'Verified' is one of the most overused words in recruiting software. It is often attached to a profile because an email address was confirmed or because a resume was parsed successfully — neither of which says anything about whether the candidate actually did the work they describe. For hiring teams, that gap between the label and the reality is expensive: it creates false confidence, wastes interview slots and, occasionally, leads to hires that do not work out.
This guide sets out a stricter definition. Verification is a specific check, performed on a specific claim, by a specific party, with a recorded outcome. Anything less is a claim or, at best, evidence. We walk through the difference between claims and evidence, the main types of verification used in India tech hiring, and a labelling approach that keeps everyone honest.
Claims, evidence and verification are different things
A claim is anything the candidate states: 'five years of Java and Spring Boot', 'led an S/4HANA migration', 'built a RAG pipeline for internal search'. Claims are the starting point of every hiring process and most of them are honest, but they are unconfirmed by definition. Evidence is supporting material that makes a claim more credible — a project description with scope and dates, a public repository, a certification ID, a portfolio link or a detailed explanation of architecture decisions.
Verification is the step where someone checks a claim or piece of evidence against an independent source or through a direct assessment, and records the result. A certification is verified when someone looks it up with the issuing body; employment is verified when the employer confirms dates and designation; a skill is verified when a qualified interviewer assesses it. Keeping these three layers separate is the single most useful thing a hiring team can do to avoid over-trusting profiles.
Why 'verified' is so often misleading
Many platforms apply a verified badge for account-level checks: a confirmed phone number, a matched email domain, or a completed profile. These are useful for fraud prevention but they say nothing about competence. When a recruiter sees 'verified' next to a candidate's name, they naturally assume the experience is verified too, and that assumption carries into interviews and offers.
In India tech hiring the problem is compounded by vendor chains. A resume can be reformatted by one or more staffing agencies before it reaches the hiring team, sometimes with skills added to match the JD. By the time it arrives, nobody is sure which parts came from the candidate. Without provenance — a clear record of where each piece of information came from — even honest profiles become hard to trust.
- Account checks (email, phone) are not skill or employment checks
- Parsed resumes are extraction, not verification
- Vendor-reformatted resumes may not reflect the candidate's own words
- Self-reported certifications need a lookup to be verified
- A badge without a recorded check, date and reviewer is just decoration
The main types of verification
Different claims need different checks, and each check proves something narrower than people often assume. Employment verification confirms that someone worked at a company for a period in a given designation; it does not prove what they built. Certification verification confirms a credential was issued; it does not prove hands-on delivery. Understanding these limits helps teams combine checks sensibly instead of relying on any one of them.
For most tech roles, the strongest signal comes from combining project evidence with a skill assessment conducted by someone who knows the domain. For an SAP ABAP developer, that might mean walking through a specific enhancement or a BTP extension they delivered; for a data engineer, discussing how they designed an Airflow DAG or handled late-arriving data in Databricks or Snowflake.
- Employment verification: dates, designation and employer confirmed with the organisation
- Certification verification: credential ID checked with the issuing body (for example SAP, AWS, Azure, Databricks)
- Project evidence: scope, role, stack and outcomes described in enough detail to be probed
- Skill assessment: a structured technical interview or task reviewed by a qualified assessor
- Reference check: a former manager or peer confirms responsibilities and working style
- Education verification: degree and institution confirmed, mainly relevant for campus hiring
Why human review is non-negotiable
AI is very good at extracting a certification ID or a project description from a resume. It is not a verifier. It cannot call a former manager, it cannot authenticate a credential with the issuer unless it is connected to that issuer's system, and it cannot tell a rehearsed answer from real depth on its own. Labelling AI-extracted information as 'verified' confuses a convenience with a check.
A sound rule is that nothing should be marked verified unless a named human reviewer has checked it and the check is recorded — what was checked, how and when. This makes verification auditable and makes it obvious when a profile has plenty of claims and evidence but no actual verification yet. That is not a bad thing; it simply tells the recruiter where the interview should focus.
A practical labelling scheme for candidate information
Provenance labels make the state of every item visible at a glance. Instead of a single badge on the whole profile, each skill, project, certification or employment record carries its own label. Recruiters can then filter or sort by what matters for the role — for example, looking only at candidates whose must-have requirements have at least candidate-confirmed evidence.
RecruitGPT uses exactly this approach. Each item is tagged as Candidate Provided, AI Extracted, Candidate Confirmed, Verified or Not Verified, and requirements show Evidence found, Partial evidence or No evidence yet. 'Verified' is reserved for items a human reviewer has checked. RecruitGPT does not currently connect to third-party verification or background-check providers, so it never implies that an external check has happened when it has not.
- Candidate Provided: stated directly by the candidate
- AI Extracted: pulled from a resume or document by AI, not yet confirmed
- Candidate Confirmed: the candidate reviewed and confirmed the extracted item
- Verified: a human reviewer checked the item and recorded the outcome
- Not Verified: a check was attempted or considered and did not confirm the item
Verification and the candidate experience
Verification should not feel like an interrogation. Candidates in India often juggle several processes at once while serving notice, and heavy-handed checks early in the funnel push strong people away. A better sequence is light at the start — structured claims and evidence — and deeper as the candidate progresses, with employment and reference checks typically reserved for the offer stage.
Consent matters too. Under the DPDP Act 2023, candidates should know what is being checked and why. Giving candidates the ability to review AI-extracted details, correct mistakes and control who can see their profile builds trust and tends to improve data quality, because the candidate is the best source for fixing an extraction error.
A verification checklist for hiring teams
Use the checklist below to tighten how your team talks about and records verification. The aim is not to verify everything for every candidate, but to be precise about what has and has not been checked at each stage.
- Define which requirements need verification for each role, and at which stage
- Never treat parsing or account checks as verification
- Record who verified an item, how and when
- Ask candidates to confirm AI-extracted information before relying on it
- Probe project evidence in interviews with specific follow-up questions
- Reserve employment and reference checks for late stages, with consent
- Flag vendor-submitted resumes and ask candidates to confirm key details directly
See evidence, not just keywords
RecruitGPT structures your job description into clear requirements and shows, for each candidate, the evidence behind every one — labelled by source and verification status. Recruiters make every decision.
Frequently asked questions
What does 'verified talent' actually mean?
It should mean that specific claims about a candidate — such as employment, certifications or skills — have been checked by a named reviewer against an independent source or through a direct assessment. A profile with a confirmed email is not verified talent. Precision about what was checked is what makes the label meaningful.
Can AI verify a candidate's skills?
AI can extract and organise evidence, and can highlight where evidence is missing, but it is not a verifier on its own. Verification needs a check against an issuer, employer or qualified assessor. That is why responsible tools only use the 'Verified' label after human review.
When should employment verification happen?
Usually late in the process, around the offer stage, and with the candidate's consent. Running it earlier adds friction for candidates who may not progress, and many are still employed and serving notice. Early stages are better served by structured project evidence and technical assessment.
Does RecruitGPT run background checks?
No. RecruitGPT does not currently connect to third-party verification or background-check providers. It structures claims and evidence, labels provenance, and marks items as Verified only when a human reviewer has checked them.