"We caught the fake candidate before the offer went out!"

Great news.

Now count the hours it took to get there.

A recruiter read the resume. A hiring manager blocked an hour for a video call (maybe two). Someone wrote a scorecard for a person who, it turns out, doesn't exist.

That's not a fraud win. That's a fraud tax you paid in full... and then got a receipt for.

Where most fraud tools sit

Last week LinkedIn announced new tools to fight fake profiles and invented work histories. Candidate fraud has officially gone from "weird story" to "line item."

The market has noticed. Go shopping for help and you'll find a crowded field, almost all of it downstream of the application:

• Identity verification. A government ID checked against a selfie, usually at the offer or onboarding stage.
• Interview integrity. Deepfake and off-screen-help detection on live video calls.
• Background and employment checks. History confirmed after you've already decided you like them.

These are real tools solving real problems. For some roles you absolutely need them.

But notice what they have in common.

They all run AFTER somebody has spent time on the candidate.

And they're almost always a separate contract (quote-only or priced per check) sitting next to whatever you screen with.

The application already told you something was off

Every applicant hands you four things: a resume, a LinkedIn URL, a phone number and an email address.

Three of those can be checked against the outside world before anyone reads a word.

That's the cheapest moment you'll ever get to notice a problem, because nobody has invested anything yet.

It's also where SquarePeg's authenticity checks run. Every applicant we score is checked against eleven named signals:

• LinkedIn. The URL doesn't resolve to a profile, in our 400-million-person database or in a live search. Fewer than 100 connections. No photo.
• Phone. Voicemail-only, toll-free or non-fixed VoIP. Barely any call activity in the past year. Invalid or can't receive calls. Registered to a different name than the one on the application.
• Email. Only appeared in public records within the last three months. Registered to a different name. Comes from a disposable or low-reputation provider.
• Resume. Mirrors the job description's structure and language a little too perfectly.

Ten of the eleven don't touch the resume at all.

That's on purpose. The resume is the one document the applicant fully controls... and it's also the easiest one to generate.

"Just give me a fraud score"

It sounds efficient. It's actually useless.

A number like "72% likely fraudulent" can't be checked and can't be argued with. You either trust it blindly or ignore it (and most people end up ignoring it).

So SquarePeg doesn't do that. You get a flag next to the candidate's name, and when you open it you see exactly which signals fired:

• "The name registered to this phone number doesn't match the name on the application."
• "This LinkedIn URL doesn't resolve to any profile."

The signals roll up into a confidence level (high, mixed, limited, low or unknown). But the level isn't the point. The REASONS are.

A named reason is something a recruiter can check in thirty seconds and agree or disagree with. Same glass-box idea as the score itself.

A flag is a question, not a verdict

An authenticity flag in SquarePeg never rejects anyone.

No auto-decline below a threshold. No rule that quietly moves people out of your pipeline. A flag means exactly one thing: a person should take a closer look.

That's not caution for its own sake. Every one of these signals has an innocent explanation:

• Plenty of excellent engineers barely touch LinkedIn.
• Lots of real people use Google Voice.
• Someone who just changed their name will trip a name mismatch.

One signal on its own is weak evidence.

Several together, on a remote role that attracts fraud, is worth fifteen minutes of a recruiter's attention before the interview hits the calendar.

So you filter a job by confidence level or by individual signal, look at the flagged candidates first, and decide. Advancing or rejecting is still something a recruiter physically does.

The software tells you where to look and why. You make the call.

What it doesn't do

I'd rather tell you the edges now than have you find them in month two.

SquarePeg doesn't check government IDs. It doesn't watch your video interviews for deepfakes.

If your biggest risk is a stand-in showing up on the call, or you're hiring for roles where identity verification is legally required, you still want a tool built for that stage.

What application-stage checks do is shrink the problem before it gets there. The obvious fakes get flagged before anyone spends an hour on them, and the expensive checks downstream are saved for candidates you're actually serious about.

Do you need a second contract for this?

No.

The authenticity checks aren't an add-on. They run as part of scoring every applicant, inside the same single credit that parses the resume, enriches it and scores it against your criteria.

No separate fraud product. No per-check fee. Nothing extra to integrate.

If a candidate comes through your ATS and gets scored, they get checked.

Catching a fake at the offer stage makes a good story.

Catching it at the application makes a cheaper one.