# Candidate Fraud Checks: Compare Résumés with LinkedIn

Candidate fraud detection starts before the interview – an AI agent compares each applicant's résumé with their LinkedIn profile and flags what to verify.

Candidate fraud detection – catching fake job applicants and résumés that don't match the person behind them – now starts before the first interview, especially for remote roles. This guide is for recruiters and talent acquisition leads holding a shortlist. In [Checkr's 2026 survey](https://checkr.com/blog/hiring-hoax-2026-ai-fraud-outpacing-detection) of 3,000 US hiring managers, 40% had interviewed a candidate later found to be using a fake identity, and 65% had uncovered candidates who misrepresented their skills or experience. The cheapest first check is whether the résumé and the LinkedIn profile tell the same story.

> **The short version.** Before you interview anyone on your shortlist, compare their résumé with their LinkedIn profile – employers, job titles, start and end months, schools. Check that each employer they claim has a LinkedIn company page, and where the two disagree, see whether other people list the same employer and a similar title in the same years. Treat every difference as a reason to ask or verify, never a reason to reject: profiles are written by their owners, often out of date, and a convincing profile can be rented. An AI agent on your own LinkedIn account runs the comparison across the whole shortlist in minutes and returns a result, the reason and a next step for each applicant – it rejects no one and contacts no one.

## A worked run: check a shortlist against LinkedIn

The agent works through [Linked API](/) on your own LinkedIn account, in a real cloud browser at a human pace – Linked API never stores your password. Connect it through the [MCP server](/mcp/overview) in Claude, ChatGPT or Cursor, give a coding agent the agent-friendly [CLI](/cli/getting-started), or install ready-made [skills](/skills); the [LinkedIn AI agent guide](/guides/linkedin-ai-agent) covers the setup.

**You:** I'm hiring a remote senior backend engineer. Here are 12 shortlisted applicants with their LinkedIn URLs and the employers, titles and dates from their résumés. For each one, compare the résumé with the LinkedIn profile – employers, titles, start and end months, schools. Check that each employer they list has a LinkedIn company page. Where the résumé and the profile disagree, find two or three people who list the same employer and a similar title, and compare their dates. Flag only differences between the résumé and the profile – not how active someone is on LinkedIn. Return a result, the reason and a next step for each. Don't reject anyone and don't contact anyone.

**Agent** (used Linked API: fetch_person ×12, fetch_company ×27, search_companies, search_people ×2, fetch_person ×3):
- experience and education of each applicant
- compared with the résumés – 8 consistent, 2 mismatches, 1 profile with no experience listed, 1 profile not found
- every employer listed on the 10 profiles with experience – 27 unique, all with company pages
- "Quillon Systems", an employer on one résumé but not on that profile – no matching company
- a similar title at the two mismatched employers – Brightline Pay: 3 found; Quillon Systems: LinkedIn offered no company option
- the three Brightline Pay profiles – their dates there

12 applicants: 8 consistent, 2 mismatches, 1 profile with no experience listed, 1 profile not found. These are reasons to ask or verify, not verdicts – nobody was rejected or contacted. Five of the rows:

| Applicant | Result | What the comparison found | Next step |
| --- | --- | --- | --- |
| Alex M. | Consistent | Employers, titles and months match within a month | Normal process |
| Jordan K. | Mismatch | Résumé: Staff Engineer, Brightline Pay, 2022–2025. Profile: Senior Software Engineer, Mar 2023 – Jan 2025. 3 other profiles list a Staff title there, all from 2024 on – their own entries, not proof either way | Ask Jordan about the title and dates; verify with Brightline Pay |
| Sam T. | Mismatch | Résumé: Backend Lead, Quillon Systems, 2020–2023. Profile: freelance work for those years, no Quillon role. No LinkedIn company under that name | Ask Sam; check the company outside LinkedIn – website, business registry |
| Riley P. | Not enough on LinkedIn | The profile lists no experience, so there is nothing to compare – this says nothing about Riley | Ask Riley to clarify, or for another record of the work history |
| Casey W. | Profile not found | The LinkedIn URL on the résumé opens no profile | Ask for the right URL – a missing profile is not a red flag on its own |

*Nothing is sent and no one is rejected – you decide each next step.*

*Example run – the applicants and companies are invented. Profiles are self-reported: a match is not proof of employment, other people's entries are their own claims, and identity is a separate check for everyone you shortlist.*

This run used 15 profile visits, 27 company pages and 3 searches. By default Linked API allows 80 profile visits and 100 company page visits a day and 300 people and company searches a month – your settings can differ, see [limits](/docs/admin-limits). A shortlist of 30 usually fits in a day or two. Look up other people's dates only for the roles that disagree, not for everyone.

