Graduates are using AI to write CVs, cover letters and application answers, and a 2026 survey of 155 large UK graduate employers found they are becoming “more accepting” of it at some stages of hiring. The line is still blurry: AI help with polish is tolerated, AI that misrepresents a candidate is not. Meanwhile, a Justice Department warning about North Korean remote IT workers shows why identity and authenticity checks matter more as AI-written applications become normal. HR teams should publish a clear AI-use policy for candidates this quarter, stop treating polished prose as a signal of effort, and verify skills and identity later in the process instead.
This week the BBC asked a question that nearly every recruiter now faces: is it OK for jobseekers or employers to use AI? The report followed graduates who say they have little choice but to use the technology, and employers who say they are relaxed about it, up to a point. At the same time, US prosecutors announced a sweep against North Korean operatives who used stolen identities to get hired as remote IT workers. The two stories look unrelated. For HR leaders they describe the same problem from opposite ends: when text, voices and even faces can be generated cheaply, the old signals that told you who a candidate was and how hard they tried no longer work.
What the BBC reported about AI and job applications
The BBC piece, from its October 2026 business coverage opens with a simple scenario: would you use AI to help write a CV or cover letter, or during an online assessment or video interview? Graduates and recruiters interviewed said the technology is increasingly part of the application process, and many employers are happy for candidates to use it “within reason.” The harder question is where reasonable ends and cheating begins.
The story’s central voice is Tom, 22, from Maidstone, who graduated last year with a geography degree. He says he applied for more than 200 jobs before landing a role as a restaurant manager last month. The process, he said, was “exhausting” and “demoralising,” with employers often not replying at all. He used AI to help with cover letters and CVs and to find the right “buzz words” that early-stage screening would pick up, but says he never used it in an assessment or interview. His argument is the one many candidates make: “Employers want you to put your absolute best into the application, but if I’m applying for 10 to 20 jobs in a day, the reality is you can’t do that.”
He also called it “hypocritical” for companies to ask candidates not to use AI when some use AI to screen applicants. In his words, it is a “weird tightrope”: use AI too much and you are not human, but if you do not write like an AI you may be skipped by the AI screening.
The employer side of the picture comes from the Institute of Student Employers (ISE). This summer it surveyed 155 of the UK’s biggest graduate employers and found they were becoming “more accepting” of candidates using AI at some stages of recruitment. The most commonly accepted use was writing cover letters or CVs, followed by completing online application questions. The BBC also reported research suggesting that graduates are increasingly using AI to misrepresent themselves in applications, which is the part that worries employers.
- Acceptance is rising, unevenly. ISE found large UK graduate employers more accepting of AI at some stages, mainly CVs, cover letters and online application questions.
- Volume drives behaviour. Candidates sending 10 to 20 applications a day cannot hand-craft each one, and employers’ own screening tools reward keyword matching.
- Misrepresentation is the real line. Polishing is different from inventing experience or having AI take an assessment.
- Authenticity risk is growing. The FBI says “thousands” of North Korean state-controlled IT workers are employed at Western companies, many hired through fabricated identities.
What the trend data says, and what it does not
We checked Google’s trending searches for the United States on October 14, 2026. The list was dominated by entertainment and sport, with Hugh Jackman, Lindsay Lohan, tennis results and a Star Wars game among the leaders. The most visible business story was ASML, whose search interest topped 1,000 queries after the chipmaker reported third-quarter 2026 net sales of €11.7 billion and net income of €5.0 billion, and said it was seeing order interest through 2030. There was no breakout search term for “AI job applications” itself.
That absence is informative. The AI-and-hiring story is not a spike driven by one headline; it is a slow structural shift. What the ASML result tells us is that investor and employer confidence in AI as a long-run investment is high, which means the tools your candidates use will keep getting better, cheaper and more embedded in every stage of work. HR policy written around today’s tools will age quickly, so it should be written around principles instead.
Why the old signals no longer work
For decades a well-written cover letter signalled several things at once: the candidate could write, cared enough to try, and understood the role. Generative AI separates those signals. A fluent letter now says almost nothing about writing ability or effort, and only a little about understanding, because the model can echo the job description back convincingly. Tom’s experience shows the other side: when employers screen with software, candidates optimise for the software, and the resulting documents converge on the same phrases.
Three consequences follow for HR teams.
- Written materials become weaker evidence. If you still shortlist on the quality of a cover letter, you are mostly selecting for access to good tools and time, not for ability.
