Age Bias in CV Screening: Are You Losing Experienced Candidates Without Meaning To?

Where Experienced Candidates Drop Out

Research on age and hiring keeps pointing to the same place: the first sort. In April 2026, France’s labour ministry research department published a DARES review of the studies measuring hiring discrimination against older candidates. Most of the field experiments it covers, in which two identical CVs are sent to real job ads and only the age changes, conclude that older candidates are treated less favourably. As reported by Previssima, younger candidates are on average about 1.5 times more likely to be called back, and in one Swedish study cited by the review, the callback rate already starts to fall around the age of 40.

Candidates feel it too. In the Défenseur des droits and ILO survey published in December 2024, which focused on workers aged 50 and over, nearly a quarter of seniors reported discrimination at work, with hiring as the most exposed moment, and a quarter of unemployed seniors said they had been made to understand during an interview that they were too old for the job.

A callback test measures what happens before any conversation. That is the useful lesson for an HR team: the loss happens at the CV screen, the step with the most volume, the least time per application and the least written reasoning. It is also the step where nobody looks back, because rejected CVs rarely get a second reading.

Five Ways Age Enters the Criteria

French law is clear on the principle: article L1132-1 of the Labour Code lists age among the grounds on which no candidate can be excluded from a recruitment process, and a job ad cannot state an upper age limit unless a legal text imposes one. That is precisely why age rarely appears as such. It enters through criteria that each sound reasonable on their own.

• The experience range read as a ceiling. “3 to 5 years of experience” was written to describe a level. Under time pressure, it becomes a maximum, and a candidate with 15 years is set aside as “not the profile” without anyone checking what those years contain.
• “Overqualified” as a verdict. The word stands for several assumptions at once: they will be bored, they will leave, they will not accept being managed by someone younger. None of these is checked on a CV, and all of them can be asked in a ten minute call.
• The salary guessed from the career length. “Too expensive for us” is decided without knowing what the candidate expects. Some experienced candidates are looking for exactly the scope and the pay on offer.
• The dates on the page. A graduation year, a first job in the 1990s, a CV that runs to two pages: these are read in seconds, long before the skills are.
• The team description. “Young and dynamic team” in the ad, or “will they fit in?” at the screen, turns the average age of the current team into an unwritten requirement.

None of this calls the screener’s judgment into question. Each shortcut exists because it saves time on a real concern. The problem is that the concern is never checked, and the shortcut lands on the same group of candidates every time.

Why Volume Makes It Worse

Shortcuts grow with the pile. When an HR manager has a hundred or more applications for one role and an afternoon to sort them, the first visual cues decide more than they should. Our article on what manual CV screening really costs shows how criteria drift as the volume rises. Age proxies are one of the first things that drift in, because dates and career length are the fastest things to read on a page.

Without an applicant tracking system

CVs arrive in a shared inbox or a job board account, and the sort happens by opening attachments one after another. There is no written list of criteria in front of the screener, so each CV is compared with the previous one and with a mental picture of the “right profile”. Rejections leave no trace beyond a folder, which means nobody can ever see who was set aside, or why.

With an applicant tracking system

The risk moves into the configuration. A knockout question on “years of experience” with a maximum, or a filter set up years ago for another role, rejects automatically and silently. Anonymous screening modes help with names and photos, but they usually leave dates and career length visible, which are exactly the cues at stake here. And when screening relies on AI, a model asked to match “3 to 5 years” will apply that range faithfully, at scale. Our article on who is liable when AI screening discriminates covers that side in more detail.

A Worked Example

The figures below are illustrative, but the pattern is common. A 140 person company hires an operations coordinator. The ad asks for “3 to 5 years of experience”. The role receives 210 applications, and the HR manager sorts them over two afternoons, at roughly one minute per CV. 24 candidates go to a first call.

A month later, before closing the role, the HR manager rereads the rejected pile against the three must haves that actually define the job: running supplier planning, using the company’s ERP, managing two people. 31 rejected candidates had more than 12 years of experience. 9 of them met all three must haves, more than the shortlisted candidates did on average. None had been rejected for a stated reason. They had been rejected for the length of their career, through the experience range.

Nobody in this example decided to exclude older candidates. The range in the ad did it, applied at speed, with no written criteria to compare against. That is the useful point: the fix is not to screen with more good intentions, it is to change what the screen measures.

Taking Age Out of the Screen

The goal is not to favour any group. It is to make sure the screen judges what the job requires, for every candidate, in the same way.

• Write the criteria before reading the first CV. Three to five must haves, each one a capability or a result, not a number of years. Our article on writing a job description that filters the flood shows how to phrase them so the ad already does part of the sorting.
• Replace ranges with thresholds. “At least three years running supplier planning” describes a level. “3 to 5 years” adds a ceiling the job does not need.
• Move the assumptions into the first call. Motivation, salary expectations, comfort with the scope: ask, in ten minutes, instead of deciding on the CV. Publishing the pay range, which the EU pay transparency directive already pushes employers toward, settles most of the salary question before it is asked.
• Reread the ad for coded words. “Young team”, “digital native”, “junior to mid level” when the job does not require it: each one narrows who applies before you screen.
• Score every candidate on the same grid. A short written note per criterion turns a feeling into a decision you can explain, and it is the same logic that makes scoring candidates across the whole process more reliable than a single impression.
• Reread a sample of the rejected pile. Once per role, take ten rejected CVs at random and check them against the must haves. If the same profile keeps coming back, the criteria need adjusting, not the people.

Where tooling helps

Every step above takes time a small HR team does not have in a week with five open roles. That is the real reason shortcuts win: not a lack of care, but a lack of time and of a place where the criteria live. A tool does not remove bias on its own, since it applies the criteria it is given. What it can do is make the written criteria the thing every CV is scored against, keep the reasons for each rejection visible, and leave the HR manager the time to reread the pile. That is how we built Kynto for HR teams in growing companies: sourcing, CV screening, scoring and interview scheduling in one place, with scores tied to criteria you write and can read back.

Key Takeaways

• Age bias in hiring mostly happens at the CV screen, before any conversation, which is where field experiments measure it.
• It rarely comes from a decision. It comes from proxies: experience ranges, “overqualified”, guessed salaries, dates, and “fit”.
• Volume makes the proxies stronger, with or without an applicant tracking system, and AI screening applies them at scale if they are in the criteria.
• Write capability based criteria before reading, use thresholds instead of ranges, and ask in a call what you would otherwise assume.
• Reread a sample of the rejected pile for each role: it is the only place the filter shows itself.

FAQ

Can I reject a candidate because they are overqualified?

Being overqualified is not a protected ground in itself, but if the word stands in for age, the rejection can be discriminatory. The safe approach is also the useful one: base the decision on documented job requirements and on what the candidate actually said about their motivation, not on an assumption read from their CV. For a specific dispute, an employment lawyer remains the right person to ask.

Does anonymising CVs solve age bias?

Partly. Hiding the name, the photo and the date of birth removes the most direct cues, but graduation years, career length and the dates of each job usually stay visible. The criteria matter more than the masking: a screen built on capabilities treats a 15 year career as evidence, not as a warning sign.

How do I know whether our screening filters out older candidates?

Look at the rejected pile, not the shortlist. For a closed role, reread a sample of rejections against the must haves and note how often long careers were set aside without a stated reason. If you use an applicant tracking system, also check the knockout questions and filters for any maximum on years of experience.