Fraudulent Candidate Detection

Last updated: November 16, 2026

Candidate fraud is one of the most common issues affecting trust on the platform. This article will cover the best practices when reviewing a candidate’s profile to determine if the candidate is fraudulent. Recruiters choose their own screening methods. This guide shares signals we’ve seen that may be useful.

This guide is informational. It doesn’t guarantee fraud detection, and Paraform doesn’t warrant that candidates who pass these checks are legitimate.

While we understand fraudulent candidates are getting more difficult to detect, recruiters may not knowingly enable fraud or alter a candidate’s information. See 📄 Recruiter Standards → Submission Specifics for more detail.

Quick tips (TL;DR)

  • Screen every candidate for fraud before submitting. Most fake profiles look mostly plausible, with a few details that don't hold together.

  • Use the quick rule. If one signal looks suspicious, investigate. If several signals back each other up, pause and verify. If there's a major contradiction that can't be resolved, do not submit.

  • Check the profile for consistency:

    • Photo quality and revese-image matches

    • LinkedIn depth and connections

    • Exactly what LinkedIn has verified (identity is not the same as employment)

    • Contact details

    • Whether the resume dates, titles, and employers match LinkedIn

  • Watch for resume red flags:

    • Experience with a technology that's longer than the technology has existed

    • Current buzzwords added to old roles

    • Overlapping full-time jobs

    • Another person's name in the file metadata

    • Contract work presented as direct employment

  • Use the live screen to test the story. Confirm the person matches their profile. Then ask detailed follow-up questions about one project. Real experience gets more specific under follow-up questions, and fabricated experience gets vaguer. See 📄 Candidate Screening for more best practices.

  • Be fair. Accent, fluency, disability, and connection problems are never fraud signals. Give candidates a chance to explain anything.

  • Know when to stop. If there's an identity mismatch, a proxy interviewer, a contradicted claim or a refusal to verify, pause the submission. Keep the evidence, write down the discrepancy and email the Paraform Trust and Safety team via standards@paraform.com.

  • Recruiters may never knowingly enable fraud or alter a candidate's information.

In-Depth Specifics:

The profile detection process

Since all candidates must be screened before being submitted on platform, we recommend incorporating these detection practices as a means to always check whether the candidate you are screening is real.

The quick rule

  • One suspicious signal: investigate.

  • Several signals that reinforce one another: pause and verify.

  • A material contradiction that cannot be resolved: do not submit.

Fake candidates rarely look completely fake. More often, the profile is mostly plausible with a few details that do not hold together.

1. Start with the profile photo

A grainy, unusually polished, or AI-generated-looking photo is a good reason to investigate further.

Look for:

  • Strange lighting or an overly smooth face

  • Distorted glasses, ears, hair, teeth, hands, or background objects

  • Pixelation or tearing around the eyes, ears, glasses, or edge of the face

  • Different photos that appear to show different people

  • A face that appears elsewhere online under another name

  • An extremely low-resolution image that makes comparison difficult

Run the photo through a reverse-image search when the profile already feels questionable. Microsoft has documented fraudulent workers using AI tools to alter headshots and apply the same face across different identities. Use general-purpose reverse-image search only; don’t use facial-recognition services on candidate photos.

2. Check the thoroughness of the LinkedIn profile

A senior candidate with fewer than 100 connections, little activity, no meaningful professional interactions, and almost no trace of prior colleagues should attract more scrutiny.

Look for:

  • Few connections from claimed employers

  • No credible recommendations or professional interactions

  • A recently created or recently populated profile

  • Almost no public work, portfolio, or professional history

  • A long, prestigious career history that exists only in polished resume bullets

  • An “Open to Work” banner on an otherwise fresh or manufactured-looking account

The question is whether the digital footprint makes sense for the story being presented.

3. Check exactly what LinkedIn verified

We suggest clicking any verification badge to see what it confirms.

LinkedIn supports separate verification for:

  • Identity: A verification partner confirmed a government-issued identity document.

  • Workplace: The person confirmed an association with their current company using a work email or another supported workplace credential.

  • Education: The person confirmed an association with a supported educational institution.

An identity verification does not verify the candidate’s employment history. If the application is anchored on a recent or current role, check whether the workplace is verified. Where LinkedIn supports the employer, ask the candidate to verify the current workplace using their company email.

Workplace verification is not available for every company and generally applies to the current workplace. A missing badge is still useful as a prompt to verify the role another way. See the LinkedIn verification guide for more detail.

4. Check if the writing feels generated rather than remembered

AI-written resumes are common. The suspicious part is a profile filled with polished claims that the candidate cannot make concrete.

Look for:

  • Every bullet following the same polished structure

  • Long lists of current AI terms with little evidence of actual use

  • Vague achievements such as “revolutionized,” “optimized,” or “transformed” without a specific result

  • Large volumes of generic LinkedIn content

  • Nearly identical descriptions across unrelated jobs

  • Technical claims that do not match the candidate’s title, tenure, or stated responsibilities

  • A large gap between the polished written profile and the candidate’s ability to explain the work live

Rather than guessing whether AI wrote a resume, focus on one key claim and test whether the experience underneath it is real.

