Two CVs land for a platform engineering role. Both list Kubernetes. The first has it in a skills row, twelfth in a list of nineteen technologies. The second devotes a paragraph to rebuilding a deployment pipeline on it, including why the team abandoned their first approach to ingress and what the migration cost them in downtime. A keyword parser scores these identically. A human who reads both knows immediately that they are not comparable candidates.
The gap between those two documents is the difference between claimed versus evidenced skills in recruitment, and it is the single highest-use distinction in screening. Get it right and shortlist quality improves without any change to your sourcing. Get it wrong and you optimise your entire pipeline for whoever is best at formatting a skills section.
Why the Distinction Exists at All
CV screening inherited its core assumption from document retrieval: that the presence of a term indicates the relevance of a document. That assumption works tolerably for searching a library and poorly for assessing a person, for one reason. A library catalogue is not trying to persuade you.
A CV is an advocacy document written by an interested party who knows, or can trivially discover, what terms you are searching for. The moment screening rewards term presence, the rational candidate strategy becomes term inclusion. Both sides know this, and that is why "tailoring your CV to the job description" is standard advice and why skills sections have inflated steadily for twenty years.
Generative AI finished the job. The cost of producing a CV that contains every term in your advert, phrased fluently, is now zero. Term presence has been fully commoditised as a signal. Anything downstream of it inherits that worthlessness.
The Evidence Hierarchy
The practical alternative is to treat a skill claim as a position on a scale rather than a binary. Six tiers cover almost everything you will encounter.
Tier One: Bare Mention
The term appears in a skills list, detached from any role, project or period. It is free to include and free to fabricate. Its evidential value is close to zero, and it should be scored accordingly, which means near-zero, not "a match".
Tier Two: Contextual Mention
The term appears within a role description, so it is at least anchored to an employer and a timeframe. Still cheap, but now constrained: it has to be consistent with the role it sits inside.
Tier Three: Applied Context
The candidate describes doing something specific with the capability. "Used Terraform to manage our staging environments" is a real statement about real activity. It can be exaggerated but it cannot be produced by pasting a keyword.
Tier Four: Outcome-Linked
The application connects to a result. "Moved environment provisioning to Terraform, cutting new environment setup from two days to under an hour." Now the claim carries a consequence, and consequences can be discussed.
Tier Five: Scale and Constraint
The description carries the dimensions of the problem: volume, concurrency, team size, regulatory limit, legacy obstacle, or the thing that did not work first time. "Managed provisioning across 40 environments for six teams, after the initial module structure proved unmaintainable and had to be reorganised around workspaces."
This tier is the hardest to fake convincingly, because it requires knowing how the work behaves in practice. Constraints and failures are the fingerprint of real experience, and they are exactly what generated text omits, because a model asked to make someone look good does not volunteer what went wrong.
Tier Six: Corroborated
The claim is internally consistent with the rest of the document: the seniority claimed matches the outcomes described, the timeline supports the depth asserted, and the capability appears consistently rather than once in a list. Inflated CVs frequently fail here in ways that are quick to spot once you look.
What This Changes in Practice
Three consequences follow, and all three show up in commercial outcomes rather than just in screening philosophy.
Ranking changes materially. A candidate with six genuinely evidenced capabilities outranks one with nineteen listed. That is the correct result and it is the opposite of what keyword matching produces. It also tends to surface candidates from adjacent backgrounds who described their work substantively, which widens a shortlist that title matching would have narrowed.
Gaming stops paying. Under evidence weighting, adding keywords achieves almost nothing. To score well a candidate would have to fabricate detailed, internally consistent, technically coherent project narratives with plausible constraints and failures. That is far more effort, much easier to dismantle in a fifteen-minute call, and beyond what a generic prompt produces.
Interviews get sharper. The screen can hand the interviewer the specific weak points: which required capabilities sit at tier one or two, and therefore which claims need testing. Instead of a general competency conversation, the interviewer opens with the two questions that actually resolve the uncertainty.
