AI, deepfake and voice cloning scams
AI has not created new scams. It has removed the mistakes that used to give the old ones away, and it can now copy a familiar voice from a few seconds of audio.
What it is
This covers fraud where AI generates the deception itself: a cloned voice on a phone call, a synthetic face on a video call, a fabricated endorsement from a public figure, or a persona whose photographs belong to a person who does not exist.
The important framing is that nothing here is a new category of scam. A voice clone of your child in trouble is the grandparent scam with better production. A deepfake investment endorsement is investment fraud with better production. What has changed is that the signals people were taught to rely on no longer work.
Poor grammar, obvious artificiality, an unwillingness to speak on the phone, a refusal to video call: all of these were reasonable tests, and all of them have stopped being reliable.
How it starts
A phone call in a familiar voice. A relative, distressed, saying they have been in an accident, arrested, or are in trouble abroad, and need money now. Cloning a convincing voice needs only a few seconds of audio, and social media supplies it.
A video call with a face you recognise, in business fraud, where an instruction to transfer funds comes from someone who appears on screen to be an executive.
An advertisement featuring a public figure endorsing an investment or a product, generated from real footage of them. They have never heard of it.
A persona built for a relationship, where the photographs are AI-generated, which is why reverse image search now returns nothing rather than returning someone else.
A brief, low-quality video call in a romance or investment scam, just long enough to answer the objection that you have never seen them.
Why it works
A familiar voice bypasses evaluation entirely. You do not assess whether a call is from your daughter. Recognition happens before judgement, and it happens instantly.
Distress compresses the decision. The scenario always involves someone frightened and out of time. Verifying feels like a delay you cannot afford, which is the design.
The old tests now pass. Grammar is perfect. The video call happens. The photograph is unique because it was generated, not stolen. Every heuristic people were taught has been quietly inverted.
Public figures carry unearned trust. A recognisable face saying something plausible is persuasive in a way an unknown one is not, and consent has nothing to do with it.
Detection is not a consumer skill. Telling a synthetic voice from a real one is difficult for trained listeners and near-impossible on a phone line under stress. Advice built on spotting artefacts is advice that will fail.
How to spot it
Because the artefacts are unreliable, this section is about behaviour rather than detection.
An urgent call from a relative in trouble, where the caller discourages you from contacting anyone else
Any request for money by gift card, wire, crypto or cash courier following an emotional call
A number you do not recognise, or a claim that they are calling from someone else's phone
A public figure endorsing an investment, appearing only in an ad and nowhere on their own channels
A video call that is unusually brief, dark, low-quality, or where the person will not turn their head or move naturally
A profile whose photographs return no reverse image search results at all
Any high-value instruction arriving only by voice or video, with no independent confirmation
What to do
Agree a family phrase, and do it before you need it. A word or question only your family knows, asked on any distressed call. This is the single most effective countermeasure available, it costs nothing, and it defeats voice cloning completely.
Hang up and call back on the number you already have. Not the number that called you. If the story is real, the call back reaches them. This works regardless of how convincing the voice was.
Verify high-value instructions on a second channel, always, in business as much as at home. A video call is no longer confirmation of anything.
Reduce what is available to clone. Public video and audio of your voice, and your children's, is the raw material. This is not a reason for panic, but private accounts materially reduce the supply.
If you have already paid. Contact your bank immediately. Report to the FBI's IC3 at ic3.gov, and note that AI was involved, because that reporting is how the scale of it becomes visible.
Expect the tests to keep changing. The reliable defence is procedural rather than perceptual: verify on a channel you initiated, using contact details you already had. That holds no matter how good the generation gets.
Have this emailed to you
Send yourself this page so you can forward it, or read it later when you are not standing in a shop.
Email it to myselfPaste a text, email or direct message
A text, an email or a chat message. Take out names and account numbers first.
Add a screenshot
Check it
No account needed.
What happens to your message
Checking the message now. Stay on this page.
Flggd could not check this message just now. Try again in a moment, or use the Flggd app.
Try again
Flggd could not read this image, so it has not been checked. This is not a verdict. Try a clearer screenshot, or paste the text instead.
Paste a message or add a screenshot first.
Text Submission
Image Submission
Flggd Verdict
Scam
Red flag
High confidence of a scam. The content, the sender, the intent and where its requests lead all point the same way. Do not reply, click, call or pay.
Likely to be a scam
Yellow flag
It asks you to act: a link, a number, a reply or a payment. Where that request leads is where the harm starts.
Looks normal
Green flag
High confidence this is normal. The content reads as expected, the sender has a safe history, and anything it asks leads somewhere legitimate. Accounts can be taken over, so if a request seems out of character, check again.
Likely to be a scam
Grey flag
Nothing is asked of you yet. The content, the sender or the intent do not add up. Scams like this start friendly and ask later.
This result could not be shown. Check the message again. If it happens twice, use the Flggd app instead.
Score
Scam type
Confidence
Why Flggd says this
What this message wants you to do
Has anything happened yet?
You do not have to answer. Nothing you tap here is sent or recorded, and the verdict above stands either way.
Change answer
Nothing has happened yet
Good. Checking first is the safest thing to do. Do not tap any link, reply, or call any number in the message. Block the sender and delete the message. If it claims to be from an organisation you know, contact them using details you already trust.
Change answer
You tapped a link
Opening a link does not usually take money or details on its own. The harm comes from what you do on the page it opens. If you entered nothing and installed nothing, close the page and do not go back to it. If it asked you to sign in, enter a code, or install something, and you did, choose Gave details instead.
Change answer
You gave details
Work through this in order. First, if you gave bank or card details, call your bank now on the number on your card. Next, change the password for any account you entered, and for any other account that uses the same password. If you shared a one-time code, tell the organisation it belongs to straight away. Then turn on two-step verification where you can, and watch your statements for anything you do not recognise.
Change answer
You sent money
Call your bank now, on the number on your card or in your banking app. Say: "I have been scammed and I need to stop a payment." The sooner they hear from you, the more they may be able to stop or recover. In the UK you can call 159 to reach your bank. Do not pay anyone who offers to get the money back. That is another scam.
Report it. US: ReportFraud.ftc.gov, and ic3.gov for online crime. UK: reportfraud.police.uk or 0300 123 2040; in Scotland, Police Scotland on 101. Australia: scamwatch.gov.au. Have ready the message, the sender's number or address, and any payment details.
What to do next
Do not reply, click any link, call any number in the message or send money. If it claims to be from an organisation you know, contact them using details you already trust.
Do not reply, click any link, call any number in the message or send money. If it claims to be from an organisation you know, contact them using details you already trust.
No action needed. If the sender later asks for money, codes or personal details, check that message too.
Do not reply at all, not even to say wrong number. A reply tells them the number is live. Block the sender and delete the message. If a request follows, check it again.
Text read from your image
Your image
The closest scam in the Flggd library. Similar, not necessarily the same message.
How this scam works
The closest match is the scam described on this page.
Share this result
Email it to myself
Copied. Paste it anywhere to share it.
Could not copy. Select the text and copy it, or use Email it to myself.
Checked with Flggd at flggd.com
My Flggd verdict
Check the next one on your phone
The Flggd app checks texts, emails and screenshots wherever you are, and teaches you the warning signs as you go. Free on iPhone and Android.
Get the app
What Flggd does about this
Worried about someone else rather than yourself?
Helping someone else
