You’re not just thinking it, the applicant tracking system really is getting more crowded than it was a couple of years ago. A lot of recruiters, whether they work for staffing agencies, are part of internal HR teams, or focus on corporate talent, are saying they’ve never seen so many applications come in – and a big chunk of those aren’t even real.

Generative AI has handed every job seeker on the planet a free resume writer, interview coach, and personal brand consultant. Used responsibly, that is a good thing. The problem is the other end of the spectrum: applicants who use AI to fabricate entire work histories, invent employers, inflate titles, and produce resumes so perfectly tailored to the job description that they sail past automated screening filters.

For staffing agencies and HR leaders running high-volume pipelines, fabricated applications waste recruiter hours, distort hiring metrics, and introduce real legal and operational risk when a fraudulent hire slips through. This guide breaks down what is actually happening, how to spot the warning signs, and the ATS workflow changes that meaningfully reduce the noise.

The Scale of the Problem in 2026

A few numbers to set the stage:

Most hiring managers, about 74%, have seen content made by AI in job applications, and a lot of them, 58%, are worried about it. This is what a survey done in January 2025 by Resume Genius found out when they asked 1,000 hiring managers in the US about their experiences.

  • 77% of employers now actively screen for AI-generated resume content, per Resume Now’s 2025 AI Applicant Report.
    * Many bosses think people are using computers to pretend to be someone they’re not when they’re looking for a job. In fact, about 60% of the people in charge of hiring are suspicious of this. And get this, about one out of every three of them has actually caught someone using a fake name or having someone else do the interview for them.
  • Almost a quarter of companies have seen identity fraud with their new employees, says Checkr, a company that does background checks.
  • Gartner projects that by 2028, one in four candidate profiles worldwide will be fake.
  • The FTC’s Consumer Sentinel Network logged a 457% increase in reported job-search fraud losses between 2020 and 2024 — from $90 million to over $501 million annually.
  • Deepfake fraud attempts in hiring jumped 1,300% from 2023 to 2024, and Amazon’s CSO disclosed in late 2025 that the company had blocked over 1,800 suspected North Korean (DPRK) applicants since April 2024.

This is no longer something that only affects a small group of people, it’s a big change in the way companies hire employees.

Not All AI Resumes Are Fraud — Know the Three Categories

Not All AI Resumes Are Fraud

Before you tighten your filters, draw a clear line between three very different behaviors. Treating all AI-assisted applicants as fraudsters will cost you good hires and expose you to discrimination claims.

  1. Genuine resumes that use AI for assistance. These are from candidates who have actual experience and use tools like ChatGPT or Gemini to enhance their resume’s language, structure, or formatting. The facts in the resume are true. This is similar to hiring a professional to help with your resume, and most resumes that have been improved with AI fall into this category. There’s no need to view this negatively.
  2. Inflated resumes are a problem because candidates take their real jobs and make them sound way more impressive using AI. For example, they might say “I just helped with quarterly reports” but then change it to “I was in charge of a big data project that made reports come out 35% faster.” They’re not making up the company they worked for or the dates, but they’re definitely stretching the truth about what they actually did. This can be tricky to figure out, but one way to catch it is by asking them lots of specific questions in an interview and also checking with the people they used to work with to see if their stories match up.
  3. Fake resumes are a big problem. Some people make up their whole work history, including the companies they worked for, their job titles, and how long they worked there. With the help of artificial intelligence, they can create a resume that looks perfect in just a few seconds. In some cases, the whole person is fake – they have fake photos, use someone else’s identity, and even use AI to answer questions in interviews. This is the kind of thing that your hiring system needs to be able to catch.

Why Traditional ATS Workflows Are Missing Them

Applicant tracking systems were designed on a simple assumption: applicants are real people, applying honestly, on their own behalf. That assumption no longer holds, and most ATS keyword and ranking logic actively rewards what fraudsters produce best — perfectly tailored, keyword-dense, well-formatted resumes.

A few specific mismatches:

  • Keyword matching favors fabrication. A fabricated resume can be tuned in seconds to hit every keyword in your job description. A genuine candidate with messy, real experience often scores lower.
  • Usually, when recruiters go through resumes, they don’t check if the person is who they say they are. It’s not until something seems weird that they realize it, and by then, they’ve already spent a lot of time on it.
  • Hiring someone without meeting them in person first can be a problem. When you only talk to someone through a video call and look at their resume, it’s easy for them to pretend to be someone they’re not. This can lead to trouble because you’re not really getting to know the real person. In the past, meeting someone in person would give you a sense of who they are, but now that’s not always possible.
  • With automated tools and submissions done by bots, one person trying to cheat can send in many fake applications every day to lots of different companies.
    Don’t give up on your applicant tracking system just yet. Instead, try adding some extra checks and looking at how people behave, on top of the keyword-matching tool you’re already using. This way, you can get a better sense of who’s really a good fit for the job.