Implementing It Without Overcomplicating It
You do not need a large project to move in this direction. The sequence that works is roughly:
- Separate must-have capabilities from nice-to-haves, and be honest. Most job specs list twelve requirements where three determine success.
- Set a minimum evidence tier for the must-haves. If deep experience genuinely matters, a tier-one mention should not satisfy the requirement. State that explicitly.
- Stop scoring breadth. A long skills list should not accumulate points. If anything, an implausibly long list against a short career is a signal to look harder.
- Require the source passage behind every assessment, so a reviewer sees the evidence rather than a number, and so the record is defensible later.
- Generate probes from the gaps and hand them to whoever runs the first conversation.
Step three is the one teams resist, because a candidate with more listed skills feels stronger. They are not. They are a candidate who wrote a longer list.
The Objection Worth Taking Seriously
There is a real counterargument: evidence-based scoring can penalise candidates who are genuinely skilled but write terse, understated CVs. Some excellent engineers submit a page of bullet points and expect the technical interview to speak for them. Under evidence weighting, they score lower than they should.
That is true, and it needs handling honestly rather than being waved away. Three mitigations help. First, weight the must-haves only, so an understated CV that evidences the three things that matter is not punished for brevity elsewhere. Second, treat tier-one claims on critical requirements as unresolved rather than absent, which routes the candidate to a probe instead of to rejection. Third, monitor your own outcome data for patterns, because groups less coached in self-presentation may be affected differently, and that is worth knowing rather than assuming.
The honest position is that evidence weighting is substantially better than keyword matching and not perfect. It replaces a signal that has been fully gamed with one that is harder to game, at some cost to terse writers, mitigated by routing uncertainty to a conversation rather than to a rejection.
Frequently Asked Questions
What Is the Difference Between a Claimed Skill and an Evidenced Skill?
A claimed skill is a term the candidate has included, most often in a skills list, with nothing behind it. An evidenced skill is a capability supported by a description of applying it: what was done, in what context, with what result, at what scale. The first costs nothing to produce; the second requires having done the work.
Do Applicant Tracking Systems Distinguish Between the Two?
Most traditional parsing and keyword matching does not. It identifies term presence and optionally frequency, which means a skills list and a detailed project account register as equivalent matches. Distinguishing them requires assessing the context around the term rather than the term itself.
Does This Mean Candidates Should Remove Their Skills Section?
No, it is a useful index and some systems still rely on it. The advice that follows from evidence weighting is that a skills list should summarise capabilities the CV goes on to demonstrate, rather than standing in for demonstration. Candidates are better served describing three things properly than listing twenty.
How Many Evidence Tiers Do You Actually Need?
In practice the workable distinction is three bands: unsupported mention, applied with context, and applied with outcome and scale. The six-tier version is useful for designing a scoring model; reviewers work faster with three.
Does This Slow Screening Down?
It does if done by hand, which is why it is rarely done consistently at volume. Automated evidence extraction removes that cost: the system locates and classifies the supporting context, and the reviewer assesses structured evidence rather than searching for it. The judgement stays human; the reading does not.
Why This Is the Differentiator That Matters
Every screening tool claims accuracy. Very few will tell you what they score, and the answer usually reduces to term matching with additional statistics layered on top. In a market where any candidate can generate a term-perfect CV for free, that entire category of tooling is measuring something that no longer carries information.
CVSense is built around evidence-based skill validation: it evaluates the context, outcome and scale attached to a claim rather than counting occurrences, and it shows the source passage behind every assessment so a human can check the reasoning. If your shortlists keep producing candidates who match on paper and disappoint on the call, this distinction is almost certainly where the problem lives.
Sources
Chartered Institute of Personnel and Development.
https://www.cipd.org/uk/
Information Commissioner's Office. Guidance on AI and Data Protection.
https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/
National Careers Service. CV Sections and Advice.
https://nationalcareers.service.gov.uk/careers-advice/cv-sections
InsightCircle
