Assessing AI-Generated Applications: Why CVs Alone Are No Longer Enough
Moritz SteinbachCo-Founder & CEO of Talentigo
Well-written cover letters and polished CVs reveal less than they once did about how much support was involved in producing them. This is not a reason to distrust applications. It is a reason to base selection more strongly on verifiable evidence of capability.
DGFP reports that AI is already frequently used in recruiting for job ads, interview preparation and communication. Applicants use it too. Documents therefore remain a useful starting point, but decisions need additional evidence.
This topic should be distinguished from AI candidate analysis: that is about pre-screening and profile matching from a large applicant pool. This article focuses on what happens next – when candidates are in conversation and their skills need to be verified.
From document to verifiable capability
- Use the CV to identify relevant situations, not to treat a skill as proven.
- Ask about specific decisions, trade-offs and outcomes.
- Use short, work-relevant tasks where proportionate.
- Score answers against a predefined rubric rather than overall impression.
This shifts attention from a polished document to evidence that can be discussed. It is fairer to candidates because they can show how they work, and more reliable for the hiring team because everyone evaluates the same information.
A lean assessment design
- Define three to five critical capabilities.
- Assign each one an interview question or work sample.
- Agree what strong, sufficient and weak evidence look like.
- Record reasons immediately after the interview.
- Compare candidates against the same rubric.
Use AI well – and decide as humans
AI can structure interview notes or make criteria easier to see. It should not become a black box claiming that a person is suitable. Final decisions require accountable human judgement, documented criteria and room for challenge.
Example: same CV, different evidence
Two applicants for a project manager role list “stakeholder management” and “budget responsibility.” In a structured interview, candidate A describes a concrete situation with conflict, decision and measurable outcome. Candidate B describes general tasks without numbers or scope boundaries. On paper both profiles look similar – the evidence in conversation does not. That is why a fixed scoring rubric before selection pays off.
Should I automatically filter out AI-generated applications?
Blanket distrust or automatic downgrading is risky. A better approach is a process that verifies skills in conversation or an appropriate work sample – regardless of how polished the CV reads.
Does every role need a work sample?
No. For many positions, structured behavioural questions and a clear scorecard are enough. Work samples make sense where the core task can be represented in a short, bounded form.
Where does AI fit sensibly in selection?
In structuring notes, preparing interview questions or making criteria visible – not as the sole authority for yes or no decisions.
How many capabilities should a scoring rubric include?
Usually three to five critical capabilities. More criteria dilute the assessment; fewer make differences between candidates harder to explain.
Conclusion
The right response to AI-assisted applications is not an arms race. It is a process that makes relevant skills visible to both candidates and decision-makers.
Improve recruiting where it matters?
If you want to know whether active sourcing and AI-assisted profile analysis make sense for your roles, book a no-obligation demo – we’ll focus on your specific case.