5. Verify if the employment story blurs employer and client

Contract work is common. Misrepresenting contract work as direct employment is not.

Consider looking more closely when a profile shows:

  • A large IT consultancy followed by several prestige technology companies

  • Multiple short engagements at well-known companies with no mention of the staffing firm or employer of record

  • A famous company logo with no mention of the staffing firm or employer of record

  • Overlapping employment dates that are difficult to explain

  • Repeated movement between full-time and contract roles without clear labels

  • A templated career arc followed by a string of difficult-to-verify client assignments

To get clarity and confirmation, you might ask:

  • Were you a direct employee, contractor, consultant, or vendor?

  • Which company employed and paid you?

  • Which company was the client?

  • What team did you support?

  • Who managed your work?

  • How should I describe the relationship to the hiring manager?

The answer should be clear enough that you can represent it in one accurate sentence.

6. Check if the compensation matches the profile

An unexpectedly low salary expectation can be a useful signal.

Ask only about the candidate’s expectations, not their current or past pay. If someone claims to be a senior engineer at a top technology company but is targeting compensation far below the normal range, clarify:

  • Their actual level

  • Their current location

  • Whether their prior work was full-time or contracted

  • Whether the figure is base salary, total compensation, or an hourly rate

  • Whether they personally set the expectation

There may be a reasonable explanation. Resolve the mismatch before submission.

7. Check if the contact information holds together

Investigate when:

  • Names are spelled differently across profiles and documents

  • Email addresses repeatedly change

  • The email domain closely resembles, but does not match, a real company

  • The candidate claims current employment but cannot offer a reasonable way to verify it

  • Phone numbers, locations, and working hours conflict

  • The same phone number appears on unrelated profiles

  • Different accounts appear to belong to different people

  • The candidate relies on a disposable or VoIP number while claiming a well-established local identity

Reading the resume

Signal

What to inspect

What to do

Impossible technology tenure

“10+ years of Claude,” “MCP since 2022,” or another claim that predates the tool.

Ask when, where, and in what form they used it.

Current buzzwords added to old roles

A 2017 role suddenly includes RAG, agentic workflows, LangGraph, or modern LLM tooling.

Test the chronology and ask what terminology the team used at the time.

Product or team did not exist

The candidate claims work before the product launched, before the company existed, or before an acquisition.

Ask for exact dates and organizational context.

Overlapping full-time roles

Several jobs are presented as full-time during the same period without a consulting explanation.

Reconcile the employment structure and hours.

Resume contradicts LinkedIn

Different dates, universities, degrees, titles, employers, locations, or contact information.

Pause until the differences are reconciled.

Another person in the document

The filename, PDF author, hidden metadata, headers, links, or footer reference a different name.

Treat as a strong concern and preserve the original file in a private folder labeled “unverified” until their identity is authenticated.

Copied resume language

A distinctive sentence appears in other resumes or profiles under different identities.

Search the sentence in quotation marks and compare the profiles.

Resume matches the job description perfectly

Every unusual requirement appears almost verbatim, but the candidate cannot explain the work.

Test one claimed project in detail.

Impact is implausibly broad

A junior candidate claims sole ownership of major systems, migrations, and large business results in every role.

Ask about team size, boundaries, attribution, and what they personally shipped.

Education details conflict

The degree, program, attendance period, campus, or confirmation of background does not match the resume.

Ask for the exact program and dates, then verify through public records if available.

Technology timeline cheat sheet

Use this only as a sanity check. Earlier research or predecessor technologies may exist, but the candidate should be able to explain the distinction.

Technology

Earliest credible public date

Example that should trigger a question

Retrieval-Augmented Generation (RAG)

May 2020

“Eight years building RAG systems”

ChatGPT

November 2022

“Five years of ChatGPT development”

Claude

March 2023

“10+ years of Claude”

vLLM

June 2023

“Five years operating vLLM”

LangGraph

January 2024

“Using LangGraph since 2021”

Model Context Protocol (MCP)

November 2024

“Built MCP servers in 2023”

Timeline sources: RAG paper, ChatGPT launch, Claude launch, vLLM launch, LangGraph launch reference, and MCP launch.

Check the wider digital footprint

Reused identity

Search the candidate’s name, email, phone number, profile photo, portfolio URL, and one distinctive resume sentence.

Strong concerns include:

  • The same face appearing under different names

  • The same phone number or email appearing across different candidates

  • Multiple resumes connected to one portfolio but presenting different identities

  • Several profiles with nearly identical wording, timelines, or project claims

  • Overlapping full-time roles the candidate can’t explain or has misrepresented

Portfolio and GitHub provenance

A portfolio or GitHub account is not automatically proof of experience.