How to Detect AI-Generated and Fabricated Resumes

Detection signals

There is no single tell. Detection works best as a stack of weak signals that, together, identify applications worth a closer look. Train your recruiters to spot the following.
Linguistic and Structural Signals

  • Generic, buzzword-saturated phrasing. Look for stock phrases like “results-driven,” “synergistic cross-functional collaboration,” “leveraged data-driven insights” repeated across multiple bullets without specific details. AI-generated text trends toward statistically average word choices, while genuine human writing shows more variation.
  • Identical or near-identical phrasing across multiple applicants. If three resumes for the same role describe their previous job in nearly identical sentences, that is template reuse, not coincidence.
  • Hyper-tailored content that mirrors your job description verbatim. Real candidates rarely echo every requirement word-for-word. AI does this by default.
  • Suspiciously polished formatting with hollow specifics. Beautiful layout, perfect grammar, no typos, but vague accomplishments and round-number metrics (“increased revenue by 30%”) with no named projects, products, or context.

Content and Career-History Signals

  • There are some problems with the timeline. For example, the employment dates overlap or there are gaps that don’t make sense with what’s happening in the person’s life. Also, the career progress is not what you would normally see in that industry – like someone who is 22 years old but says they have been a senior leader for eight years.
  • Employer ambiguity. Companies that cannot be found on LinkedIn, in business registries, or anywhere on the web. Or companies that exist but have no record of the claimed role.
  • Skill-experience mismatch. A candidate claims senior expertise in a tool, framework, or certification that did not exist when they say they used it. AI hallucinates these constantly.
  • Education red flags. Schools, degrees, or graduation years that do not check out against the National Student Clearinghouse or the institution directly.
    Behavioral and Application-Pattern Signals
  • Unusual submission patterns. Multiple applications from the same IP address with different names. Bursts of applications at odd hours. Applications submitted within seconds of a posting going live.
  • Mismatched contact information. Email domains that do not align with the candidate’s claimed location, phone numbers from VOIP services, addresses that map to mail-forwarding facilities.
  • Look for fake profile pictures by searching for them online. Pictures made by AI tools like ThisPersonDoesNotExist or Stable Diffusion often have small mistakes – like earrings that don’t match, blurry backgrounds, or eyes and ears that aren’t quite right. If you search for the picture online, it might show up on a website that sells stock photos, or it might not show up at all, which can be a sign that the person is not real.

ATS Workflow Changes That Filter Out Fakes Early

The verification funnel

Spotting individual signals is helpful. Building those signals into the workflow is what scales. Consider adding the following to your ATS configuration.

  1. To really get a sense of a candidate’s experience, you need to ask them to provide short answers to specific questions about their past projects. Instead of just looking at their resume, have them write a few paragraphs – around 50 to 150 words each – about what they’ve actually done. This way, you can see how they think and what they’ve accomplished. AI systems can generate generic answers, but they won’t be able to provide the same level of detail and personal touch that a human can. By comparing their answers to their resume, you can also look for any inconsistencies and get a better sense of whether they’re being honest about their experience.
  2. Knockout questions with role-specific detail. Ask about a specific tool version, methodology, or industry-standard process the candidate claims to know. Real practitioners answer easily; fabricators stumble or produce textbook descriptions.
  3. When you apply, we need to check that you’re a real person. So, we use a simple two-factor confirmation by phone and SMS. This helps stop bots from submitting fake applications and makes sure we have a working phone number for you. It’s like a security check to keep everything safe and honest.
  4. Async video introductions before recruiter screens. A 60-second recorded answer to a randomly assigned question — not the candidate’s choice — is a low-cost way to expose proxy candidates and obvious deepfakes before you spend live recruiter time.
  5. Checking skills early on is a good idea. A short test that’s watched over by someone can help figure out who’s lying about their technical skills. It’s not a big deal for people who are telling the truth, but it’s a major obstacle for those who are faking it. This way, you can quickly find out who’s really qualified and who’s not.
  6. Identity verification at offer stage. Government-ID verification, paired with a live liveness check, before any onboarding paperwork moves. This catches the “multi-hired ghost worker” pattern where a real person interviews and a different person shows up to work.
  7. Reference checks that reach earlier in a candidate’s career. Fabricated identities are easier to maintain across one or two recent references. Asking for a reference from five or seven years ago, or from a former direct report rather than only managers, is harder to fake.
  8. Cross-reference candidate data against your existing database. A modern ATS should flag when the same phone number, email, address, or resume content appears under different names. This is one of the highest-signal flags you can configure.

