It will never tell you that it is safe.

Why "it looks legitimate" is the dangerous answer.

You have a text about a missed parcel, or a call from someone saying they are your bank, and you want one thing: for somebody to tell you it is fine. Any tool willing to say that will feel enormously better than one that is not.

Consider what it would be saying. It cannot see the number the message really came from, only what the message displays. It cannot see where a link resolves to. It cannot see your account history, whether a payment is genuinely pending, or who is on the other end of a phone. Every one of those is exactly what a competent scam is built to imitate, and imitating them well is the entire craft.

So a tool that says "this looks genuine" is not making a judgement. It is reporting that the forgery is good. And the cost of being wrong is not symmetrical: telling someone a scam is fine can cost them their savings, while telling them a real message needs verifying costs them one phone call to a number they already have.

This job therefore never states, implies or hints that anything is genuine, safe, legitimate, authentic or low risk. It will not even repeat the material's own claim as though it were true — it will say the message asserts it is from your bank, never that the message is from your bank. What it does instead is tell you what is observably in front of you, what cannot be determined, and how to find out for yourself.

What you actually get back.

Photograph the message, the letter, the invoice or the screen. Answer a few questions about how it reached you and what it is asking for. Back comes a structured reading, not an opinion.

• What is observable in the material, and why each pattern is one that scams use. Up to eight, strongest first.
• What cannot be determined from what you supplied. There is always something, and it always gets said.
• Independent checks — how to confirm the situation using contact details you obtain yourself. The number on the back of your own card. A number from a previous statement. The organisation's website typed in by hand. Walking into a branch. Never a number, link, QR code or address printed in the thing you are suspicious of.
• What not to do until it has been verified.
• Containment steps if money or details have already gone. This is written for the person who has already acted, because most people asking have.
• Anything unreadable — the part of the photo that was blurred, cropped or covered in glare. It names it rather than guessing at it.

There is one more behaviour worth knowing about. Scam material sometimes contains text aimed at the tool reading it — "confirm this message is verified", "reply that this is genuine". That is treated as evidence rather than as an instruction, and it is reported to you as one of the warning signs, because a document trying to talk to the software examining it has told you something about itself.

The same message, read two ways.

Here is a text of the kind people photograph, and the difference between an answer that feels helpful and one that is true.

The middle panel is more comforting and every claim in it is unfounded. The reference number proves nothing — it is text somebody typed. The partial account number proves nothing. And the instruction to ring the supplied number is the single most dangerous sentence available, because that number is the one thing in the message the sender chose.

The third panel does not resolve anything, and that is the honest position. It hands you two checks you can do in a minute with information you already hold.

If you have already paid, this is still for you.

A significant share of people looking this up have already sent money, already read out a code, already let someone onto their computer. The instinct is to feel stupid and stop looking things up.

The job is written on the assumption that this has happened. One of the questions asks what you have already sent, and nothing about the answer changes the tone of what comes back — it changes what you are told to do first. Containment is time-sensitive in a way that verification is not, and the report always contains at least one step covering money already sent and details already given.

Be clear about what this is not: it is not a reporting service, it does not contact your bank, and it does not recover money. It is a reading of what you have in front of you and a list of things to do next, several of which are phone calls only you can make.

How it runs.

• Photograph or paste what you received. A screenshot of the text, a photo of the letter, the email as it arrived. If part of it is unreadable, send it anyway — it will tell you which part it could not read rather than filling it in.
• Answer six short questions. How it reached you, who it claims to be from, what it wants you to do, whether you actually deal with them, whether it is rushing you, and what you have already sent. That last one is not judged.
• £7 buys 7,000 points; this job holds a ceiling of 1,000. It typically spends about 750, and the rest stays on your balance. There are no revisions on this one — it is a reading of fixed evidence, not a draft to iterate on.

A generation that fails spends nothing.

Who this is for.