AI in recruitment and CV screening: Why HR should review how candidates are assessed

by Lilia Dangi, Brightmine Legal Editor

As candidates use AI to produce more polished applications and employers use AI to support screening and shortlisting, HR needs to reassess whether their recruitment processes still produce reliable evidence and fair outcomes.

Recruitment is entering a new phase in which AI can influence both sides of the hiring process. This doesn’t mean that recruitment has become artificial or that candidates who use AI are necessarily being dishonest in their applications. In many cases, AI is simply another tool candidates use to present their experience clearly. However, it does mean that CVs, covering letters and written application answers may provide less reliable standalone evidence of a candidate’s skills or suitability for the role.

For HR, the challenge is to ensure that recruitment processes continue to identify reliable evidence of a candidate’s potential, skills and experience when AI may have shaped both the application and the employer’s assessment of it.

The “dual AI loop” in recruitment

AI is now embedded in many parts of the recruitment process. On one side, candidates are using it to find roles, tailor CVs, draft covering letters, prepare answers and practise interviews. On the other side, employers are using AI to manage high application volumes and reduce repetitive administration by drafting or reviewing job descriptions, communicating with applicants and summarising applications. Some tools go further by screening, scoring, ranking or filtering candidates or recommending who should progress to the next stage of the recruitment process.

For HR, this creates a “dual AI loop”, where AI shapes the application a candidate submits and an AI-assisted process then assists with the assessment process.

This is particularly relevant to early careers, graduate, apprenticeship and high-volume customer-facing recruitment, where employers often process large numbers of applications with limited recruitment resource. If candidates use AI to produce more polished and tailored applications, the number of apparently strong applications may increase. At the same time, if employers use AI to screen or rank those applications, the process may become faster without necessarily becoming more reliable.

A polished written answer may show that a candidate can use AI effectively, but not necessarily that they have the judgment, technical knowledge, communication skills or experience required for the role. An automated ranking may help narrow a long list, but it may also make it harder to understand why particular candidates were deprioritised.

The challenge is therefore not simply to process a greater volume of applications, but to ensure that the assessment continues to identify relevant and reliable evidence.

Why HR needs to be involved

AI tools that influence who is screened in or out, how candidates are ranked or who is recommended for interview can significantly affect an individual’s opportunity to progress, even where a human makes the final decision.

Using AI to draft a job advert, organise applications or produce a summary may support the recruiter’s work without deciding who progresses. By contrast, an AI tool that scores CVs, ranks applicants, filters candidates out or recommends who should move to interview is much closer to decision-making.

In practice, this can affect a candidate’s opportunity to be considered and move through to the next round of recruitment, even if a human recruiter makes the final decision.

The Government’s Responsible AI in recruitment guidance recognises that, although AI-enabled recruitment tools may improve efficiency, scale and consistency, they can also perpetuate existing bias and contribute to discriminatory outcomes when candidates are screened, ranked or selected.

The practical implication is that employers must consider not only whether AI makes recruitment more efficient, but also how it affects the fairness and quality of recruitment decisions. Recruitment is not just an administrative exercise. It’s a decision-making process that should be fair, based on the criteria relevant to the role, accessible to candidates and supported by reasons the employer can evidence and explain.

HR should be able to explain what a tool does, what evidence it uses, how its outputs are checked and whether candidates can challenge or correct information that may affect them.

AI recruitment risks arise before the hiring decision

In November 2024, the Information Commission (ICO) published findings from audits of AI recruitment tools used for sourcing, screening and selection. The ICO found that some tools allowed recruiters to filter out candidates with certain protected characteristics. Other tools inferred characteristics such as gender and ethnicity from application data or candidates’ names, rather than using accurate information provided directly by candidates.

If a tool filters, ranks or deprioritises candidates using unreliable data, inferred characteristics or criteria that are not clearly linked to the role, some applicants may be excluded before their applications ever receive meaningful human review.

This can expose employers to legal, regulatory and practical risks. For example, a tool may produce discriminatory outcomes, the employer may struggle to explain or evidence its decisions if challenged and candidates may lose trust if they do not understand how their data was used. HR should therefore not assume that a provider’s assurance that a tool is “fair” or “objective” is enough.

Taken together, the ICO’s findings show why employers should address these risks before an AI tool affects which candidates progress, rather than waiting for a complaint, challenge or poor candidate experience to expose weaknesses in the process.

As part of its findings, the ICO identified questions that can help employers assess data protection risks and obtain evidence from providers before procuring AI recruitment tools, including:

• whether an appropriate data protection impact assessment has been completed and what risks and safeguards it identifies;
• which lawful basis applies to each way in which candidates’ personal data will be processed through the tool;
• how the respective data protection responsibilities of the employer and the provider are allocated and documented; and
• what information candidates receive about how the tool uses their personal data and affects the recruitment process.

HR should therefore seek evidence of:

• how the tool has been tested;
• what data and criteria it uses;
• how bias and accuracy are monitored;
• what information candidates receive; and
• how meaningful human review is built into decisions influenced by automated recommendations.

AI should prompt a review of recruitment design

Debate about AI in recruitment often focuses on whether it can make hiring faster or more consistent. Those benefits may be real, particularly for employers dealing with high applications volumes, but they are not automatic. AI does not remove the need for sound human recruitment judgment.

For HR, the key question is whether the organisation understands and can stand behind the recruitment process it has designed or adopted. If employers do not understand what an AI tool assesses, what data and criteria it uses and how its outputs affect who progresses, they will struggle to explain and justify the resulting decisions.

In that case, employers may also struggle to identify whether protected groups are being disadvantaged and to respond confidently if a candidate challenges the outcome.

Organisations should therefore treat the adoption of AI as a prompt to review the design of their recruitment processes, rather than considering only whether the technology can automate existing steps.