text detector |
AI-written text is not a scam sign on its own: many honest companies write with AI. It matters when the rest does not hold, for example when the company behind the email cannot be found.
What our AI detectors read in scam content
- Video (MP4, MOV, WebM, up to 200 MB and 3 minutes, or a public link): one frame per second through the photo detector, a face-swap check when a frontal face is clearly visible, the soundtrack through the voice detector, and a timeline of suspicious moments. 5 credits.
- Voice (up to 25 MB and 5 minutes): each segment is scored for cloning, and the transcript is checked for scam patterns such as urgency, secrecy, money requests and changed bank details. 2 credits.
- Photo (JPG, PNG, WebP, HEIC, up to 20 MB): content credentials and metadata first, then an AI-image classifier and visible anomalies. 1 credit.
- Text (at least 300 characters, English, French or Arabic): scored passage by passage, so a partly generated message shows up as mixed. 1 credit.
A free account gets 10 credits a month for these detectors, with no credit card. Each verdict has three levels (probably human, uncertain, probably AI) with its reasons, never a bare percentage.
The checks that beat any detector
A detector can be wrong, especially on short, compressed or re-recorded content. These habits work whatever the content looks like.
- Call back on a number you already have. The FTC’s alert on voice cloning in family emergency scams gives this advice: use a phone number you know is theirs.
- Agree on a secret word with your family, and a verification step with your finance team. The FBI recommends the family secret word in its December 2024 announcement.
- Treat a new payment method, new bank details or a request for secrecy as the signal, whatever the voice or the face.
The AI voice and deepfake scams guide walks through each scenario, and the guides on spotting a deepfake video and telling if a voice is AI cover what you can still see and hear.
Before you check
Is the score proof that a business is legitimate, or not?▼
No. It is an indication built from public signals, not a legal or financial assessment. A young or small business can be legitimate, and an established-looking site can still be fraudulent. Use the report to decide what to verify next.
Which countries can you check in an official registry?▼
France, the United Kingdom, US public companies (SEC EDGAR) and any entity with an LEI, worldwide. For other countries, including Germany, Turkey, Saudi Arabia, the UAE, Qatar, Israel and Canada, there is no free registry we can query in this version: we show the registration numbers printed on the site as found, marked not verified.
Do you show who runs the business?▼
Only when a public registry or the company’s own pages (legal notice, about, team) name them, with a link to that source. If our AI cannot confirm anyone, the report says so and our team takes the case manually, with an answer within 48 hours: we never guess a name, and an unidentified operator lowers the score.
What does a low score mean?▼
That several things could not be verified or point the wrong way: no registration confirmed, no legal notice, a very recent domain, a Google Safe Browsing listing, or repeated complaints in search results. It is a reason to check further, not a judgement. For investment platforms, always check the regulator of your own country before sending money.
What do you not check?▼
We do not log in, fill in forms, buy anything or contact the company. We do not read court records, credit ratings or private databases. Regulation is checked against the UK FCA register only; other claims, such as “regulated by CySEC”, are reported as not verifiable in this version.
How is AI used in the report?▼
Claude, Anthropic’s model, reviews the facts we extracted and short excerpts of the pages. Each of its findings must quote the text or the page it is based on; findings without evidence are dropped, and its effect on the score is capped. It also writes the summary, from the checks only.
What does it cost?▼
Any free account can run a Business Check (3 credits, out of 10 free credits a month) and see a preview: how many checks were run and how many warning signs were found. The full report — trust score, company, people, every source and the PDF — is part of the Business plan ($49 a month): unlimited Business Checks and AI checks under a fair-use limit of 5,000 credits a month, and a manual review by our team, with an answer within 48 hours, when the AI cannot confirm everything or when you are not satisfied with a result. When you upgrade, the checks you already ran unlock.
Is there an AI scam checker that checks everything at once?▼
No single test covers a company and its content. The Business Check verifies the website and the company behind it; the voice, video, photo and text detectors check what was sent to you. Use the one that matches what you have, starting with the company.
Can AnyFact check a live video call?▼
No. The detectors work on files and public links, after the fact. During a call, end it and call the person back on a number you already have; if you recorded the call legally, the recording can then go through the deepfake video detector.
Is a probably AI result proof of a scam?▼
No. It says the content shows signs of generation, and honest people use AI too. Combine it with the Business Check and with a call-back before you decide anything.
Check the company first, then the content it sent you
What you need
The website address of the business
Free preview with a free account (3 credits a check). Full report on the Business plan — $49 a month, unlimited Business Checks (fair use).
The score is an indication, not a verdict on the business.
We only read public pages and public registries.
You might also want to check
Is This Voice Real or AI-Cloned?
AI voice detector for voice notes and recordings: each segment is scored for cloned or synthetic speech, and the words are checked for scam patterns. Free test.
Is This Video Real or AI-Generated?
Deepfake video detection and AI video detector: frames, faces and soundtrack are checked separately and suspicious moments are marked on a timeline. First test free.
Website Scam Checker: Check a Site Before You Pay
Website scam checker: paste a link and see the signals scam sites share — regulator warnings, Safe Browsing, domain age, a hidden operator — with sources.