The use of AI in recruitment is one of those topics where the noise is currently much louder than the knowledge.
Candidates are being told to “beat the bot”. Employers are being told AI will fix hiring, and recruiters are being accused of outsourcing judgement to machines.
And somewhere in the middle of all this, a very real legal and ethical question is getting lost
“Can AI reject a job application without a human being involved?”
The answer is not as simple as “yes” or “no”. But it is important.
In the EU, there are legal frameworks that restrict fully automated decision-making in recruitment, particularly where that decision has a significant impact on a person’s employment prospects. In plain English, if a system automatically rejects a candidate without meaningful human involvement, that can create serious legal, ethical and reputational risk.
That does not mean all automation in recruitment is illegal. It does mean employers need to understand the difference between useful automation, lawful screening, poor process design, and decisions that should never be left entirely to an algorithm.
Why this matters now
Recruitment has always involved filtering. Long before generative AI, recruiters were using keyword searches, Boolean strings, database matching, job board filters and applicant tracking systems.
Keywords have been part of recruitment for decades. Search has been part of recruitment for decades. Sourcing came into action in the 1990s, and by the early 2000s Boolean search was already being used as part of digital transformation in recruitment. So this is not really a brand-new debate.
What has changed is the scale, speed and opacity of the technology. AI tools can now screen CVs, rank candidates, summarise applications, assess video interviews, infer skills, compare profiles against job descriptions, and recommend who should move forward. Used well, this could help recruiters and hiring managers manage volume and spend more time doing the human parts of hiring.
Used badly, it could quietly exclude people without anyone understanding why. That is where the law starts to matter.
GDPR Article 22 and automated rejection
Under the General Data Protection Regulation, Article 22 gives individuals the right not to be subject to a decision based solely on automated processing, including profiling, where that decision produces legal effects or similarly significant effects.
In recruitment, being rejected from a job may well be a similarly significant effect.
Work matters. Income matters. Career progression matters. Being excluded from an opportunity is not a trivial outcome.
The key phrase is “solely automated”.
If an AI system screens an application and automatically rejects the candidate without a human meaningfully reviewing the result, that is where Article 22 becomes relevant.
There are exceptions – Solely automated decisions may be allowed in limited circumstances, such as where they are necessary for entering into or performing a contract, authorised by law, or based on explicit consent. But these exceptions are not a free pass.
Even where automated decision-making is permitted, candidates may still have rights to obtain human intervention, express their point of view, and challenge the decision. Employers also need to think about transparency, fairness, data protection, bias, accuracy and accountability.
Consent is especially tricky in recruitment. A candidate may technically tick a box, but if the alternative is not being considered for a role, how freely was that consent really given?
That is why relying on consent alone is rarely the strongest foundation for automated hiring decisions.
The EU AI Act and high-risk recruitment systems
The EU AI Act classifies certain AI systems used in employment, recruitment and worker management as high-risk. This includes AI used to analyse, filter or evaluate job applications. High-risk does not mean banned, it means governed.
Employers and technology providers using AI in recruitment need to think about risk management, data quality, transparency, record-keeping, accuracy, bias controls and human oversight. Human oversight is not meant to be a decorative feature.
It should mean that a competent person has the authority, training and practical ability to understand, monitor, question, intervene in, and where necessary override the system.
A human rubber-stamping an AI output they do not understand is not meaningful oversight. A recruiter clicking “approve” on a rejection list generated by AI, without reviewing the actual candidates, is not meaningful judgement.
A hiring manager relying on a score without knowing how it was produced is not robust decision-making. The human has to matter.
The difference between screening and rejection
This is where recruitment needs a more grown-up conversation. Not every filter is the same.
There is a difference between a simple knockout question and an algorithmic assessment of a person’s suitability.
For example, a role may genuinely require the legal right to work in the UK without sponsorship. If the employer cannot sponsor, asking “Do you now or in the future require sponsorship to work in the UK?” is a practical eligibility question.
Likewise, some roles may have genuine non-negotiables: a specific licence, a mandatory qualification, a location requirement, or the ability to work a particular shift pattern.
These questions can help stop non-viable applications from overwhelming the process and taking oxygen away from suitable candidates. In a high-volume market, that can actually support better human recruitment, because it allows hiring managers to spend more time with people who may realistically be right for the role.
But knockout questions can also be abused. Sometimes through poor intent. Sometimes through poor understanding. Sometimes through lazy configuration.
“Do you have five years’ experience in a technology that has only existed for three?” is not intelligent screening. It is bad recruitment with a form attached.
So the question is not simply, “Was automation used?”
The better questions are:
- Was the question genuinely relevant?
- Was it proportionate?
- Was it transparent?
- Could it create indirect discrimination?
- Was there a route for correction or review?
- Did a human remain responsible for the outcome?