Be Careful With Off-the-Shelf “AI Detector” Tools

Don’t just rush into adding an AI tool to spot fake texts to your hiring system and think you’re all set. That’s not a good idea. When you look at the results of testing these tools on their own, you might start to worry.

  • Scribbr’s August 2024 evaluation found GPTZero correctly identified only 52% of AI-generated texts overall.
  • Stanford HAI analysis of more than 10,000 samples showed that AI detectors produce false-positive rates above 20% on non-native English writers and on creative writing.
  • Claims about how accurate vendors are, usually saying they’re right 99% of the time or more, are based on special tests they’ve chosen themselves.

Relying solely on a simple AI-detection score to automatically reject something is not a good idea. This approach can unfairly penalize people who aren’t native English speakers or those who have unique writing styles. Using AI-detection scores in this way can lead to real issues with equal employment opportunities and can have a disproportionate impact on certain groups. Instead, these scores should be considered as just one piece of a larger puzzle, and they should always be reviewed by a human to make a more informed decision.

What This Means for Staffing Agencies Specifically

For agencies, the stakes are even higher than for internal HR teams. A fraudulent placement does not just create a bad hire — it puts your client relationship, your contract, and your reputation at risk. A few things to think about:

  • Make sure you thoroughly screen candidates before they meet with your client. After all, your client is paying for you to find the best fit, not just to find someone – they expect a thorough check, not just a quick search.
  • Document your verification steps. When a placement does go wrong, the difference between a recoverable client conversation and a contract loss is whether you can show a defensible process.
  • Keep an eye on fraud warnings as a way to measure how well recruiters are doing their job. If recruiters are good at spotting and writing down potential problems before sending candidates to clients, they’re helping to keep your company safe – and that’s something to appreciate.
  • It’s time to review and update your client contracts to make sure everyone is on the same page. You need to clearly spell out what you’re responsible for when it comes to verifying information and what your clients are expected to handle. These days, it’s especially important to include clauses about backdoor hiring and misrepresentation – they’re way more crucial now than they were just a few years ago.

The Role of Your Applicant Tracking System

Getting the right tool for the job is crucial, but it’s not a magic solution – it’s just a starting point. If you’ve got the wrong tool, though, you’re already at a disadvantage. So, what should you expect from a decent applicant tracking system? For starters, it should at least allow you to:

  • Configure custom application questions per job and per requisition type
  • Detect duplicate phone numbers, emails, addresses, and resume content across candidate records
  • Capture and store IP addresses, application timestamps, and submission metadata
  • Run automated screening rules and route flagged applications to a human-review queue
  • Integrate with identity verification, background screening, and skills assessment vendors
  • Keep a record of everything that happens with each application, so you can check what was done and when, in case you need to prove something or resolve a disagreement.

If your current system is not able to do those things, or if it’s really hard to set them up, that’s a sign that your ATS is outdated – it was made for the way things were done in 2015, not for how things will be done in 2026.

For over twenty years, StaffingSoft has been creating a special hiring system just for staffing agencies and company HR teams. This system can do lots of things like ask custom questions to job applicants, find duplicate applications, keep track of everything that happens, and connect with other tools. We work really closely with our clients to set up the system in a way that helps them catch fake applicants without slowing down the good ones.

The Bottom Line

Companies that thrive in this new landscape won’t be those trying to completely eliminate AI-generated resumes from their hiring process – that’s just not feasible or legally possible in many places. Instead, they’ll be the ones who learn to adapt and work with this new reality. They’ll find ways to effectively identify and filter out fake resumes, while still making sure they’re not missing out on talented candidates who are using AI tools to enhance their applications.

It’s all about finding a balance and being proactive, rather than trying to fight a losing battle against technology. By embracing this change and developing strategies to navigate it, agencies and HR teams can stay ahead of the curve and come out on top.

  1. Distinguish legitimate AI assistance from fabrication
  2. Layer multiple weak signals into a strong screening process
  3. Add verification steps at the right points in the funnel
  4. Train recruiters to spot the patterns
  5. Use an ATS built for the way hiring actually works in 2026

The recruiters and agencies that get this right will spend more time talking to real candidates and less time chasing ghosts.
________________________________________
Want to learn more about how StaffingSoft‘s system can help you catch fake applicants, avoid duplicate entries, and create a hiring process that fits your needs? Just ask for a demo or give us a call at (866) 267-7477.