What really happens to your CV before any human sees it and how to clear every filter
Few weeks ago I got on a call with a software engineer in Eindhoven. Eight years of experience. Embedded systems background. Had applied for a Senior Software Engineer role - good fit on paper.
He hit submit at 5PM. The rejection arrived at 5.04PM. Then the words: "We have decided to move forward with other candidates at this time." No human reads that fast.
He wanted to know what had gone wrong. Was his experience not relevant enough? Was it his English? He had written the CV in English rather than Dutch? Had someone more qualified applied? He had been through this three times in two months and was starting to question whether he was the problem.
I pulled up the role next to his CV. Within two minutes I could see exactly what had happened and it had nothing to do with his experience, his English, or his competition.
His CV was a two-column PDF built in Canva. Contact details in a styled header the parser couldn't read. Skills listed as visual rating bars - graphic elements, invisible to the machine. The ATS had extracted a partial profile, scored it below the threshold, and filtered him out before a single recruiter opened a single document.
No human had seen his CV. Not one.
I have versions of this conversation regularly. Talented people, solid CVs by any human standard, eliminated by a system they didn't know existed, for reasons nobody told them.
This is what's happening across tech hiring right now. The machine isn't assisting the process. It is the process.
Here's what it actually does to your CV from the moment you click submit.
Three Automated Layers. Most Candidates Know About One.
Layer One: The ATS - Older, Simpler, Still Causing Damage
The Applicant Tracking System is not intelligent. It is a structured database with text-extraction logic. It pulls your job titles, employment dates, education, and skills into fields and if your formatting gets in the way, it either misreads the data or loses it entirely.
Roughly 98% of enterprise employers across Europe now route all applications through ATS platforms like Greenhouse, Workday, Lever, SAP SuccessFactors etc. This is the baseline, not a trend.
I have seen candidates with genuinely strong profiles eliminated at this stage because their contact details were placed inside a document header; a formatting choice that looked clean on screen and was completely invisible to the parser. Technically, incomplete profile. Systemically, never had a chance.
A two-column layout, a graphic skills bar, an embedded table, a beautifully designed PDF, - all of these can produce a document that looks impressive and parses as near-blank. Up to 75% of CVs are rejected by ATS before a human sees them. Not because the candidate was unqualified. Because the document wasn't readable by the machine.
Layer Two: AI Resume Screening - Where It's Actually Getting Complicated
By the end of 2025, 83% of companies were using AI to review resumes, nearly double the adoption rate of the prior year. This layer sits on top of ATS infrastructure and operates very differently.
Modern AI screening tools don't just keyword-match. They run semantic analysis: understanding that "cross-functional team leadership" and "interdepartmental project coordination" describe the same capability; that "Python development" sits within "software engineering." In principle, this should benefit strong candidates who write naturally rather than stuffing keywords. In practice, I have seen something more complicated.
These models are trained on historical hiring data, which means they reflect historical hiring biases. In markets like Germany and the Netherlands, where certain universities and employer brand names carry outsized weight, the AI can systematically disadvantage candidates from non-traditional paths even when their actual capability is equal or stronger.
I have sat in enough vendor demos to know that the recruiters buying these tools rarely understand what the model is actually optimising for. And I have had enough talent leadership conversations to know the honest answer when you push on it: most don't know. They trust the score. They rarely interrogate it.
That's the honest state of the technology in 2026. Powerful, partially opaque, making first-round decisions across European hiring at scale.
Layer Three: AI Video Screening - Real, But Not Yet Everywhere in the Netherlands
Let me be precise here, because this is where most career content overstates the picture.
Globally, 56% of organisations use AI-powered video interview analysis. In the Netherlands, it depends entirely on where you're applying. Dutch multinationals - Philips, ASML, ING, Shell are already deploying it for high-volume and graduate roles. For the growth-stage tech companies and high-tech suppliers across Brainport and the Randstad, it's less common for now because hiring volumes are lower and most still open with a recruiter call.
That's changing. The Netherlands is Europe's highest AI adopter, with 95% of organisations running AI programmes. Infrastructure and appetite are both there. A tightening recruiter market is pushing even mid-sized companies toward automated first-round screening. This layer is coming to more Dutch hiring processes faster than most candidates expect.
The more urgent issue is what most Dutch candidates don't know: you have legal rights here. The EU AI Act classifies AI tools used in hiring as high-risk systems, meaning employers must tell you when AI is assessing you, and you can request human review of an automated decision. Many companies are not yet compliant. If you hit a video screening tool and nobody has told you AI is scoring your responses, you are entitled to ask. In the Netherlands, that's regulation not just good practice.