## What candidate fraud looks like before the interview

Candidate fraud runs from a stretched title to a person who is not who they say they are. Only part of it shows when you compare a résumé with a LinkedIn profile:

| Kind | Example | Shows in the comparison? |
| --- | --- | --- |
| Inflated titles or dates | "Staff" on the résumé, "Senior" on the profile | Often – as a mismatch |
| Invented employers | A company with no trace on LinkedIn or anywhere else | Sometimes – as not found |
| A borrowed or stolen identity | A real person's details used by someone else | No – the profile may be genuine |
| A rented or claimed account | A profile with years of real activity, used by someone else | No |
| A different person at the interview | A stand-in or a deepfake on video | No – a later stage |

US government guidance describes the same pattern. A 2022 [joint advisory](https://ofac.treasury.gov/media/923126/download) from the State Department, the Treasury and the FBI on North Korean IT workers lists "inconsistencies in name spelling, nationality, claimed work location, contact information, educational history, work history, and other details across a developer's freelance platform profiles, social media profiles" as a red flag. The FBI's [October 2026 notice](https://www.ic3.gov/PSA/2026/PSA261013) adds that these workers pay to use other people's professional networking accounts "which often contain years of legitimate activity history". The roles named are remote technical ones – "information technology and computer programming, database, and software related job functions", per the FBI's [2022 notice](https://www.ic3.gov/PSA/2022/PSA220628). Gartner [predicts](https://www.gartner.com/en/newsroom/press-releases/2025-07-31-gartner-survey-shows-just-26-percent-of-job-applicants-trust-ai-will-fairly-evaluate-them) that "by 2028, one in four candidate profiles worldwide will be fake" – a forecast, not a measurement.

## Detecting candidate fraud: compare the résumé with the profile

Line up the two documents role by role:

- **Employers** – every company on the résumé should appear on the profile, and the other way round.
- **Titles** – the same level on both; "Staff" on one and "Senior" on the other is a question.
- **Start and end months** – LinkedIn stores experience by month, so compare at month level; if the résumé gives years only, compare years.
- **Overlapping roles** – two full-time jobs at once on one document but not on the other.
- **The current role** – the profile should show the job the résumé says they hold now.
- **Schools** – names only. The education section, as the agent reads it, carries no dates, so degree years cannot be checked this way.

Most differences have ordinary explanations:

- contractors who list the agency rather than the client, or the reverse;
- titles the employer renamed or levelled differently;
- a profile not updated since the last move;
- several titles under one employer, which LinkedIn groups and a résumé may not;
- a résumé tailored to the job, with older roles dropped.

That is why a difference is a question for the candidate, not a conclusion.

The comparison itself is well known. [SHRM](https://www.shrm.org/topics-tools/news/technology/how-to-spot-fake-job-applicant-online) quoted a recruiter in 2023: "A profile might say that you were with a company for two years, but your resume says you were there for three." Lever's guide to recognising fake candidates lists "obvious inconsistencies between their application and LinkedIn profile", and Gem's lists "LinkedIn employment dates that don't match the resume or application". What an agent adds is running it for every applicant on the shortlist in the same way, plus the next two checks.

## Check the employers – and who else lists them

**The company's LinkedIn record.** For each employer, the agent opens the LinkedIn company page and reads what it says – website, industry, size and the founding year if the page gives one. A founding year later than the applicant's claimed start date is worth a question. A page is a LinkedIn record, not confirmation that the business is genuine, and many small or defunct companies have no page at all. When there is none, check outside LinkedIn – the company's website, a business registry – before reading anything into it.

