How to stop AI-generated and fake applications in Lever

The order to switch on Lever's native defenses against bot and AI-written applications, which package covers each one, and where you still need a person or an ID partner.

The short answer

Lever spam applications fall fastest when you stack a few native settings in the right order. To stop AI-generated resumes in Lever, make resume and phone mandatory on the application form, add required knockout questions, and build an automation workflow that auto-archives applicants whose answers disqualify them. Then turn on the repeat application block, which is off by default, and review Fraud Signals tags before anyone gets an interview. Fraud Signals flags; it never rejects. For identity checks, the routes we could confirm are Lever marketplace partners and your background check provider.

Why fake and AI-written applications reach Lever

Generative AI made the cover letter and the tailored resume close to free. Gartner reported in July 2025 that 39% of job applicants surveyed had used AI in their application, and 6% admitted to interview fraud. Gartner also predicted that by 2028, one in four candidate profiles worldwide will be fake.

Some of that is harmless polish. Some is organized fraud. In January 2025 the FBI described North Korean IT workers reusing phone numbers, often VoIP numbers, across many resumes, and changing addresses or payment platforms during onboarding. Phone line type is one of the details Lever's Fraud Signals checks, which is why the setup below asks for a phone number early.

How to stop AI-generated resumes in Lever, step by step

Each step below makes the next one more useful. Form settings give your workflows clean data to act on. Workflows remove the obvious mismatches. Signals and ID checks then focus on the smaller group left over.

  1. Tighten the application form. Lever always requires Name and Email. Resume, Phone, Location and Company can each be set as optional or mandatory. Requiring a phone number gives Fraud Signals a line type to check later. Lever also notes that suspected bots must solve an hCaptcha before they can use the Current Location field.
  2. Write a short set of knockout questions. Custom application questions can be saved as reusable sets and marked internal or external. Keep them factual: legal right to work, willingness to work the posted hours, a required license. Required questions with fixed answers are the ones a workflow can read.
  3. Auto-archive on disqualifying answers. In automation workflows, use the trigger Application submitted, add a condition on the custom question answer, and choose Auto-archive. You can add a rejection email and delay it, so it doesn't land seconds after the candidate clicks submit. Workflows can also condition on origin or source, which helps if one channel sends most of the junk. Only a Super Admin can create these workflows.
  4. Turn on the repeat application block. This global setting stops the same email from applying to the same posting ID again. The default window is six months. It is off until an admin turns it on.
  5. Decide how each posting uses LinkedIn. With LinkedIn Apply Connect, you can switch Easy Apply off for a single posting. Required custom questions carry over to Easy Apply, except File Upload and Rating types, and a required resume carries over as well. Adding #LI-DNI to a posting opts it out.
  6. Review Fraud Signals before scheduling anyone. Fraud Signals checks for disposable email addresses, Application Volume, phone line type (mobile, VoIP or landline) and company background. Candidates are labeled Tagged Suspicious, Re-Verify or Failed Checks. Users need the AI Screening Companion permission to see them.
  7. Merge duplicates carefully. Lever dedupes profiles with the same email on its own and shows a banner when it suspects other duplicates. Merging cannot be undone, so check before you click.
  8. Close postings once they're filled. Turn on auto-close in Job site, then Posting Configuration, so a posting closes when one person is hired. A stale open posting keeps collecting applications nobody will read.
  9. Add identity checks where risk is high. Lever's marketplace lists Persona, which offers ID checks, selfie liveness and deepfake checks, and Proof. These are third-party partners, not Lever features. Your background check provider is the other option.
  10. Interview against a written plan. Build interview plans and feedback forms in Lever so every interviewer scores the same things. Gaps between rounds are easier to see when the criteria don't change.

Lever spam applications: feature, what it stops, package

Lever's package breakdown names the flagship package as the one that includes Fraud Signals, knockout questions and automation workflows. Legacy packages differ, so check yours with Lever before you plan around a feature.