What Candidates Need to Do Differently
The hiring funnel has three automated layers before a human sees your application. Each has different logic and different failure modes. Here's what actually moves the needle.
Format and file basics
Single-column layout only - no sidebars, tables, or text boxes; parsers read top to bottom and table content frequently disappears
Submit as .docx unless the application explicitly asks for PDF; it parses more reliably across major platforms
Contact details in the body of the document, never in the header or footer - many ATS systems don't read those areas at all
Standard fonts: Arial, Calibri, or Georgia at 10–12pt; no decorative fonts, no icons, no infographic-style skill ratings
One to two pages maximum; some platforms truncate before parsing
Section structure
Use standard, unambiguous headers: Professional Experience, Education, Skills, Professional Summary - not My Journey or What I Bring
Include a dedicated Skills section; AI tools specifically scan for a structured skills block
Lead with a Professional Summary containing your strongest keywords - this section is weighted heavily by screening algorithms
Non-standard job titles: add the industry equivalent in parentheses - Growth Lead (Head of Marketing)
Keywords and content
Mirror the exact language of the job description, not synonyms - if the role says "stakeholder management," use that phrase
Target 60–80% keyword alignment; going higher can trigger spam filters in newer AI screeners that penalise obvious over-optimisation
Quantify everything: headcount managed, budget owned, percentage improvements, revenue influenced, time saved - numbers are concrete signal for scoring models
Cut vague descriptors entirely: "results-driven," "passionate," "dynamic," "team player" they consume space and contribute nothing to your match score
Use current terminology; outdated tool names and methodologies reduce semantic match scores even when the underlying experience is directly relevant
Before you submit
Paste your CV into a plain text editor - what reads cleanly is roughly what the ATS extracts; this is the single fastest diagnostic you can run
Use Jobscan or Resume Worded to check keyword alignment against a specific job description for roles you particularly want
Maintain a master CV and build tailored versions per role cluster - a platform engineering CV should look different from a technical leadership CV
What Employers Are Getting Wrong
Automated screening has created a false sense of efficiency. Faster shortlists are not the same as better hires. I have watched companies cut time-to-shortlist in half and spend the next six months managing out someone the algorithm loved and the team couldn't work with.
The models optimise for pattern-matching against historical hires. If your previous Senior Engineers came predominantly from the same cluster of companies and institutions, your AI screen will keep favouring that cluster regardless of whether those were actually your best performers.
The companies hiring most effectively right now use AI for administrative filtering like volume reduction, completeness checks, basic qualification matching, while preserving human judgment for potential, trajectory, and fit. That's the right division of labour. Many have not found it yet.
The Bottom Line
Your CV doesn't land on a recruiter's desk. It lands in a parser, gets scored by a model, and if it clears the threshold, eventually reaches a human who will spend under 60 seconds deciding whether to proceed.
The rejection you felt wasn't a person passing on you. In most cases, it was a system that couldn't read your document, or a model that didn't find enough signal to rank you above the threshold.
That's fixable. But only if you know that's what's happening.
5 Key Takeaways
1. Three automated systems filter your application before any human sees it.
ATS parsing, AI semantic scoring, and AI video screening each operate differently. Most candidates only know about one and optimise for the wrong layer.
2. Formatting is a technical requirement, not a design choice.
A visually impressive CV that parses poorly is worse than a plain one that parses cleanly. Up to 75% of CVs are rejected at the ATS stage alone.
3. AI screening rewards signal, not keywords.
Quantified outcomes, specific achievements, and role-matched language give scoring models what they need. Vague descriptors and generic claims contribute nothing to your score and in some systems, actively lower it.
4. European candidates have legal rights in AI-assisted hiring - most don't know it.
Under the EU AI Act, AI tools used in employment decisions are high-risk systems. Candidates have the right to know when AI is assessing them. Many companies are not yet compliant.
5. The human interview is now the easiest part of the process to reach only if you've handled the automated layers correctly.
Most candidates prepare extensively for interviews and minimally for the three filters they face before the first human call. Reverse that ratio.
Recruiter Brief lands in your inbox every Tuesday. Free Insider EU hiring intelligence
Subscribe here - https://www.recruiterbrief.com/
Until next time,
Amruta Bhargava - Senior Tech Recruiter
The Recruiter Brief
