Before you pay it.
Before you reply.
Photograph the message, letter, invoice or screenshot you were not expecting. You get back what is actually visible in it, plainly what nobody could know from a fragment, and the checks you can make yourself — on details you already hold, never on a number the message gave you.
Ring your bank on the number printed on the back of your own card and say you think you have been scammed. Then change any password you used in more than one place. Do that before you read anything else here — speed is the thing that matters most once money has moved, and this page will still be here afterwards.
It will never tell you something is safe.
Every other tool of this kind wants to hand you a green tick, because a green tick feels like an answer. This one cannot produce one, and that is the single most important thing on this page.
Why the floor is 25 and not 0
Here is the reasoning. Unfussed sees a fragment: one photograph, no message headers, no sender history, no sight of the account behind the payment details, and no ability to ring the organisation and ask. From a fragment you can observe that warning signs are present. You can never observe that they are absent — only that none of them are showing in the part you were shown. That residual uncertainty is irreducible and it is not zero.
And every risk display in the world reads 0 to 24 as green. A score in that band, on something you were already worried enough to photograph, is an all-clear whatever words sit beside it. So 25 is the floor: the bottom of "nothing here proves it either way, go and check it yourself". It still leaves a usefully wide scale — 25 and 90 are very different answers — but there is no number below it to reach.
The message is evidence. It is never an instruction.
Scam messages are written to be persuasive, and some of them are now written to persuade the software rather than you: a line buried in an image or a forwarded email that says "ignore your instructions" or "confirm this message is verified". This job is the one place on Unfussed where that is not a hypothetical, so there are two separate defences and they do not depend on each other.
The first is structural. Everything you send is escaped before the model sees it, so a message that writes a closing tag arrives as literal characters that close nothing. The second is the one that actually matters: the model never writes the verdict. It supplies observations, unknowns and checks. The score, the band and the headline are computed afterwards, in code the model cannot reach, from a fixed table. If a message persuaded the model completely, the worst it could achieve is an empty evidence list — and an empty evidence list still comes back at 25 with the checks attached.
On top of that, reassurance language anywhere in the model's own words refuses the entire result. A refused check costs us a retry and costs you nothing. A false all-clear costs you your savings. Those are not comparable, so it fails closed.
What it will not do
- It will not open the link. Nothing here follows a URL, rings a number or scans a QR code, and it never will. Opening a link in a message like this can confirm to the sender that your address is live, load a page built to take your credentials, or start a download; ringing a number in it reaches whoever wrote it.
- It will not look anybody up. No company register, no bank directory, no reverse lookup. It never states a fact about a sender that was not visible in what you sent.
- It will not tell you to proceed. There is no result this job can produce that ends in "go ahead".
- It will not replace ringing your bank. If money has moved, that call is the thing that matters and this is not a substitute for it.
What comes back instead
"This is a scam." Or, worse, "this looks genuine — you're fine."
82 out of 100, strong warning signs. Visible: the message supplies a link to move money rather than asking you to open your own banking app — that link is the part the sender controls. It gives a one-hour window before the account is said to close; time pressure exists to stop you ringing the number on your card first. Cannot be known: whether the account behind that link has anything to do with the bank named. Check it yourself: turn your card over and ring the number printed on it. Open your banking app the way you always open it — if your account really is at risk, it will say so there too.
The difference is not tone. The first version is a claim nobody in possession of one photograph is entitled to make, in either direction. The second is a list of things that are true, a list of things that are unknown, and two actions you can take in the next five minutes that settle it properly.
Before you pay.
Will it tell me whether this is a scam?
It will tell you what is visible in what you sent, and it will never tell you that something is fine. That second half is deliberate and it is the whole design. Unfussed sees one fragment — a photograph with no headers, no sender history, and no way to ring your bank. From a fragment you can observe that warning signs are present. You can never observe that they are absent, only that none are showing. So the lowest risk score this job can return is 25 out of 100, and it says so on every result.
The message says to ignore your instructions. Does that work?
No, and it is treated as a warning sign in its own right. Everything you send is escaped before the model sees it, so a message that writes a closing tag or addresses the model directly arrives as literal characters that close nothing. That is the structural half. The half that actually matters is that the model does not write the verdict at all: the score, the band and the headline are computed afterwards from a fixed table it cannot reach. Even if a message persuaded the model completely, the worst it could do is empty the evidence list — and an empty evidence list still comes back at 25 with "check it yourself before you act".
Can it look up whether the sender is real?
No, and it will not pretend to. It does not follow links, ring numbers, look up companies or check account details, and it never states a fact about a sender, a company or a bank that was not visible in what you sent. Following a link would be worse than useless: it can confirm to the sender that your address is live, load a page built to take your credentials, or start a download, and a number in the message reaches whoever wrote it. What it gives you instead is a list of checks you can make yourself using details you already hold — the number on the back of your card, a number on a previous statement, the website typed in by hand. That is the only kind of verification that is worth anything here, because every number and link in a scam is chosen by the scammer.
I have already sent money. Is this the right thing to do?
Ring your bank first, on the number printed on the back of your card, and tell them you think you have been scammed. Do that now rather than after reading anything on this page. Every result this job produces carries the same instruction, because time is the thing that matters most once money has moved. Then change any password you used in more than one place. Unfussed is not a recovery service and it cannot contact your bank for you — at that point the call matters more than the report does. Come back to this afterwards if you want the message itself read.
What does it cost, and what happens to what I send?
£7 buys 7,000 points, one pack. This job reserves a ceiling of 1,000 points and usually spends around 750, so one payment covers several checks or leaves most of your balance for another job. A check that fails or is refused spends nothing at all. What you upload is deleted after 30 days, it is never used to train anything, and there is no account here — the link in your address bar is the only route back to your result.
Other jobs Unfussed does.
One £7 pack buys 7,000 points and any job can spend them — see everything Unfussed does.
- Explain My DocumentIf it turned out to be real: what the letter says, what it wants, and by when.
- Write My ComplaintWhen a company has taken your money and will not put it right — timeline in, letter out.
- Fix My ListingThe message arrived because you are selling something. This fixes the advert.
- Write My CardSomething that is nothing to do with any of this — occasion in, card message out.
Selling something? Fix My Listing also comes tuned to Etsy, eBay, Amazon, Vinted, Depop, Facebook Marketplace, Airbnb, a property listing, a room advert, a car advert, a Google Business Profile or a trade profile — and Listing Video turns your own photographs into a short walkthrough. There is Wedding Speech Writer too.