News and views

AI-written job applications: how to spot them and design questions that still work

Written by Laura Chambers | 16 Nov 2026, 08:30:00

AI-generated job applications are no longer the exception. Around nine in ten UK professionals now use AI somewhere in their job search and three in four UK recruiters believe most of the CVs they see were written by AI. For hiring teams across the NHS, care, local government and beyond, the result is a flood of polished, similar-sounding applications that tell you very little about the people behind them.

The good news is that this is a fixable problem. In this blog, we look at how to spot AI-written applications, why trying to catch candidates out is the wrong battle to choose and how to design application questions that work in 2026.

How many candidates use AI to apply for jobs?

Most candidates now use AI somewhere in their application. Around 65% of job seekers used AI in 2025 and by 2026 only 10% of UK professionals said they don't use it at all. For recruiters, that means more applications that look polished but reveal very little about real ability.

The impact on hiring teams is real:

  • 67% of UK hiring managers say screening for AI-generated material has increased their time to hire (Robert Half, 2026)
  • 27% struggle to tell AI-generated CVs from authentic ones (Robert Half, 2026)
  • 75% of UK recruiters believe most CVs and profiles they see are written by AI (TestGorilla, reported by People Management, November 2026)
  • 73% of recruiters spend an hour or more on a single candidate before trusting they can do the job (TestGorilla, 2026)
  • 92% of graduate employers suspected or confirmed cheating in recruitment this year, a record high, with undeclared AI use in interviews (66%) and online assessments (47%) the most common forms (Institute of Student Employers, October 2026)

Why are open text boxes hit the hardest?

Put yourself in the candidate's shoes. Faced with a question and a large empty text box, the easiest thing for them to do is paste the question into ChatGPT or Claude and copy the answer back in. Multiply that across hundreds of applicants and every personal statement starts to read the same.

This creates three problems for employers:

  1. No signal of which candidate is better. If every answer is fluent and well structured, writing quality no longer tells you anything about ability.
  2. Wasted time. Recruiters spend hours reading statements that all say similar things, and even more time verifying claims at interviews.
  3. Unfair comparisons. Personal statements have always been difficult to score consistently. Now it's even harder to make confident, fair decisions between candidates because of AI.

How can you spot an AI-generated job application?

You can often spot an AI-generated job application by its generic answers, unsupported claims, wording copied from your job advert and near-identical structure across applicants. But spotting them is not enough. AI detection tools are unreliable, so a better fix is redesigning the questions you ask, rather than trying to catch candidates out.

There's no single giveaway, but experienced recruiters tend to notice the same patterns:

  • Generic, could-be-anyone answers. The response fits any organisation and any role, with no mention of your services, values or the setting the job sits in.
  • Claims without evidence. Shallow phrases like "passionate about delivering excellence" or "a proven track record" with no names, numbers, dates or outcomes behind them.
  • Mirrored language. The answer repeats the wording of your job advert or person specification almost line by line.
  • Too polished and too uniform. Perfect structure, the same tidy three-part shape and similar phrasing across many different applicants.
  • A mismatch in tone. A highly formal statement alongside a CV, email or interview manner that is very different.
  • Surface-level detail. When you ask a follow-up question during the interview, the candidate can't expand on the example they gave.

Can AI detectors spot AI-written applications?

It's tempting to reach for an AI detection tool. Be cautious though. A 2025 review by Jisc's National Centre for AI found that once AI text had been paraphrased or edited, even the leading detector caught only between 7.9% and 30% of it in the studies reviewed. Jisc concludes that decisions "should never be based solely on AI detection, whether by a tool or a person."

For UK employers, that's a fairness risk as well as an accuracy one. Rejecting candidates because a tool guessed their writing was machine-made could disadvantage people on the basis of language or background, which sits uncomfortably with the Equality Act 2010.

There is a bigger point too. Using AI to polish a CV is not cheating; many employers now accept it. The real question is not "did a machine help write this?" but rather "does this application show me whether the person can do the job?" If your form questions can't answer that, the problem lies there.

Which application questions still work when candidates use AI?

The application questions that still work are structured, specific and hard to paste into a chatbot. Replace open text boxes with rating, forced-choice and ranking questions based on real scenarios from the role. Then add time limits and randomised question banks, so answers reflect judgement and experience rather than fluent writing.

The most effective response to AI-written applications is not to ban AI. Trying to stop it is a bit like King Canute trying to hold back the tide. Instead, design questions that reward judgement and real-world experience and that a chatbot can't answer well on a candidate's behalf.

1. Swap open text boxes for structured questions

Structured question types are far harder to "AI game" and much easier to score fairly:

  • Rating questions. Ask candidates to rate their confidence or approach on a 1 to 5 or 1 to 10 scale, with clear labels so expectations are consistent.
  • Multiple choice with required picks. Ask candidates to "choose three" from a list, forcing trade-offs that reveal priorities.
  • Order of importance. Ask candidates to rank options by dragging them into order, showing how they think.
  • Situational judgement. Present a realistic scenario from the role and ask which response they would choose.

2. Make questions specific to your organisation and role

Generic questions get generic answers. Anchor every question in your own values, services and the realities of the job.

 Instead of this

 Try this

Tell us why you want to work here.

A resident tells you they don't want their care plan followed today. From the options below, which would you do first?

Describe your communication skills.

Rank these four ways of updating a family member about a change in a patient's care, from most to least appropriate.

Give an example of teamwork.

Your shift is short-staffed and two urgent tasks land at once. Choose the two actions you would take.

Why are you suitable for this role?