FeatureWhat it stopsPackage
Mandatory form fields and hCaptcha on Current LocationThin, bot-filled applications missing basic detailsNot named in the help article
Custom application questions (knockout questions)Applicants who don't meet a hard requirementKnockout questions: flagship package
Automation workflows with Auto-archiveDisqualified applicants sitting in the active pipelineFlagship package; Super Admin to create
Block repeat applicationsThe same email applying to the same posting againNot named; off by default
LinkedIn Apply Connect controlsUnfiltered Easy Apply volume on a postingNot named in the help article
Fraud SignalsNothing on its own; it tags risk for a person to reviewFlagship package
Duplicate dedupe and mergeSeveral profiles for one personNot named in the help article
Auto-close on hireApplications to roles you've already filledNot named in the help article
Persona or Proof (marketplace)Applicants using someone else's identityThird-party partner, not Lever-native

Caveats that change how you set it up

  • Fraud Signals never removes anyone. Lever's spring 2026 release post said "no one is ever automatically rejected." Plan for someone to work the tagged list every day during busy periods.
  • API applications skip the repeat block. So do referrals and internal applicants. If a job board or agency feeds Lever through the API, repeat applications from that channel can still land.
  • Merges are permanent. A wrong merge mixes two people's records. Train recruiters to check before merging a suspected duplicate.
  • Each required field costs something. More mandatory fields and questions mean fewer bots, and also fewer honest applicants who finish. Use the heavier setup on the postings that draw the most junk.

Adding Carolina to a Lever pipeline

Filters shrink the pile. They don't tell you who is good. NTRVSTA's AI recruiting agent, Carolina, connects to Lever and scores every applicant who gets through against your scorecard. She writes down her reasons for each candidate, so a recruiter can check the logic instead of trusting a number. She can also run live first-round interviews, by video or voice and at any hour. The questions adapt to each answer, and she asks follow-ups on what the candidate just said. Her scorecards and transcripts are written to Lever. Setup details are on our Lever integration page.

Follow-up questions make a memorized script or answers fed by someone else harder to sustain, and a human reviewer can often spot the strain in the transcript. Still, that is evidence for a person to weigh, not identity verification. NTRVSTA doesn't verify who someone is. Pair it with a marketplace ID partner or your background check provider; Checkr, Sterling and HireRight connect to NTRVSTA.

How much she does without asking is your call. Each action can be set to suggest, ask first or run on its own, and offers always wait for a person. For the bigger picture, see how to stop AI resume spam across your funnel, or estimate what junk applications cost your team with our spam cost calculator.

Where NTRVSTA isn't the right fit

If knockout workflows and Fraud Signals already bring each posting down to a list your recruiters can read in an afternoon, you may not need another layer. Lever's own tools may be doing the job.

If your worry is identity fraud, buy an ID partner or a stronger background check first. NTRVSTA won't solve that problem. And for hourly hiring built on text-to-apply and chat scheduling, tools designed for that flow are a better match.

Related reading: our guide to proxy candidates in remote interviews, and our look at hidden help during interviews. If you also run Greenhouse, read how to stop fake resumes in Greenhouse.

Questions

Does Lever Fraud Signals reject suspicious candidates?

No. Fraud Signals tags candidates as Tagged Suspicious, Re-Verify or Failed Checks based on email, application volume, phone line type and company background. Lever said at launch that no one is ever automatically rejected. A recruiter reviews the tag and decides.

Is blocking repeat applications on by default in Lever?

No. The global setting is off by default. Once on, it blocks a second application to the same posting from the same email for a window that defaults to six months. It does not apply to applications that arrive through the API, referrals or internal applicants.

Do my knockout questions apply to LinkedIn Easy Apply candidates in Lever?

Mostly. With LinkedIn Apply Connect, required custom questions carry over to Easy Apply, except File Upload and Rating questions, and a required resume carries over too. You can also turn Easy Apply off for a single posting, or add #LI-DNI to a posting to opt it out.

How do I verify a candidate's identity if I use Lever?

Use a partner. Persona and Proof are third-party providers listed in the Lever marketplace, not Lever-native features. Your background check provider is the other route. NTRVSTA connects to Checkr, Sterling and HireRight, but NTRVSTA itself does not verify identity.

What does NTRVSTA add on top of Lever's own tools?

Carolina, NTRVSTA's AI recruiting agent, scores every applicant who gets past your filters against your scorecard, with written reasons, and can run a live first-round interview by video or voice. Scorecards and transcripts go back to Lever. Each action can be set to suggest, ask first or run on its own.

Curious what written reasons look like on your own Lever postings? Book a demo and we will run one of them live. Plan details are on the pricing section of our homepage.

Sources

NTRVSTA
‍