**Other people who list the same employer.** Where the résumé and the profile disagree, the agent searches for people who list the same employer and a similar title, and reads their dates there. If two other profiles list that employer and similar roles in overlapping years, that is all it shows – their entries are their own claims too, since LinkedIn members [add and edit their own experience](https://www.linkedin.com/help/linkedin/answer/a593695) – so independent verification is still needed. Overlap does not show that the applicant worked there, or worked with them. Each profile read is one visit, so do this only for the roles that disagree. And keep it read-only: messaging those people is a reference check – a separate step the candidate should know about.

## The four results – what each means and what to do next

| Result | Raised when | What it means | Next step |
| --- | --- | --- | --- |
| **Consistent** | Employers, titles and months agree within a month | Nothing on LinkedIn contradicts the résumé – not proof of employment | Normal process |
| **Mismatch** | An employer missing, a different title, months apart, a different current role | The two documents disagree – often for an ordinary reason | Ask the candidate, then verify with the employer |
| **Not enough on LinkedIn** | The profile lists no experience, or not the years in question | There is nothing to compare – it says nothing about the person | Ask the candidate to clarify those years |
| **Not found** | No profile at the URL, or no LinkedIn company for a claimed employer | The record is not on LinkedIn under that URL or name | Ask for the right URL; check the employer outside LinkedIn |

Results come only from a difference between the résumé and the profile, or from something that cannot be found. Follower count, how often someone posts, how many roles they list and whether other people turned up never raise a result on their own – plenty of genuine candidates barely use LinkedIn. Identity verification is one policy for everyone at the same stage, such as every finalist before an offer, not an extra hurdle for quiet profiles. The FBI's 2026 notice makes the same point about fraud controls: aggressive ones risk "generating false positives that could harm legitimate employees and filter out qualified applicants", and need "careful detection calibrations, substantive review, and continuous refinement". The check never uses photos, and names, photos and graduation years are never used to infer anything about a person – the same [fairness rule](/guides/ai-candidate-sourcing-linkedin#guardrails-fairness-review-pace-and-limits) as in AI candidate sourcing.

## What LinkedIn data cannot prove

- **Employment.** Experience is entered by the member, and so are other people's entries that seem to back it up.
- **Identity.** A rented or claimed account comes with real history, and the FBI notes these workers "cannot be identified reliably with a single indicator".
- **Verification badges.** A badge can stand for a checked government ID, a workplace or an education – open it and read the type and the organisation. An identity verification "means that a member's government-issued ID was verified by one of LinkedIn's verification partners" ([LinkedIn Help](https://www.linkedin.com/help/linkedin/answer/a1359065)). A workplace verification is obtained for the employer listed as current at the time, and it can stay visible after a job change ([LinkedIn Help](https://www.linkedin.com/help/linkedin/answer/a568484)). No badge verifies the full work history, or that the person applying is the person who owns the account. The agent does not read badges; apps can read verification details only with the member's consent, through [Verified on LinkedIn](https://www.linkedin.com/help/linkedin/answer/a9385085) – more in [LinkedIn API access](/guides/linkedin-api-access).
- **Profile age and connection count.** Neither is in the data the agent reads – and an old profile proves little when accounts with years of history can be rented.
- **A working relationship.** Overlapping dates at one employer are not the same as working together.
- **That a company exists.** A LinkedIn page is a record, not a registration.

The next step sits outside LinkedIn. The 2022 advisory puts it plainly: "Verify employment and higher education history directly with the listed companies and educational institutions, using contact information identified through a search engine or other business database, not directly obtained from the potential employee or from their profile." Candidate identity verification and employment verification are separate services – the profile comparison tells you which questions to ask first.

## Deepfakes and stand-ins come later

Some fraud shows only at the interview – a different person on camera, someone answering off-screen, or video whose "actions and lip movement … do not completely coordinate with the audio", as the FBI's [2022 notice](https://www.ic3.gov/PSA/2022/PSA220628) describes deepfake interviews. A profile check catches none of this; it narrows what to ask before you get there.

