Evidence · strength 0 of 5 · searched to 2 Nov 2026
Does AI make better hiring decisions?
Not enough evidence yet to say whether AI in screening and interviews leads to better hires.
What field trials found when employers used AI to interview or screen applicants, alone or as advice to human evaluators: who got hired, and whether they stayed.
VerdictNot enough evidence yet
No good evidence
ImpactNot shown
We show no impact figure until at least two studies qualify.
Evidence strength0 of 5
There is not enough evidence to rate this yet.
Cost per user a year?
AI interview and screening tools are priced by contract or per applicant; too varied to band.
Pilot rating under our v0.3 rules. The rules are published on our method page but not yet independently reviewed, so ratings may change as the method is checked and new studies arrive. Our method.
What the studies found
No impact figure yet. Fewer than two studies qualify, so under our rules we do not show how big the effect is. The studies we found are listed below, with what each measured.
How sure are we?
- Starting level3 of 5, from 2 grade A or B studies.
- Down one: old models0 of 3 studies show they used a model released since May 2025 (the 18-month window); the rest used older models or do not say which.
- Down one: not independent2 of 3 studies were run, funded or written by the company whose AI was tested.
- Down one: mostly not peer-reviewedMost of the weight is on working papers or preprints, together with another problem above.
- Publication biasnot testable with fewer than 10 studies.
- Result0 of 5. There is not enough evidence to rate this yet.
Behind the average
- Applicants allowed to choose (Jabarian and Henkel (2026)): 78% of applicants offered a choice picked the AI voice agent; those who chose it had lower language and analytical test scores.
- Interview experience (Jabarian and Henkel (2026)): AI-led interviews were rated significantly less natural by applicants (p = 0.014); other experience ratings were similar.
- Applicants who accepted an offer (Jabarian and Henkel (2026)): 7% higher likelihood of starting the job (p = 0.003) and 6% higher one-month retention (p = 0.025) under the AI interviewer.
- Applicants invited to the AI interview (Aka et al. (2026)): About 75% of invited applicants did not complete the 30-40 minute AI interview, moving screening cost onto applicants.
- Junior vs non-junior candidates (historical platform data) (Aka et al. (2026)): Adding AI Interview Report ratings raised out-of-sample AUC by 0.18 for junior candidates against 0.08 for others.
- Evaluators given AI suggestions (Awuah et al. (2026)): Evaluators overrode the AI in over 80% of cases where their grade differed, mostly when the AI suggested a higher grade.
- Applicants using LLMs (Awuah et al. (2026)): About 60% of applicants submitted at least one essay that a detection tool (Pangram) flags as LLM-generated.
How, where and why it works
Why it works
2 of 4. AI interviews ask every applicant the same structured questions; human evaluators given AI grades mostly overrode them.
For whom and where
1 of 4. Studies rarely split results by job or applicant group.
What it takes
2 of 4. Trials describe whether the AI decides alone or advises people, which changed the result.
The full cost
1 of 4. One trial reports extra evaluator time; none counts the tool's full cost.
Four separate scores from 0 to 4, never added up: how much the studies tell us about why it works, for whom, what it takes and what it really costs (the EMMIE method).
What to do next
Measure your pilot →If you are trying it yourself
- If you are applying, expect structured AI interviews and prepare as for a standard interview.
If you are rolling it out
- Pilot with a randomised comparison and track who stays, not just who is hired.
- Check employment and data-protection law, and tell applicants.
What it takes
- Decide whether the AI advises people or decides: in one trial, evaluators shown AI grades took longer and hired no better.
What can go wrong
- Applicants rated AI-led interviews as less natural.
- In one trial, most invited applicants did not complete the AI interview.
- Hiring decisions about people are high-risk under the EU AI Act; check the rules before using AI to screen.
The studies
| Study | Grade | Design | People | AI model (released) | Company role |
|---|---|---|---|---|---|
| Voice AI in Firms: A Natural Field Experiment on Automated Job Interviews Brian Jabarian, Luca Henkel, 2026. arXiv 2607.28222 (working paper) | A | field RCT, deployed work | 70,884 applications for entry-level customer-service jobs in the Philippines; 67,056 eligible and randomised (40,103 AI interviewer, 13,557 human interviewer, 13,396 choice); 131 recruiters evaluated applicants; 2,764 applicant survey responses | LLM-based AI voice agent (model and supplier not named in the paper) | authors employed by vendor: The AI interviewer is the recruitment firm's own |
| Better Together: Quantifying the Benefits of AI-Assisted Recruitment Ada Aka, Emil Palikot, Ali Ansari, Nima Yazdani, 2026. arXiv 2507.08029 (preprint) | C | field RCT, deployed work | Experiment 1: 4,323 applicants to four real engineering roles, 1,108 completed the AI interview and were randomised to recruiters with or without the AI Interview Report; 140 finalists (108 attended). Experiment 2: 34,493 applicants to a junior frontend role randomised 70/30 (25,536 AI pipeline, 8,957 resume pipeline); 70 finalists (64 attended). | micro1 'AI Recruiter' (LLM-powered interviewer and LLM skill ratings; underlying model not named, internals proprietary) | authors employed by vendor: Ali Ansari and Nima Yazdani are affiliated with micro1, which built the AI Recruiter and runs the recruitment platform on which both experiments ran |
| Augment or Automate? An Early Field Experiment with Generative AI in Hiring (earlier title: Automation with Generative AI? Evidence from a Teacher Hiring Pipeline) Kobbina Awuah, Ursa Krenk, David Yanagizawa-Drott, 2026. Author PDF dated 7 April 2026 (working paper) | A | field RCT, deployed work | 697 applicants to teaching fellowships at a Ghanaian non-profit, randomised to Human-Only, Human-with-AI-Assistance or AI-Only screening; 4,182 graded essay answers; evaluators were staff or programme alumni | OpenAI GPT-4 (gpt-4-0314 API) (Mar 2023) | none |
Grades: A, a well-run randomised trial in real work; B, a strong natural experiment in real work or a well-run trial on realistic tasks; C, a trial with clear limits or a matched comparison; D, before-and-after, surveys or case studies. How studies are graded.
Does the rating hold up?
- Run without company-run studies1 studies: strength 0, verdict no good evidence, headline band no clear effect.
- Run without working papers and preprints0 studies: strength 0, verdict no good evidence.
How up to date is this evidence?
Method →Evidence searched to 2 Nov 2026. Studies cover models released Mar 2023 to Mar 2023. Rating computed 5 Nov 2026 under Module 15 v0.3 (pilot).
How this page was made: AI research agents searched for studies and read each paper; a second pass checked every number on this page against the paper itself, with the quote and where it appears. The rating is computed by fixed rules, not chosen. No person outside AI Adoption has reviewed it yet. Editorial owner: AI Adoption editors. Spot a mistake? Corrections.
Disclosure: AI Adoption uses AI tools, including Claude from Anthropic, in its research and editorial work. No company pays for or has any say in a rating.