How do you screen job applications now that candidates use AI?
With AI, a polished cover letter takes a minute to write. So what does an application still tell you, and how do you screen without losing good people?
By Ken Pauw, Head Of Recruitment
According to Eurostat, 44.7 per cent of people aged 16 to 74 in the Netherlands used generative AI in the three months before the 2025 survey. Among unemployed people in the Netherlands it was 49.9 per cent. A polished cover letter now proves little. Screen for what AI cannot supply: hard requirements, real experience and a short call.
In late September, LinkedIn's chief executive said in an interview that job seekers are sending noticeably more applications than before the pandemic, and that employers find it ever harder to tell who can actually do the job. Nobody measures how much more it is in the Netherlands. What is clear is that a cover letter and a CV now cost the applicant less effort than ever, which means they also tell you less about the person behind them.
How many applicants use AI?
Nobody knows exactly, because nobody records which letters were written with AI. What Eurostat does show is how ordinary the tools have become. The Dutch figure is well above the EU average of 32.7 per cent. Among Dutch people aged 16 to 24, 73.6 per cent used generative AI in those three months. If you advertise a role for graduates or young professionals, you can assume nearly every applicant has one of these tools within reach.
That in itself is not the problem. Using AI on a cover letter is not so different from asking a friend to read it over, only faster. The problem lies elsewhere: the effort an application used to take was a signal for you as an employer. That signal is gone.
Why does a cover letter say less than it used to?
A cover letter did three things. It showed that someone had read the advert, that they could write, and that they had made an effort for your role in particular. A language model does all three in a minute. It picks up the wording of your advert, writes without errors and sounds motivated, even on the twentieth application of the day.
The same tools also tune the CV to the words in your advert. A filter that searches for keywords then mostly lets through whoever copied best. You end up with a pile that looks better than ever without containing more suitable people. If a lot comes in and little of it fits, the advert itself is often part of the cause; more on that under wrong applicants.
And it takes time. Reading every letter properly means more hours on the same vacancy, while the strongest candidates are already talking to someone else. What to do when that becomes too much for your people is covered under an overloaded recruitment team.
How do you screen applications when every letter looks good?
By looking less at the writing and more at what someone can show. Five things help.
Put the hard requirements first. Which two or three things must someone really have: a driving licence, a diploma, weekend availability, a work permit? Ask for them in the application form with a short question each. An answer to a specific question can be checked; a nice paragraph cannot.
Replace the letter with a question about the work. Do not ask why someone wants to work for you. Ask something only a person with experience can answer. Which machine did you last set up, and what went wrong? How many beds did your ward have? That answer can be written with AI too, but in the conversation you quickly notice whether it really came from the applicant.
Call within a few days. A ten-minute phone call tells you more than any letter. It is also the moment the candidate notices that you are different from the other employers they applied to that week.
Do not reject a letter because it sounds like AI. A detector gives a probability, not proof, and someone who had help with the letter may still be the best person for the job. Judge the content: does what the letter says match the CV and what the person tells you on the phone?
If you filter with AI yourself, keep a person in the loop. According to the Dutch Data Protection Authority, a decision quickly counts as automated under Article 22 of the GDPR when a score is produced entirely automatically and decides who goes through. That is only allowed under conditions. On top of that, regularly read a sample of rejected applications by hand. Which other rules apply, and what is added in December 2027, is set out in Which EU rules apply to AI in hiring right now?
If the pile grows too large to screen well yourself, you can also hand that work over. With RPO, an outside partner runs recruitment under your name, including first contact and follow-up, while you keep the final say.
What do employers ask us about AI in applications?
Can you tell whether a cover letter was written by AI?
Sometimes, but not reliably enough to reject anyone over it. Text from a language model tends to be general, repeats the wording of the advert and mentions little you can verify. That last point is the useful one. So do not ask whether a letter was written with AI. Ask whether it contains something specific you can follow up on in the interview. If it does not, the letter tells you little, whoever wrote it.
Should you turn down applicants who use AI?
No, and you would lose good people by doing so. According to Eurostat, 44.7 per cent of Dutch people aged 16 to 74 used generative AI in 2025, and for young people a chatbot is simply a tool. If the role requires someone to write well on their own, test that within the process: a short writing task during the interview says more than a letter written at home.
Can you use AI yourself to screen applications?
Yes, with conditions. A system that decides on its own who drops out quickly falls under Article 22 of the GDPR, and from December 2027 stricter European rules apply to AI in hiring. Use AI to pre-sort and summarise, let a recruiter who understands the role make the decision, and check now and then what has been filtered out. Where AI does help in hiring is covered under AI recruitment.
Should you drop the cover letter altogether?
For many roles, yes. Especially for practical and operational jobs, a letter can put off people who do not like writing, while saying little about how they do the work. Replace it with two or three short questions about experience and availability. For roles where writing is the job, you can keep it, but judge it together with a writing task in the interview.
SocialFind recruits for employers who need people on a structural basis, and wrote this article from Eurostat's figures and the guidance of the Dutch Data Protection Authority, checked on 6 November 2026. If you would like articles like this in your inbox, sign up for our newsletter.
Our newsletter
What we run into in our work for employers, gathered in one email: new articles, labour market figures and what follows from them in practice.