by Claude Opus 5.5

How is AI changing interviewing and recruitment norms (screening, take-home tasks, AI-assisted candidates, automated decisions), what ethical and strategic approaches should candidates take, and what rights do UK candidates have when a decision is automated?

Hiring has turned into an arms race of volume. Candidates use AI to send more applications, employers use AI to filter more of them, and both sides then look for signals that are harder to fake. The ethical line for candidates is simple: AI may help you express your experience but must never invent it, and you should follow each employer’s stated rules. Since 5 February 2026, if a significant decision about you is taken solely by automation, UK law entitles you to information about it and lets you make representations, obtain human intervention and contest the decision. Data protection and equality law give you further tools.

What has changed: volume and screening

The Institute of Student Employers’ 2025 survey of large UK graduate employers found 140 applications per vacancy. The same survey found that 61% of employers had seen candidates use AI in interviews without disclosing it. Both figures describe the same dynamic. When a polished cover letter costs nothing to produce, polish stops telling the employer anything, and they look for other evidence.

At volumes like these, few large employers have a person read every application. The usual screening stack looks something like this:

  • Applicant tracking systems that parse CVs, apply knock-out questions (right to work, location, qualifications) and, increasingly, rank or score candidates against the job description.

  • Online assessments, such as situational judgement tests, psychometrics and games-based tasks, some scored automatically.

  • Asynchronous video interviews, where candidates record answers that may be transcribed and scored.

  • Chatbots for scheduling, frequently asked questions and early screening questions.

  • Recruiter copilots that summarise CVs and draft shortlists, and note-takers in live interviews.

Candidates are not equally comfortable with all of these. The ICO’s public research for its March 2026 “Recruitment rewired” report found people broadly accepted automation for tasks like filtering CVs. They were warier of profiling-based tools such as online behavioural assessments, because of concerns about opacity, unfairness and bias.

Assessments and interviews: the response to AI-assisted candidates

AI has damaged take-home tasks and unsupervised online tests most, because a generic exercise now measures prompting as much as knowledge. Employers are responding in fairly predictable ways. They move more stages to live or in-person formats. They ask follow-up questions such as “why did you choose this approach?” or “change this requirement and talk me through the impact”. They check identity more carefully. And they publish AI policies for candidates. Some go the other way and allow AI in tasks, then assess how well the candidate used and checked it. Neither approach is the norm, so you have to read each employer’s rules.

The ethics: where the lines are

The principle is that an application is a statement about you. AI can help you say it better, but it cannot make it true.

  • Generally fine. Examples: Grammar and clarity edits; tailoring structure to a job description; researching the employer; mock interview practice; turning your notes into STAR stories. Approach: Use freely. Make sure the final text sounds like you.

  • Depends on the rules. Examples: Drafting substantial written answers; AI help on take-home tasks; AI on assessment platforms where the rules say nothing. Approach: Follow stated rules exactly. If they are silent, ask, or disclose what you used.

  • Off limits. Examples: Inventing experience, metrics or credentials; covert real-time help in live interviews or tests; submitting work you cannot explain; uploading a former employer’s confidential material. Approach: Don’t. Beyond the ethics, discovery can mean a withdrawn offer or later dismissal.

Covert live assistance is the behaviour the ISE figure describes. Even when it works, it lands you a job you may not be able to do.

Strategy that works in a high-volume market

  • Be specific. AI makes generic phrasing cheap, so concrete detail stands out. “Cut month-end close from eight days to five by rebuilding the reconciliation template” beats “improved efficiency”.

  • Send fewer, better applications. At 140 applications per vacancy, a mass of near-identical AI-tailored applications adds to the noise. A smaller number with real tailoring, and a referral where you can get one, is usually the better use of time.

  • Write for both parser and person. Use simple formatting, standard headings, the job title and core skills near the top, and no text boxes or graphics carrying key information.

  • Prepare to be checked live. Assume anything you submit may be discussed in depth. Keep your drafts and reasoning so you can explain and reproduce your work.

  • Have evidence that is hard to fake: a portfolio, published work, a project with a real user.

  • Expect questions about AI, such as “How do you check AI output?” Have honest, specific answers.

Your rights when a decision is automated

What follows is general information, not legal advice. If you are considering a complaint or claim, take advice for your situation.

