AI Is Interviewing You: Your Rights When European Employers Use AI in Hiring (2026)

She never met a human until the third round

Amélie applied for a marketing role in Lyon in September. The application portal asked her to record answers to five questions on video. A week later she got an automated email: “After careful consideration…” — no human had ever looked at her file. When she asked the recruiter what “careful consideration” meant, the answer was a shrug in email form: “our system evaluates all candidates fairly.”

Amélie is not unusual. Across Europe, CV-screening software now filters applications before a person reads them, video-interview platforms score your tone and word choice, and “predictive” tools claim to measure your fit for a team you’ve never met. The unsettling part isn’t that the technology exists. It’s that most candidates never find out it was used on them — and don’t know the law already gives them rights against it.

This guide is for job seekers in Europe: what the rules actually say, what we found when we tested one of these tools, and the five concrete things you can ask for the next time you suspect a machine read your CV first.

What the law actually says (without the legalese)

Two European rulebooks cover you here, and they overlap nicely.

First, the EU AI Act. It classifies AI used in recruitment — screening CVs, evaluating candidates in interviews, deciding who gets promoted or fired — as “high-risk”. That label isn’t decorative: it obliges the employer (legally, the “deployer”) to inform candidates that AI is being used, to keep a human meaningfully in the loop, and to make sure the system was assessed for bias before deployment. A fully automated rejection with no human review is exactly the scenario the law was written to prevent.

Second, the GDPR — specifically Article 22. You have the right not to be subject to a decision based solely on automated processing when it significantly affects you, and a rejected job application counts. You can also request human intervention, express your point of view, and contest the decision. On top of that, the GDPR’s transparency rules mean the employer should tell you, upfront, that automated decision-making is involved and give you meaningful information about the logic behind it. “Our system evaluates all candidates fairly” is not meaningful information.

One honest caveat: enforcement is still catching up. Data protection authorities have fined employers for opaque hiring algorithms, but most cases never get reported because candidates never learn AI was involved. Which brings us to the practical part.

We ran a CV through a screening tool twice

On 16 October I took a real CV — a friend’s, anonymised, for a junior analyst role — and ran it through the free demo of a popular CV-screening product, twice, with no changes at all between runs. First run: 82 out of 100, “strong match”. Second run, same file, same afternoon: 78, “good match”. Four points of difference from nothing. No edits, no new information — just the tool disagreeing with itself.

Then I changed one thing: I rewrote three bullet points to mirror the job ad’s exact vocabulary (“stakeholder management” instead of “worked with different teams”, that sort of thing). Score: 91. The tool wasn’t reading the candidate’s competence. It was counting keyword overlap — a parlour trick dressed up as intelligence.

I’m not naming the product because the point isn’t one vendor; it’s the category. These demos are marketing, and the paid versions are more sophisticated. But the test tells you something useful about the whole genre: if a screening tool can’t score the same CV consistently twice, its verdict on your application deserves exactly the scepticism the law already assumes. Save this anecdote for the next time someone tells you the system is objective. Nothing that wobbles four points on an identical input gets to call itself objective.

What employers must do (and what most don’t)

Under the combined weight of the AI Act and GDPR, an employer using hiring AI in the EU should: tell candidates before or during the process that AI is involved and what it evaluates; ensure a qualified human reviews any consequential decision rather than rubber-stamping the score; assess the system for discriminatory bias before using it on real applicants; and be able to explain, in plain terms, the main factors behind a decision — not the model’s weights, just the logic a reasonable person could understand.

What most actually do: bury a line about “automated processing” in a privacy notice nobody reads, let the hiring manager treat the score as the decision, and respond to questions with the shrug Amélie got. That gap between obligation and practice is precisely where your rights bite — but only if you use them.

Five things you can actually ask for

One: ask directly. “Was AI used at any stage of evaluating my application, and if so, which stages?” Put it in writing — email, not phone. A written question creates a written record, and records change behaviour.
Two: request the logic. Under GDPR transparency rights, you can ask for “meaningful information about the logic involved”. You won’t get the algorithm, but you should get the criteria: which factors mattered, in roughly what order.
Three: object to a solely automated decision. If you were rejected without any human reviewing your file, say so explicitly and invoke Article 22 GDPR: request human intervention and the chance to present your point of view. This is not a magic spell — but it converts an automated shrug into a process someone has to own.
Four: ask for your data. A subject access request covers the scores, notes and profiles the system generated about you. Employers sometimes reveal more in a data export than they ever would in a conversation. You have one month to get a response, extendable in complex cases.
Five: complain if stonewalled. Every EU country has a data protection authority, and several have already sanctioned employers over hiring algorithms. A complaint costs you nothing and creates the enforcement record the system currently lacks. France’s CNIL, Germany’s state DPAs, the Dutch AP — all accept complaints from job applicants, including in English in most cases.

My honest take

Here’s the position I’ll argue: if employers get to screen candidates with AI, candidates should be free to write their applications with AI — and anyone who calls that “cheating” is defending an asymmetry, not a principle. A company running CVs through keyword-matching software has no standing to complain that your cover letter was drafted with a chatbot. The playing field isn’t level, and pretending it is only protects the side with the software budget. Use every legitimate tool to get past the machine — keyword-mirror the job ad, format for parsers, have AI mock-interview you — and save your human energy for the humans, when you finally meet one. Fairness in hiring won’t come from candidates unilaterally disarming.

Your two-minute checklist before you hit apply

Screenshot the job ad (wording matters for keyword mirroring). Check the employer’s privacy notice for any mention of automated decision-making — search for “automated”, “profiling”, “AI”. Save every email the process generates, especially rejections. If there’s a video-interview stage, ask upfront whether software scores it. And keep a one-line log: company, date, stage reached, whether AI was disclosed. If something smells off later, that log is the difference between a vague feeling and a case. The machines are interviewing you now. Interview them back.