Rate how confident you are in each of these five tasks from the job description, from 1 (not yet) to 5 (fully confident).

These examples come from health and care, but the same approach works for local government, education, housing and charity roles.

3. Add time limits and randomisation

  • Timed questions encourage candidates to answer in the moment, rather than having time to consult and copy questions into AI.
  • Randomised question banks mean candidates applying for the same role see different questions, so answers can't simply be shared or reused.
  • Branching questions guide applicants down different routes based on their earlier answers, making the form harder to predict.

4. Keep it short and purposeful

Shorter, relevant forms feel fairer to candidates and improve completion rates too. If a question doesn't help you decide who to shortlist, remove it.

What else can employers do about AI-written applications?

Beyond the application form, employers should publish clear guidance on acceptable AI use, probe examples at interview and use short practical tasks. Being open about your own use of AI and keeping people in charge of every hiring decision, protects both fairness and candidate trust.

  • Be clear about acceptable AI use. Tell candidates where AI help is fine (for example, tidying a CV) and where it's not (assessments and interviews). The Institute of Student Employers found that 82% of graduate employers now offer this guidance, up from 55% a year earlier.
  • Probe at interview. Ask follow-up questions about decisions the candidate made personally and problems they faced. Real experience holds up under questioning, whereas borrowed answers usually don't.
  • Use short, practical tasks. A task that reflects the actual job tells you far more than a written statement.
  • Be transparent about your own AI use. Under the UK Government's AI principles, candidates should understand how AI is used in decisions that affect them and how to query an outcome.
  • Keep humans in charge. AI can support shortlisting, but people should make every hiring decision.

How does Jobtrain help employers manage AI-generated applications?

Jobtrain's applicant tracking system helps employers replace personal statements with structured, scored assessment forms. Rating, ranking, timed and randomised questions are hard to game with AI and every score is transparent and explainable. Employers across the NHS, care, local government, education, housing and charities use it to shortlist faster and more fairly.

We hear the same question from employers again and again: what is our applicant tracking system doing to catch AI applications? Our answer is simple. The issue isn't candidates using AI; it's application processes that still rely on personal statements and empty text boxes.

That's why Jobtrain is built to help you move from personal statements to structured, scored assessments aligned to your organisation's values, skills and role-specific behaviours.

Structured assessment forms that are hard to AI game

Jobtrain's built-in assessment form builder supports rating scales, multiple choice with required picks, ranking by order of importance, branching questions, randomised question banks and timed questions. Together, they reward judgement and evidence rather than fluent writing.

Transparent scoring that you can explain

Each answer option carries a weighting that your team sets and approves before any candidate applies. Responses are scored and ranked automatically and every score can be explained at question level, for example: "Candidate A chose option B on question 3, weighted 8 out of 10 for this role." That supports UK GDPR obligations and gives candidates a clear route to query their outcome.

Crucially, Jobtrain does not use AI to read free-text answers and produce opaque scores. Large language models tend to score AI-written text highly, so this kind of "black box" scoring can end up rewarding the very applications you are trying to see past, while penalising authentic answers written differently.

AI that helps you build better questions

Jobtrain uses JobBrain AI where it saves you time without taking over decisions:

  • Assessment question suggestions generated from your job description so you can build stronger forms faster
  • On-brand image generation for image-based assessment questions, without leaving the platform
  • Bias-checked job adverts that reduce unintentionally exclusionary language

Every question is reviewed and approved by your recruiters and people make every hiring decision.

Faster, fairer shortlisting

Because responses are scored the moment a candidate applies, your team can instantly see the strongest applicants and spend 80% of their time engaging the top 20%. No more wading through near-identical statements and no more wondering whether that polished paragraph was written by a machine.

Use AI as a catalyst for change

AI-assisted applications aren't going away, and that doesn't have to be a bad thing. They expose what was already true: personal statements were never a reliable or fair way to compare candidates. With the right questions and the right tools, you can make your recruitment faster, fairer and far more effective for candidates, recruiters and hiring managers alike.

Ready to rethink your application forms? Book a demo to see how Jobtrain's structured assessment forms and transparent scoring can help you see the real person behind every application.

You can also download our guides, When candidates use AI to apply and How to responsibly harness AI in recruitment.

Related reading:
AI in recruitment: 6 things every HR leader needs to know in 2026

Frequently asked questions

What is an AI-generated job application?

An AI-generated job application is a CV, cover letter or application answer written wholly or partly by a generative AI tool such as ChatGPT. Candidates often paste an employer's question into the tool and copy the answer back, which produces polished but generic responses.

Can AI detectors reliably spot AI-written applications?

No. Research from Stanford University found that seven widely used AI detectors wrongly flagged an average of 61.3% of essays by non-native English speakers as AI-generated. Relying on detectors risks unfair rejections, so redesigning application questions is a more reliable approach.

Should employers ban candidates from using AI?

Banning AI is hard to enforce and often unnecessary. Many employers now accept AI help with CVs and cover letters, but draw the line at assessments and interviews. Publishing clear guidance on acceptable use, and designing questions AI cannot answer well, works better than a ban.

Which application questions are hardest to answer with AI?

Structured questions are hardest to answer with AI. These include rating scales, multiple choice with required picks, ranking tasks and situational judgement questions based on the role. Adding time limits, randomised question banks and branching makes them harder still.

How can an applicant tracking system help with AI-generated applications?

An applicant tracking system such as Jobtrain lets you replace open text boxes with structured, scored assessment forms. Responses are scored automatically against criteria your team approves, so you can shortlist quickly, explain every score and focus on candidates' real ability.