De facto automated rejection
There is another issue that receives far less attention: “de facto automated rejection”. This happens when a candidate is not formally rejected by AI, but in practice is never meaningfully reviewed by a human being.
That might happen because an applicant tracking system ranks them lower. It might happen because a platform makes their application less visible. It might happen because the hiring manager receives 300 applications overnight and only reviews the first 50. It might happen because automated scoring, keyword matching or workflow design quietly pushes some candidates out of sight.
Legally, this may not always be the same as a fully automated rejection decision. Experientially, it can feel identical.
The candidate receives silence. Or a rejection. Or no genuine consideration at all.
That is why this debate cannot only be about whether a piece of software sent the rejection email. It has to be about whether the hiring process gives people fair, meaningful and proportionate consideration.
Why “beat the bot” is the wrong advice
Candidates are often told they need to “get past the ATS” or “beat the recruitment algorithm”.
I understand why. When people feel shut out of hiring processes, they look for an explanation. The bot becomes the villain, however, sometimes technology is part of the problem.
But “beat the bot” is poor advice because it encourages candidates to write their CV for a machine rather than communicate with a human.
It leads to keyword stuffing, unnatural CVs, inflated claims, generic applications and content that technically matches a job description but says very little about the person behind it.
The better approach is what I would call human compliance. Make it easy for a human being to understand why you are relevant.
That means using the right keywords, yes. But not because you are trying to trick a system. Keywords matter because they are the language of your profession, your skills and your evidence.
Human compliance means:
- Your CV clearly reflects the role you are applying for.
- Your LinkedIn profile is searchable and easy to understand.
- Your achievements are evidenced, not just asserted.
- Your application explains fit quickly.
- Your career story makes sense.
- Your skills are described in language recruiters and hiring managers actually use.
- Your networking, follow-up and interview approach all reinforce the same message.
In other words, make yourself discoverable, understandable and credible. That works whether the first search is done by a recruiter, a hiring manager, an ATS or an AI-enabled sourcing tool.
What employers should be doing
Employers should not wait for UK case law to change, enforcement action or reputational damage before getting their house in order.
If AI or automation is being used in recruitment, employers should be asking some basic questions:
- Where exactly is automation being used?
- Is it sourcing, screening, ranking, shortlisting, rejecting, assessing or communicating?
- Could the system materially affect whether a candidate progresses?
- Is the decision solely automated?
- Is the candidate told what is happening?
- What data is being used?
- Could the system disadvantage certain groups?
- Who is responsible for reviewing the output?
- Can the human reviewer genuinely challenge or override the system?
- Is there an audit trail?
- Can a candidate ask for human intervention or challenge a decision?
These are not just compliance questions. They are quality questions.
A hiring process that cannot explain how decisions are made is not just legally exposed. It is commercially weak.
Good recruitment depends on trust, candidates need to believe they have been considered properly and hiring managers need to trust the shortlist. Employers need to know they are not building bias, opacity or unfairness into the front door of their organisation.
What recruiters should be doing
Recruiters have a particular responsibility here. We sit between candidates, employers, technology and the labour market. We see the pressure from all sides: rising application volumes, candidate frustration, hiring manager expectations, skills shortages, automation tools and the demand to move quickly.
That means we should be careful not to reduce this debate to “AI good” or “AI bad”. AI can be useful. Automation can be useful. Filtering can be useful. But none of it replaces professional human judgement.
The best recruiters will not be the ones who use the most automation. They will be the ones who know where automation helps, where it harms, and where human judgement is non-negotiable.
AI should free recruiters to be more human, not less. It should take away some of the administrative weight so recruiters can do more of the work that actually matters: understanding people, challenging assumptions, advising clients, opening doors, spotting potential, and treating candidates with dignity.
The real issue is not the bot
Recruitment is a challenge right now. But it’s not because of AI alone. It’s also human-driven.
- Poor job design is human.
- Unrealistic wish lists are human.
- Slow feedback is human.
- Vague criteria are human.
- Ghosting is human.
- Overloaded recruiters are human.
- Hiring managers changing their minds halfway through a process is very, very human.
AI may amplify the problems. It may speed them up. It may hide them behind a shiny interface. But it did not invent them.
That is why the answer is not panic. It is better knowledge, better design and better accountability.
Final thought
The law around AI candidate screening is still developing, but the direction of travel is clear.
If technology is being used to influence who gets seen, who gets shortlisted, who gets rejected or who gets hired, employers need to be able to explain it, govern it and keep meaningful human judgement in the process.
For candidates, the answer is not to become robotic in order to beat the robots. The answer is to become clearer, more relevant, more discoverable and more human.
Because good recruitment has never really been about matching words on a page. It is about understanding people.
And if AI is going to have a useful place in hiring, it should help us do more of that, not less.
If you are looking for work across the North and East Yorkshire region, get in touch here.
If you are looking to hire, you can click here to see how we can support you.