- Volume rises. If applying takes minutes, applicants send more, and recruiters see more near-identical submissions. That pressures teams to automate screening, which in turn pushes candidates to optimise for the machine. It is a loop, and breaking it needs a deliberate design choice.
- Fairness questions get sharper. Candidates who use paid or advanced tools may look stronger on paper than equally capable candidates who do not. Blanket bans are difficult to enforce and tend to penalise honest applicants.
The second story: when the candidate is not who they say
On October 13, 2026, the US Justice Department said it had indicted and arrested at least two people and searched dozens of suspected “laptop farms” in Texas, according to TechCrunch. Authorities seized more than 180 computers that US companies had shipped to newly hired employees who were in fact sanctioned North Korean workers using stolen identities of US citizens. Prosecutors said more than 40 laptop farms were raided, run from homes owned by at least two American facilitators, and that at least 70 US companies across healthcare, cyber support, retail, communications and manufacturing had been infiltrated, with around $3 million in stolen funds linked to the scheme. CrowdStrike, which assisted the FBI, called laptop farms “vital” to North Korea’s operations.
The Justice Department added that North Korean IT workers are increasingly recruiting American facilitators to counter improved private-sector detection, and in some cases subcontracting entire jobs to third parties. In a separate public service announcement the FBI said “thousands” of state-controlled workers are employed at Western companies and pose an insider-threat risk for national security and intellectual property theft. The cover included realistic email addresses, spoofed LinkedIn profiles and extensive social-media histories.
The link to AI-written applications is not that candidates who use ChatGPT are spies. It is that both problems exploit the same weakness: hiring processes that trust documents and profiles. A convincing résumé, a convincing LinkedIn page and a convincing written assessment can all be produced by someone who does not exist as described. Everyday AI use by honest candidates makes the noise louder and the genuine fraud harder to spot.
A practical framework for HR
1. Write and publish a candidate AI-use policy
Most employers have no public stance, which leaves candidates guessing. A short policy removes the “weird tightrope” Tom described. A reasonable structure: allowed (spelling, grammar, structure, tailoring a CV to a job description, practising interview questions); allowed with disclosure (drafting answers to open-ended application questions); not allowed (using AI during live assessments, interviews or proctored tests; inventing qualifications, employers or achievements; impersonation of any kind). Put the policy on the careers page and link it from the application form.
2. Move verification later and make it human
If written materials are weak evidence, give them less weight early and test real ability closer to the decision. Short live work samples, structured interviews with follow-up questions about the candidate’s own examples, and paid trial tasks are harder to fake and fairer than trying to detect AI text, which detectors do unreliably. The aim is not to catch AI use, it is to find out whether the person can do the job.
3. Treat identity checks as a security control, not an HR courtesy
The Justice Department case involved company laptops shipped to addresses that were really laptop farms. Basic controls help: verify identity with documents and a live video check before the offer; require an in-person or notarised step for roles with privileged access; ship equipment only to the verified home address and flag mismatches or last-minute changes; and watch for remote workers whose activity patterns, time zones or payment details do not match their stated location. Involve security and IT early, because HR alone cannot see these signals.
4. Audit your own AI screening
Candidates call out the hypocrisy when employers use AI to screen and forbid them from using it. If you use automated screening, document what it does, test it for bias on a regular schedule, and tell applicants. Tools that reward keyword stuffing encourage precisely the behaviour you may later complain about. Where the law requires notice or human review of automated decisions in your jurisdiction, confirm with counsel; requirements vary and are changing.
5. Train recruiters and hiring managers
Give interviewers a short guide: what AI-assisted answers tend to look like, how to probe depth with “walk me through the decision you made when…” questions, and what to do if they suspect a candidate is not who they say. Make escalation easy so that a suspicion does not depend on one recruiter’s confidence.
Bottom line
AI in job applications is not going away, and policing it with blunt bans is likely to fail. The more durable response is to decide what you actually need to know about a candidate, ask for evidence that is hard to fake, and verify identity where the stakes are high. The employers best placed this year will be the ones that say clearly what is allowed, test real skills, and treat hiring integrity as shared work between HR, security and the business.
For related reading on workforce compliance, see our note on what HR should take from this week’s rulings.
Sources
- BBC News, “Is it OK for jobseekers or employers to use AI?” (October 2026)
- TechCrunch, “Justice Department says North Koreans have infiltrated dozens of U.S. companies” (October 13, 2026)
- Google Trends, US trending searches, October 14, 2026; CNBC and Yahoo Finance coverage of ASML’s Q3 2026 results as surfaced in Trends
This article is general information, not legal advice. Check hiring, data-protection and automated-decision rules in your jurisdiction.