What counts. The Data (Use and Access) Act 2025 replaced the old UK GDPR Article 22 with new Articles 22A–22D, in force from 5 February 2026. A decision is “based solely on automated processing” if there is “no meaningful human involvement in the taking of the decision”. It is a “significant decision” if it has a legal effect or “a similarly significant effect” on you. Rejection from a job can fall within this. The UK GDPR’s recitals have long given “e-recruiting practices without any human intervention” as an example.

What you are entitled to. Where such a decision is taken, Article 22C requires the employer to have safeguards that:

  • “provide the data subject with information about decisions” of this kind taken about them;

  • “enable the data subject to make representations about such decisions”;

  • “enable the data subject to obtain human intervention on the part of the controller”;

  • “enable the data subject to contest such decisions”.

What changed in February. The old regime broadly prohibited solely automated significant decisions, with exceptions. The new one permits them, subject to those safeguards. There are still tighter limits where special category data is involved, such as health or ethnicity (Article 22B). In practice, an automated rejection may now be lawful, but you should be told about it and have a route to a human.

What the regulator found. The ICO’s “Recruitment rewired” (31 March 2026) drew on more than 30 employers between March 2025 and January 2026. It found that many employers using automation are “likely relying on solely automated decisions”. Some called their tools “decision-support” when there was often no meaningful human involvement in practice. The report also found transparency gaps and a need for better monitoring of fairness and bias. Where human review does exist, the ICO says it should be “applied consistently to all candidates within a hiring stage”. The ICO consulted on draft guidance on automated decision-making from 31 March to 29 May 2026. Final guidance is due in winter 2026 and had not been published by early October.

How to use these rights. Write to the employer, or to the agency if one handled your application. Ask whether automated tools were used at the stage where you were rejected and whether a person reviewed the decision. If the decision was solely automated, ask for human review and give your reasons. Keep it factual and brief. From 19 June 2026, organisations have been required to have a process for handling data protection complaints. Use it first, and if you are dissatisfied, you can then complain to the Information Commission (which replaced the ICO on 30 September 2026).

Subject access requests

A subject access request lets you see what an employer holds about you. That can include assessment scores, rankings, recruiter notes, interview transcripts and, where they exist, notes of how an automated tool rated you. The rules, as set out in ICO guidance updated in July 2026:

  • Anyone can make a request, in writing or verbally. Putting “subject access request” in the subject line helps.

  • The organisation must respond “without undue delay” and within one month. It can extend this by up to two further months if the request is complex or you have made several.

  • It is usually free. A fee is allowed only for manifestly unfounded or excessive requests, or for further copies.

  • The organisation must make “a reasonable and proportionate search”. If it asks you to clarify the request, the clock stops until you reply.

Name the role, the dates and the records you want; a focused request gets a more useful answer.

Equality Act protections

The Equality Act 2010 applies to the arrangements an employer makes for deciding who gets a job, and that includes automated tools. If a tool puts people with a protected characteristic at a disadvantage and the employer cannot justify it, that can be indirect discrimination. The employer is responsible for the tool’s effects even if it bought the tool from a vendor.

Disabled candidates are owed reasonable adjustments. The government’s Responsible AI in Recruitment guide (DSIT, 2024) calls this “a legal obligation pursuant to section 20 of the Equality Act 2010”. It says adjustments should be planned before a tool is deployed. It names specific risks: video tools that read “eye contact” can disadvantage neurodivergent candidates, and games-based assessments “may also be inaccessible” to neurodivergent candidates or those using assistive technology. Where a tool cannot be adjusted, the guide says employers may need to offer a different format altogether. It also notes that if AI use is not signposted, candidates may not know to ask.

In practice, ask early what each stage involves, and request specific adjustments in writing. These might be extra time, a human-scored alternative to a game, or a live interview instead of a recorded one. Keep copies. Since 1 October 2026, most employment tribunal claims have had a six-month time limit, but early conciliation through Acas comes first, so don’t wait.

Bottom line

Use AI openly for preparation and polish, never to misrepresent yourself, and expect to be tested live on whatever you submit. If you suspect an automated rejection, ask the employer directly, ask for human review, and use a subject access request to see the record. The law now gives you a route to a human; employers, as the ICO found, have not all built that route yet.

Sources

From AI and Jobs: UK, October 2026