Responsible AI
Audit AI recruitment screening with an evidence-led review
A screening score is easier to display than to evaluate. Start an audit by asking what the tool is intended to do, what evidence would show it doing that task well and what a consequential mistake would look like. Keep review criteria visible so an apparently neat ranking does not become the only definition of quality.
Freeze the criteria and assemble a review sample
Save the vacancy specification and the version of the screening configuration. Use a suitably governed sample that includes short CVs, unfamiliar qualification descriptions and varied career paths. Ask reviewers to identify relevant evidence before showing them a model’s ranking. This creates a comparison point and makes disagreements about the role distinguishable from errors introduced by the tool.
Classify errors before adjusting the system
Separate missed evidence, invented evidence, extraction failures, inappropriate requirements and reviewer disagreement. A different prompt will not fix an unreadable document or an unfair person specification. Record severity and potential applicant impact, then identify a concrete corrective action. Avoid presenting a single overall accuracy figure without explaining the sample, the task and the types of mistake it hides.
Examine fairness without inventing personal characteristics
The ICO’s recruitment audit findings describe concerns including inappropriate inference of protected characteristics. Do not guess someone’s ethnicity, gender or disability from a name or CV to create a fairness report. Work with appropriate specialists on what information can be used and how. Record the limitations of the sample, and do not claim that removing names proves the entire process is free from bias.
Make challenges and corrective action visible
Give reviewers a way to override an unsupported assessment and applicants an appropriate route to raise concerns. Link an issue to the affected vacancy and document version. Assign an owner and a due date, and check whether similar applications were affected. A useful audit ends with changes, a retest and a record of remaining uncertainty rather than a badge claiming the system is bias-free.
Sources and further reading
Sources checked 25 October 2026. The workflow recommendations are MyValidCV editorial guidance. Check current requirements with the relevant authority; this article does not establish legal compliance or certify a product.
Explore a human-reviewed CV workflow
MyValidCV Enterprise supports bulk CV-to-job comparison, evidence review and applicant email drafts for review before sending. A comparison score is advisory; your team owns hiring decisions. The operational practices described here may require separate systems and policies.
Connect your HR software
The MyValidCV API pilot brings CV-to-job comparison into your own recruitment workflow, with owner-approved company access, capped usage and human review. Request a company pilot.