You send fifty applications and hear back from two. The most popular explanation at the moment is that artificial intelligence threw your CV out before a person opened it, and that three quarters of applications never reach human eyes.

AI is more present in hiring than it was a few years ago, so the worry is reasonable. In this article, we’re going to walk you through the processes and how they are built nowadays.

What an ATS does with your CV

An applicant tracking system (ATS) started as a database with a search bar. It reads your CV, pulls out details like job titles, skills and dates, and files them so they can be searched later.

But with AI rising, more and more companies are starting to use AI agents on top of an ATS. Screening is one of the most common places employers put AI to work, and plenty of tools score and rank applicants before a recruiter gets to see anything.

So what is the software doing to your CV?

• It extracts. Your file is turned into structured data: employers, job titles, dates, skills. This is the step that breaks when text sits inside an image, or in a second column, or in the header of the page.
• It compares. Older systems look for the exact words from the advert, but newer ones compare both your CV and the job description to understand how they match. Research on how these systems are built explains why a model can register that “software developer” and “software engineer” are the same job, or that Django and Flask belong with Python.
• It scores and ranks. After analysing your CV, the system ranks you based on the previous comparison and gives the recruiter a ranked list of the best candidates for the position.

This is why a skills list at the bottom of the page does less work than it used to. “Kubernetes” on its own is a word. “Migrated 40 services to Kubernetes and cut deploy time from an hour to six minutes” is the same word with proof attached, and it works whichever kind of system is reading.

Before any of that, there are the questions you sometimes answer before you even upload your CV: can you work here legally, when can you start, what salary are you expecting. Those can cause more immediate rejections than CVs do.

Then there is the volume, which is the reason ranking exists at all. A single role at a company people have heard of can draw hundreds of applications in a few days. A recruiter works down from the top of the ranking and rarely reaches the bottom, so a low ranking does the same job as a rejection, without anyone deciding to reject you.

How to write a tech CV that gets found

Whether a person or a model reads your CV first, both are looking for the same things:

• Use the words from the job advert. Recruiters search their database by skill and tool. If the ad says Kubernetes and your CV says container orchestration, you will not appear in that search. Describing the same work in your old team’s vocabulary makes it look like you have not done it.
• Keep the layout readable by software. Formatting will not get you rejected, but it can stop a system extracting your CV correctly. Text in images, multi-column layouts, tables and details buried in headers and footers are the usual culprits. A single column with ordinary headings usually does the job.
• Answer the application questions accurately. If you need visa sponsorship or you cannot start for three months, the honest answer costs you some applications but saves you from processes that were never going to work.

A field experiment with 9,022 real job applications in the UK, published in Nature Human Behaviour, found that listing each role by the number of years you worked there, instead of by start and end dates, increased callbacks by around 15% for candidates who had time out of work.

Your rights when AI screens you in the EU

If you are applying in the EU, there are some rules that already apply to you.

Under the EU AI Act, AI used to filter applications, rank candidates or evaluate interviews is classified as high risk. The full obligations were due in August 2026 and were pushed to 2 December 2027 by an amendment last July.

The delay doesn’t change these three, which already apply:

• Emotion recognition in the workplace has been banned since February 2025, and the European Commission treats recruitment as part of the workplace. Video interviews that score your facial expressions are not legal in the EU.

• Since August 2026, you have to be told when you are interacting with an AI system.
• Article 22 of the GDPR has applied since 2018. You have the right not to be subject to a decision based solely on automated processing where it significantly affects you.  That means a person has to be meaningfully involved, and you can ask for human review and contest the outcome. You can write to the employer and ask for three things: a human to review the decision, an explanation of how it was reached, and the chance to add anything the system missed. They have a month to reply and if nothing comes back, you can complain to your national data protection authority. 

Our Tech Talent Trends 2026 report found that only 3% of tech professionals are comfortable with fully AI-driven recruitment, and 47% want AI used with human oversight. The law is pointing the same way as the people it covers. 

Write so the software can place you, and so a person can see why you are worth a call. With that in mind, there’s nothing that can stop you from getting that dream job.