## When a bundled fraud tool is the better choice

If you want a fraud check on every applicant inside your applicant tracking system, with identity verification attached, a bundled tool fits better. ATS vendors such as Ashby and Gem build fraud signals into the application flow, and dedicated tools already automate this comparison: Endorsed "compares the education and work experience on a candidate's resume to what's on their LinkedIn profile", and HireFoundry's candidate trust agent "shows where information differs and points back to each source without making assumptions about intent". The own-account route fits a recruiter who already has a shortlist and an AI assistant, wants every result explained, and wants the check on other people who list the same employer.

## What this does not do

- No identity or ID check, and no background check.
- No employment verification – it compares two documents the candidate wrote.
- No score, ranking or rejection – a person decides every next step.
- No contact with the applicant or anyone else, and no write to your ATS.
- Nothing your own account cannot see, and nothing about how active a person is on LinkedIn.
- No inbound screening – it runs on a shortlist a person chose. For agents that source and screen, see [AI recruiting agents](/guides/ai-recruiting-agent).

Each profile the agent reads is a real visit and shows in that person's viewer list under your profile viewing setting – see [LinkedIn private mode](/guides/linkedin-private-mode).

## Frequently Asked Questions (FAQ)

#### What is candidate fraud?

Candidate fraud is when a job applicant misrepresents who they are or what they have done – a fake or borrowed identity, an invented employer, an inflated title or dates. It ranges from stretched dates on a résumé to a different person at the interview, and only the first kinds show up when you compare a résumé with a LinkedIn profile.

#### How do you spot fake candidates?

Compare each shortlisted applicant's résumé with their LinkedIn profile, check that each claimed employer has a company record, and for the roles that disagree, see whether others list the same employer and title in those years. Then ask the candidate about every difference and verify employment and identity through independent sources. No single signal identifies a fake candidate.

#### How do you prevent candidate fraud?

Layer the checks: compare documents before the interview, verify identity for everyone at the same stage, and verify employment and education with the employers and schools themselves, using contact details you find independently – the approach the 2022 State, Treasury and FBI advisory recommends. Apply the same steps to every candidate.

#### Can LinkedIn verify a candidate's employment?

Not their work history. A workplace verification is obtained for the employer listed as current when the member verified – through a work email or similar – and can stay visible after they move. The experience section is entered by the member. Open any badge to see its type and organisation, and verify past employment with the employer.

#### Does a strong LinkedIn profile prove a candidate is real?

No. The FBI warns that fraudulent applicants pay to use other people's accounts "which often contain years of legitimate activity history". A long, active profile proves nothing for anyone – and a quiet profile proves nothing against anyone.

#### Will the candidate see that I checked their profile?

Yes, the way any visit shows. Each profile the agent reads appears in the person's viewer list under your profile viewing setting, exactly like a manual visit – see [LinkedIn private mode](/guides/linkedin-private-mode).

#### Is it fair to check applicants' LinkedIn profiles?

It can be, if the check is the same for everyone on the shortlist, every result is a question rather than a decision, the candidate can explain, and nothing is inferred from names, photos or graduation years. Where automated tools in hiring are regulated – New York City and the EU, for example – see [screening and assignment](/guides/ai-for-staffing-agencies#screening-and-assignment-why-they-go-last) in our staffing guide.

## Build it in, or hand it to an agent

Get it out of the box: connect the [MCP server](/mcp/installation), the agent-friendly [CLI](/cli/getting-started) or a ready-made [skill](/skills), and give it your shortlist in plain language – the [LinkedIn AI agent guide](/guides/linkedin-ai-agent) compares the options. Or build the same check into your own tools – for example, run it when a candidate reaches the shortlist stage in your ATS – with the REST API, the Node and Python [SDKs](/sdks/installation) and the shell CLI; the experience data it reads is documented under [person experience](/docs/action-st-retrieve-person-experience). Either way it runs on your own LinkedIn account. Plans are flat per seat, from $49/mo billed annually – see [pricing](/pricing). For the step before, see [ChatGPT for recruiting](/guides/chatgpt-for-recruiting) – shortlist and reach candidates; for re-checking contractors you have already placed, [AI for staffing agencies](/guides/ai-for-staffing-agencies).

*Facts verified November 14, 2026 – Checkr's and Gartner's research, the FBI notices and the State, Treasury and FBI advisory, LinkedIn Help on verification and profile experience, and the Lever, Gem, Endorsed and HireFoundry pages checked on that date.*
