Over time, AI has become an indispensable tool for candidates and recruiters alike. For the former, it helps polish a CV, analyse the companies they are targeting or practise answering a potential employer’s questions. According to a JobTeaser-EDHEC study published in February 2026[1], 92% of the students surveyed use it to improve their job application and 71% to prepare for interviews. For the latter, it makes it easier to sift through the influx of applications, writes interview reports and, in some cases, analyses behaviour. According to the OpinionWay barometer for Kelio, 33% of HR managers use AI tools in 2026, compared with 9% in 2024[2]. What are the consequences of these uses? Will organisations lose control over the hiring of their teams? What warnings do the first excesses call for?
AI makes recruitment more efficient, but…
Students first see an advantage: AI allows them to consider avenues they had not thought of, the tool pointing them to areas of expertise related to their background. It also increases the likelihood of a positive response[3], thanks to higher-quality cover letters and CVs, better suited to recruiters’ methods. However, this use does not amount to endorsing the automation of recruitment: 67% of early-career candidates surveyed by JobTeaser and EDHEC are not comfortable with an AI-automated process (CV screening, automated interviews, scoring)[4].
This new practice also has undesirable effects for employers, who face a sometimes larger volume of applications and a standardisation of files. According to 73% of employers, one consequence is that the correlation between candidates’ profiles and the criteria for the job is weak, below 50%[5]. AI therefore does not systematically simplify the work of HR departments. The cover letter, useful for understanding an applicant’s background, might moreover not survive this development, at least for non-managerial roles: when its writing is entrusted to AI, its contribution compared with the CV diminishes, and some companies are giving up asking for it in order to rely on the interview. This is notably the case of La Mie câline, whose human resources director, Anne Duval, told Le Monde that she had removed the cover letter from the application file[6].
Recruiters are also automating their practices more and more thanks to applicant tracking systems (ATS) that help sort CVs. A time saving of 20 to 30% is claimed, and some AI solution vendors, such as HR Engine, maintain that these tools allow a greater number of applicants to reach a first interview conducted by AI[7]. Other solutions, such as conducting pre-interviews, are offered by AI players, but several companies prefer to limit the use of AI to the screening function. This is, for example, the decision of the SME ALSEAMAR[8]: its human resources manager, Elicia Lecoq, explains that her team could not handle the influx of applications for certain positions without a first screening by AI. She refuses, however, to entrust interviews to a machine, because human contact remains, for her and her teams, essential to the company’s recruitment process. She nevertheless regrets that some files are rejected mechanically, perhaps preventing candidates who are relevant for the position concerned from accessing this new opportunity. Going further in automation would, in her view, mean dehumanising the process and multiplying the possible errors, a limit the company does not wish to cross. She also warns: with a smaller number of CVs, too strong an involvement of AI could become a handicap by reducing the number of applications selected. ALSEAMAR is therefore following the evolution of these tools with caution.
Biases and legal risks are already emerging
Some major recruiters go beyond screening. Unilever conducts video interviews in which AI analyses the content of answers to standardised questions, as described in The Camp’s white paper[9] on the evolution of the HR function in the age of AI. This step allows it to pre-qualify a large number of applicants. The company, however, originally used a more advanced system: according to The Guardian, in 2019, the tool analysed candidates’ facial expressions, body language and choice of words, and compared them with characteristics deemed predictive of professional success[10]. In January 2021, HireVue, its provider, announced it was abandoning facial analysis, while a complaint by the NGO EPIC to the Federal Trade Commission was pending[11]. Since 2 February 2025, the European AI regulation has banned systems that infer a person’s emotions from their biometric data in the workplace[12]. These practices are therefore under the radar of the authorities and the public, but the possible excesses deserve particular attention because the limits set by regulation follow the evolution of the tools, which calls for continuous vigilance from recruiters and regulators alike.
Biases have also appeared in the distribution of job advertisements. In 2023, a Global Witness experiment showed that Facebook displayed an early-years educator job ad in France to 93% women and a mechanic job ad to 94% men[13]. Following a complaint by the Fondation des femmes, the Defender of Rights ruled in November 2025 that Meta’s algorithm infringed French non-discrimination law[14].
