5 hiring trends in Spain for 2026

If you hire in Spain, 2026 is not just another year. Two forces that had been converging for a while finally meet: the EU AI Act's obligations for high-risk AI systems in human resources come into application, and the scientific evidence on which selection methods actually predict performance has become too consistent to ignore. The result: processes that survived on habit will now have to justify themselves.

These are the five trends we see consolidating this year, and what you can do about each one if you run hiring at a small or mid-sized company.

1. AI in hiring, yes — but regulated

The AI Act (Regulation EU 2024/1689) classifies AI systems used in employment and personnel selection as high risk (Annex III, category 4). In practice, that means three obligations for any company using these tools: transparency towards the candidate (they must know they are being assessed with the help of AI before the assessment, Art. 50), meaningful human oversight (a person makes the final decision, not the algorithm) and technical documentation of the system.

For a smaller company the translation is simple: before signing with an AI assessment vendor, ask for their transparency documentation, ask who makes the final decision and how a candidate can request human intervention (Art. 22 GDPR). Some vendors already publish this — JOBO, for example, publishes its transparency documentation and its AI Act classification. If a vendor cannot answer those three questions, that tells you something.

2. From reading CVs to assessing by doing

CV screening rejects candidates based on signals that research does not link to performance: the look of the document, inferred age, years in a role. The alternative gaining ground is the work-sample test: give the candidate a task representative of the job and assess the execution. The evidence backs it: the meta-analysis by Arthur et al. (2003) puts its predictive validity at around r ≈ 0.54, among the highest in the literature for a single instrument.

The other piece is competency-based assessment using a structured framework such as KSAO (knowledge, skills, abilities and other characteristics), the US Department of Labor taxonomy behind O*NET. Combining several diverse predictors is what makes the difference: the Schmidt–Oh–Shaffer meta-analysis (2016) estimates an aggregate validity of R ≈ 0.63 for the combination of cognitive testing, structured interview and work sample. A later review (Sackett et al., 2022) argues these figures downwards and the academic debate remains open, but the practical conclusion holds: combining validated predictors clearly beats any single method.

3. The structured interview becomes the standard

The classic interview — free-flowing conversation, improvised questions — has a predictive validity of roughly r ≈ 0.20 according to Schmidt and Hunter (1998): it explains around 4% of the variance in future performance. The structured version of the same instrument — same questions, same order, shared scoring rubric — rises to r ≈ 0.51. It is a validity upgrade that requires no software, only discipline.

That is why more and more companies are formalising scripts and rubrics: questions anchored to the role, rating scales defined before the first interview, and evaluators scoring independently before comparing notes.

4. Video and asynchrony in the early stages

The asynchronous video interview — the candidate records their answers when they can, the evaluator reviews them when they can — solves the scheduling problem of early stages and reduces order and day-of-week bias: the tired interviewer on a Friday afternoon no longer judges differently the candidate who could only make that Friday. The evidence (Hickman et al., 2022, Journal of Applied Psychology) places its validity in the range of the in-person structured interview when assessed with identical structured criteria for everyone.

The condition — and here it links back to trend 1 — is informed consent before recording, clear information on how the video is assessed, and dignified treatment of the candidate throughout. Asynchrony saves time; it does not waive transparency obligations.

5. Candidate experience as employer brand

The 2026 candidate also evaluates the company: they compare processes, share their experience and decide with that information. Opaque processes — no timelines, no reply, no feedback — cost reputation and rejected offers. The trend runs the other way: communicate stages and timings from the start, reply to every candidate and give useful feedback to those who reached the final rounds. No statistic is needed to justify it: it is the part of the process the candidate remembers and talks about.

What to do with all this

• If you use (or plan to use) AI in hiring, ask the vendor for their AI Act transparency documentation and confirm the final decision is human.
• Replace part of your CV screening with a short work-sample test representative of the role.
• Write your interview script: same questions for everyone, rubric defined before you start.
• If you record video, inform and obtain consent first, and explain how it is assessed.
• Set response deadlines for candidates and keep them; give feedback to finalists.
• Document your process: in 2026, being able to explain how you decide matters as much as deciding well.