AI reads your CV before a human does
Most applications submitted today are read by a machine before they are read by a human. Applicant Tracking Systems (ATS) automatically scan and score candidates, and the result often determines who progresses to an interview. This is no longer the exception. It is the norm.
What the ATS system actually assesses
What many people don’t realise is what is actually being assessed. Modern ATS systems don’t just look at your CV. They assess the CV, answers to application questions and the cover letter collectively, and use all of this to calculate a match score against the role. The system looks for matches between the words in the job advert and the words in the application: specific titles, tools and qualifications, not just a general impression of experience.
This has practical implications for how a CV should be presented. Creative layouts with columns, tables, icons or a CV formatted as an image may look appealing to a human, but can make the text difficult or impossible for the system to read correctly. The result is often that content is excluded from the assessment without the candidate realising it. A clean, standardised format with clear headings and continuous text is a safer choice, even if it seems less visually exciting.
The cover letter is back
The cover letter had ‘gone out of fashion’ for a while. Many people stopped writing them, simply because they were rarely read by anyone. Now it’s back, for one simple reason: if the letter is missing, so is the data, and less data means a lower score, no matter how well-qualified the candidate actually is. The cover letter also has a value that has nothing to do with AI: it is the place where motivation, context and the nuances that a bare CV form fails to capture still reach a human being later in the process. A short and specific letter, explaining why this particular role and what the candidate can actually contribute, achieves both of these things at once.
What candidates should do
For job seekers, this means that the structure of the CV has become more important, not less so. A brief two- or three-line summary at the top, clear contact details and a maximum length of three pages provide both people and machines with a better starting point. Outdated or irrelevant experience – ideally anything dating back further than 2018 – should be omitted in favour of what is actually relevant to the role. It’s also worth mirroring the wording used in the job advert where appropriate, as these are precisely the words the system is looking for.
The same applies to LinkedIn. Recruiters use it actively, but AI tools are increasingly doing so too: searches for skills, activity and profile content are part of how candidates are identified, not just how they are assessed after they have applied. An up-to-date profile picture, visible skills and regular activity make a real difference to how well a profile is picked up.
What this means for employers
For employers, this is first and foremost a reminder that good candidates can be filtered out for the wrong reasons if the process isn’t set up correctly. A threshold set too high, a poorly worded advert or a system that isn’t tailored to the type of role can cost a business candidates it never even got to see. It is therefore not enough simply to set up a system and blindly rely on the rankings. Someone should regularly review the candidates who were filtered out, not just those who came out on top.
There is also another side to the same issue. Automatic keyword matching tends to penalise candidates with non-traditional backgrounds: those who have changed sectors, have a gap in their CV, or who use different words to describe the same experience than those used in the job advert. These candidates are not necessarily assessed and rejected; they are often never even seen by a human at all. For an employer, this means a drop-off for which there are no statistics, because the candidates never feature in the figures for ‘assessed applicants’ in the first place.
This is also an area that is set to be regulated. The EU’s AI Regulation (AI Act) classifies AI systems used for recruitment, candidate screening and ranking as high-risk. This entails requirements for risk assessment, technical documentation, testing for bias and transparency towards candidates regarding the use of AI; and, most specifically, that no decision on rejection or recruitment may be made by the system alone, without a qualified human being involved. The requirements are expected to come into force from 2027 and are also likely to affect Norwegian businesses through the EEA Agreement. In other words, employers who have already incorporated genuine human judgement into their processes are not only doing the right thing today; they are also ahead of the curve as the regulations are being developed.
Our DNV certification is built around precisely this principle: that technology streamlines processes, but that a human being makes the final decision.
For candidates, this serves as a reminder that preparation is more important than ever – not because the requirements have become stricter, but because the first stage of the assessment is no longer carried out by a human being.