Screening algorithms have run into comparable pitfalls. As early as 2015, Amazon found that its CV-screening tool downgraded women’s applications, because it had been trained on ten years of CVs received by the company, mostly from men; the tool was subsequently abandoned[15]. More recently, in January 2026, two Californian candidates, Erin Kistler and Sruti Bhaumik, filed a class action against Eightfold AI. They argue that the score from 0 to 5 assigned by the tool to each candidate constitutes a consumer report within the meaning of the federal Fair Credit Reporting Act, compiled without their consent and without the possibility of viewing or correcting it. The company denies any unlawful data collection[16]. Other actions in the United States target screening software for age discrimination[17].
A recent study[18] further points out that a candidate rejected by an algorithmic system may see that rejection stored, then reused when applying to another company that uses the same tool. These risks are added to a legal framework that the authorities are beginning to enforce. The CNIL, the French data protection authority, has made recruitment one of its priority areas of inspection in 2026, targeting automated decision systems, the information given to candidates and data retention periods, primarily at large companies and recruitment agencies. These inspections foreshadow its role as the future market surveillance authority for AI systems applied to the world of work[19]. Beyond the obligations in force, French companies are also striving to preserve their image and the buy-in of their teams by using tools consistent with the values they claim. AI is thus making progress in HR departments, but cautiously: 53% of HR managers say they do not trust it to deliver quality work[20].
AI: an ethical and participatory approach?
Faced with predefined criteria that are sometimes oversimplifying and with biases that are hard to foresee or detect (age, gender, disability, ethnic origin), Janine Berg and Hannah Johnston, authors of a working paper of the International Labour Organization (ILO), consider the optimism surrounding AI in human resources management unjustified. They point to objectives that are often reductive, low-quality or biased data and opaque programming, which expose companies and workers to legal, ethical and practical risks. According to them, the most effective systems require the involvement of their users, HR professionals and workers alike, as well as training HR teams in how the tools work before they are deployed. They consider it unlikely that turnkey solutions will deliver satisfactory results[21].
These difficulties are a reminder that hiring remains first and foremost a human decision: tools can save time or optimise certain analyses, but without fully securing the choice of profile. For organisations, the challenge is therefore to retain their capacity for judgement, and to decide at each stage what they are willing to delegate to AI. As the tools promise to reduce the risk of error, the temptation grows to entrust them with the entire evaluation, to the point that recruitment could come down to two algorithms talking to each other: the candidate’s and the recruiter’s. Yet, beyond data and analyses, recruiting means identifying a potential, a singularity, a capacity to evolve that past experience does not always allow us to predict. For the candidate, the interview is also an opportunity to get to know themselves better and to find their direction. As we see in many professions, is the ongoing revolution not an opportunity to question the place of the human each time we decide to integrate artificial intelligence solutions into our organisations? A question that is at times dizzying, and no doubt essential.
[1]JobTeaser – EDHEC NewGen Talent Centre, “Young graduates: the new rules for recruiting in 2026”, 3 February 2026 – https://content.jobteaser.com/view/106384635/
[2]Kelio – OpinionWay, “AI and HR in companies in 2026: results of the 3rd OpinionWay barometer for Kelio”, 26 March 2026 – https://www.kelio.com/fr/societe/actualites/ia-rh-barometre.html