Check:

  • Whether repositories are original work or forks

  • Whether the commit authors match the candidate

  • Whether all activity appeared immediately before the job search

  • Whether project dates line up with the claimed role

  • Whether screenshots or demos come from public tutorials

  • Whether the candidate can walk through one contribution, decision, and failure

  • Whether the portfolio website is recent, generic, or connected to another identity

Testimonials and references

Investigate when:

  • Endorser photos appear copied or stolen

  • The people cannot be found at the stated companies

  • Every reference uses a personal email address

  • References avoid phone or video conversations

  • Different references respond in nearly identical language

  • The reference cannot describe the candidate’s actual responsibilities

When verification is needed, contact references through an approved, candidate-authorized process. Say what you will check and who with, so the consent covers what you actually do. Get consent in writing before you collect or share any references and keep their reply. Written consent can be provided within 3 days after oral consent was initially given. Prefer a phone or video conversation over email alone.

Use the live screen to test the story

We strongly recommend that a real phone or video screen be completed for every candidate before submission, and video or an equivalent verification process if the profile has suspicious signals.

Confirm the person

Compare the person on the call with the available profile photos.

Watch for:

  • A face that does not match the profile

  • Unusual lip synchronization or visual artifacts

  • Pixelation or tearing during movement

  • Lighting on the face that does not adapt with the room

  • Someone repeatedly looking off-screen before answering

  • Long delays on straightforward questions

  • Another person appearing to provide answers

  • Repeated video or microphone excuses

  • A different person, voice, or level of ability appearing in later meetings

None of these signals alone indicates fraud. Many have innocent explanations, including disability, assistive technology, connectivity, or language background. Never treat accent, fluency, or disability-related behavior as a fraud signal. If a candidate requests accommodation for the screen, provide one.

Go below the buzzwords

Choose one important project from the resume and stay with it.

Ask:

  • What problem were you solving?

  • What did you personally own?

  • Who was on the team?

  • What existed before you joined?

  • What decision did you make?

  • What tradeoff did you have to manage?

  • What went wrong?

  • How did you measure the result?

  • What would a former teammate say you contributed?

Real experience usually becomes more specific with follow-up questions. Fabricated experience often becomes more generic, repetitive, or contradictory.

For a technical candidate, choose one claimed tool and ask where it sat in the system, why it was chosen, what alternatives were considered, and what failed in production. You do not need to be more technical than the candidate. You need to notice whether the answers form a coherent story.

Additional telltale signs

Signal

Why it matters

Action

Badge verifies identity but not employment

The person may be real while the work history is not.

Verify the current role separately.

Current role is not workplace-verified

The application depends heavily on a claim that may be verifiable through LinkedIn.

Ask for workplace verification where available.

Same photo appears under other names

The identity or image may be stolen or reused.

Pause and escalate by emailing the Paraform Trust and Safety team.

Portfolio is newly assembled

Projects and testimonials may have been copied to manufacture history.

Verify ownership and dates.

GitHub does not support the story

Forked work, different commit authors, or a recent burst of activity conflicts with the claimed experience.

Ask the candidate to explain one contribution.

Phone number is reused or virtual

The contact method may connect multiple identities.

Investigate and record the number.

Location does not hold together

Residence, phone, work hours, and interview availability conflict.

Ask for a coherent explanation.

Different person across calls

A proxy may have completed one stage of the process.

Escalate immediately by emailing the Paraform Trust and Safety team.

Scripted answer behavior

The candidate may be receiving live assistance or reading generated answers.

Change the question and probe.

Written and spoken ability do not match

The profile may have been generated or authored by someone else.

Test the candidate’s underlying experience.

References are too convenient

Generic emails, similar writing, and call avoidance can indicate fabricated references.

Verify independently.

Candidate cannot name the employment relationship

A client engagement may be presented as direct employment.

Do not submit until the relationship is clear.

Know when to stop

Pause the submission and route the candidate for review when you find:

  • A likely identity mismatch

  • Evidence that someone else completed an interview

  • Employment or education that an independent source contradicts

  • A direct-employment claim that turns out to be an undisclosed contract

  • Material dates, employers, titles, or locations that keep changing

  • Work the candidate claims as their own but cannot explain

  • Shared contact or identity information across supposedly different candidates

  • A refusal to complete reasonable verification

  • An admission that part of the profile is false

Do not try to prove the entire case yourself. Preserve the original evidence, record the discrepancy and the candidate’s explanation, and use Paraform’s current review practices outlined in this document. Ensure that any record marking a candidate as unverified is kept private and deleted once identity is verified. This does not change the requirement in the Recruiter Agreement to keep consent and screening records.

Examples of how to ask

Be direct without leading with an accusation:

I want to make sure I represent your background accurately. Your resume lists you as working at Meta, while another source lists a consulting firm during the same period. Can you walk me through who employed you, which team you supported, and how the engagement was structured?

Or:

I noticed a few differences between your resume and LinkedIn profile. Before I submit you, I need to reconcile the dates, title, and employment relationship.

A legitimate candidate may have a messy profile. Give them the opportunity to explain it. Then decide based on whether the explanation and evidence line up.

What can quickly reduce suspicion

  • LinkedIn identity and current-workplace verification

  • Grounded interactions with real colleagues over time

  • Consistent dates, titles, education, contact details, and employment relationships

  • A coherent live explanation of claimed work

  • References who can describe the candidate’s actual contribution

  • Public work with clear provenance and a believable timeline

A single green flag does not prove every resume claim, but coherent evidence across the profile, screen, and verification makes the candidate easier to clear.