Why we put fraud flags inside your ATS
Table of Contents
When I joined Tofu a few weeks ago, I assumed the thing customers would talk about most was the detection itself: the signals, the model, how many fakes we catch. Reading back through our customer stories, the thing that comes up again and again is more boring: where the answer shows up.
Tofu runs on every application the moment it lands in your ATS. The result goes into the candidate's profile as a label, with the details behind it in the application notes. Recruiters don't open another tool or run another search, and the fake ones get filtered before anyone books a call.
Why does it matter where a fraud flag shows up?
Because recruiters live in the ATS. Every extra tab is a step that gets skipped on a busy Tuesday.
Before Lithic used Tofu, checking a suspicious applicant meant what their team called "time-intensive internet searches to validate candidate authenticity." Googling the name, opening the LinkedIn, checking whether the phone number looked real. That works when you're suspicious of one person. It doesn't scale to a pipeline, so in practice it only happens when something already feels off, which is usually after someone's had a phone screen.
“But Joseph, can't recruiters just check LinkedIn themselves?”
They can, and a lot of them do. The trouble is that a good fake has a good LinkedIn (sometimes a real person's, sometimes one that's years old), so the manual check takes time and still misses the ones you most need to catch.
How do you flag fake candidates inside an ATS workflow?
Basically, three things happen:
- An application comes in. Tofu checks it automatically: email, phone carrier, LinkedIn, employment history, resume patterns, and whether any of it has turned up on other applications in our network.
- The result lands in the ATS. The candidate gets a label (No issues, Moderate risk or Suspicious) and the verification details go into the application notes, so anyone who opens the profile can see why.
- Your team decides. Tofu doesn't reject anyone. You choose where the check runs, whether that's at application review, before interviews, before the offer, or across every active stage, and what to do with what it finds.
It works with Greenhouse, Ashby, Lever, Gem, Workday, Oracle and 42 more ATSs, so setup means connecting the ATS you already have.
What Lithic got out of it
Lithic is a fintech, which makes it a target. Before Tofu, their team was seeing multiple fake applicants reach phone screens every week, and engineering managers were losing precious interview slots to candidates who weren't real.
After they connected Tofu to their ATS, fraud results started showing up in candidate profiles automatically. Their numbers:
- 92% to 97% reduction in fake applicant phone screens
- Manual verification gone entirely
- One click to remove a bad actor from the pipeline
Morgan Stanley, their Head of People, put it this way: "Since we integrated Tofu with our ATS we have a much higher degree of confidence in our applicants."
To make that concrete: say fake applicants were reaching three phone screens a week, at 30 minutes each with prep. That's 78 hours a year of a recruiter's time spent talking to people who aren't who they say they are, before you count the engineering managers' interview slots.
What Tailscale got out of it
Tailscale's story is my favourite, because they didn't come to us for fraud at all. They came for help reviewing high application volume. Once they were using Tofu, the same signals started surfacing a steady stream of suspicious applications, and turning on fraud detection was, in their words, "a flip of a switch."
The impact was immediate. The number of interviews scheduled with fraudulent candidates dropped to zero, and no fake candidates made it to a hiring manager.
Laurel Kiskanyan, their Lead Talent Acquisition Partner: "their fraud offerings were a side bonus. The fraud detection ended up being just as much of a time saver for our entire recruiting team."
What should you look for in ATS fraud detection?
If you're evaluating fraud detection, ask where the result ends up. If the answer is a separate dashboard, picture your team on their busiest week and be honest about how often they'll open it. If it lands in the candidate profile with the reasons attached, it gets used.
And ask what happens with the middle cases. Some real people look odd on paper (a new email, a VoIP number, a VPN). A good setup marks those as worth a closer look and leaves the call with your team.
If you want to see what the labels look like on your own applicants, here's how it works, and we can set it up on your ATS.
FAQs
What beats standard ATS screening for detecting fake candidates?
ATS screening ranks applications against the job, and a fake application is written to match the job. What catches it is checking the person behind the application when it arrives: how old the email is, whether the phone and LinkedIn line up, and whether the same details have shown up on other applications under different names.
Which applicant tracking system is best at detecting fake candidates?
The ATS matters less than what's connected to it. Several ATSs now ship their own fraud features, and a dedicated fraud detection layer can sit on top of any of them. Tofu connects to Greenhouse, Ashby, Lever, Gem, Workday, Oracle and 42 more and writes its results back into the candidate profile.
Does Tofu work inside my ATS?
Yes. Tofu connects to Greenhouse, Ashby, Lever, Gem, Workday, Oracle and 42 more. Fraud results appear as a label on the candidate profile, with verification details in the application notes.
Does Tofu reject candidates automatically?
No. Tofu flags candidates with the reasons behind each flag, and your team decides what to do. You also choose where in the process the check runs.
Where in the hiring process does the fraud check run?
Wherever you turn it on: at application review, before interviews, before the offer, or across all active stages. Starting at application review means fake candidates never reach a phone screen.
Do recruiters need to learn a new tool?
Not for the fraud check itself. The labels and notes show up in the ATS they already use.