[3]Le Monde, Nadia Hebbar, “‘What used to take me a week now takes me two seconds’: young graduates use AI to try to get hired”, 22 April 2026 – https://www.lemonde.fr/campus/article/2026/04/22/ce-qui-me-prenait-une-semaine-me-demande-desormais-deux-secondes-les-jeunes-diplomes-utilisent-l-ia-pour-tenter-de-se-faire-recruter_6682353_4401467.html
[4]JobTeaser – EDHEC NewGen Talent Centre, “Young graduates: the new rules for recruiting in 2026”, 3 February 2026 – https://content.jobteaser.com/view/106384635/
[5]Le Point, Romane Ganneval, “When AI makes the hidden job market explode”, 17 August 2026 – https://www.lepoint.fr/economie/quand-lia-fait-exploser-le-marche-cache-de-lemploi-XSW4WSV2JVFVTKOZR2YLQEDKXI/
[6]Le Monde, Sophy Caulier, “Will the cover letter survive AI and changes in recruitment?”, 29 April 2026 – https://www.lemonde.fr/emploi/article/2026/04/29/la-lettre-de-motivation-survivra-t-elle-a-l-ia-et-aux-evolutions-du-recrutement_6684084_1698637.html
[7]Les Echos, Sarah Dumeau, “‘We estimate it saves between 20 and 30% of time’: with AI, recruitment enters a new world”, 25 March 2025 – https://www.lesechos.fr/travailler-mieux/recrutement/on-estime-que-ca-fait-gagner-entre-20-et-30-de-temps-avec-lia-le-recrutement-change-de-monde-2155928
[8]Interview with Elicia Lecoq, human resources manager at ALSEAMAR, conducted by Ludovic Deblois, 22 September 2026
[9]The Camp, “Artificial intelligence in human resources”, white paper, October 2026 – https://training.thecamp.fr/wp-content/uploads/2026/08/Livre-Blanc-IA-RH-FR-2.pdf
[10]The Guardian, Robert Booth, “Unilever saves on recruiters by using AI to assess job interviews”, 25 October 2019 – https://www.theguardian.com/technology/2019/oct/25/unilever-saves-on-recruiters-by-using-ai-to-assess-job-interviews
[11]EPIC, “HireVue, Facing FTC Complaint From EPIC, Halts Use of Facial Recognition”, 12 January 2021 – https://epic.org/hirevue-facing-ftc-complaint-from-epic-halts-use-of-facial-recognition/
[12]European Parliament and Council of the European Union, Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence, Article 5(1)(f), 13 June 2024 – https://eur-lex.europa.eu/eli/reg/2024/1689/oj
[13]Global Witness, “Facebook algorithm breaking the law, French equalities regulator decides”, 4 November 2025 – https://globalwitness.org/en/press-releases/facebook-algorithm-breaking-the-law-french-equalities-regulator-decides/
[14]Le Figaro, Romane Haquette, “Equality in employment: the Defender of Rights accuses Facebook’s algorithms of discrimination”, 6 November 2025 – https://www.lefigaro.fr/secteur/high-tech/egalite-face-a-l-emploi-la-defenseure-des-droits-accuse-les-algorithmes-de-facebook-de-discrimination-20251106
[15]Les Echos (based on Reuters), “When Amazon’s recruitment software discriminates against women”, 13 October 2018 – https://www.lesechos.fr/industrie-services/conso-distribution/quand-le-logiciel-de-recrutement-damazon-discrimine-les-femmes-141753
[16]FindLaw, Kit Yona, “Proposed class action lawsuit targets Eightfold’s AI hiring platform”, 23 January 2026 – https://www.findlaw.com/legalblogs/courtside/proposed-class-action-lawsuit-targets-eightfolds-ai-hiring-platform/
[17]Le Monde, Caroline Talbot, “In the United States, job seekers sue recruitment software using AI for age discrimination”, 6 April 2026 – https://www.lemonde.fr/emploi/article/2026/04/06/aux-etats-unis-des-demandeurs-d-emploi-attaquent-des-logiciels-de-recrutement-utilisant-l-ia-pour-discrimination-par-l-age_6677140_1698637.html
[18]The Guardian, Arielle Pardes, “Will AI give you the job?”, 19 August 2026 – https://www.theguardian.com/technology/2026/aug/19/ai-hiring-tools-discrimination
[19]CNIL, “2026 inspections: recruitment, the single electoral register and sports federations”, 3 April 2026 – https://www.cnil.fr/fr/controles-prioritaires-2026
[20]Kelio – OpinionWay, “AI and HR in companies in 2026: results of the 3rd OpinionWay barometer for Kelio”, 26 March 2026 – https://www.kelio.com/fr/societe/actualites/ia-rh-barometre.html
[21]Janine Berg and Hannah Johnston, “Artificial intelligence in human resources management: the limits of empiricism”, ILO Working Paper No. 154, International Labour Organization, Geneva, November 2025 – https://doi.org/10.54394/NMSH7611
