For anything that involves judging a candidate's quality, no. AI scores should route candidates, deciding who a recruiter reviews first, while a person makes every rejection. Automatic rejection is defensible only for objective, disclosed knockout rules, such as a legally required licence or a work authorisation the role cannot sponsor.
Every recruiter on a high-volume role has thought it: if the AI already scored 800 applicants, why click reject 600 times? It is one of the most consequential settings in an AI hiring stack, because it is where a ranking aid starts making employment decisions on its own.
What does the law say about AI rejecting candidates?
No US law says an algorithm may never reject an applicant. What the rules do say is that if a tool contributes to a rejection, you own the outcome, you must be able to show how the tool worked, and in a growing number of places you must tell the candidate it was used.
At the federal level, the EEOC removed its AI-specific guidance in January 2025, but Title VII, the ADA and the ADEA did not change. A screening tool that produces a disparate impact on a protected group can be unlawful whether a person or a model applied it.
New York City's Local Law 144 requires an independent bias audit within the past year, a public summary of results, and notice to candidates at least 10 business days before an automated employment decision tool is used. The notice must explain how to request an alternative selection process or accommodation, if one is available.
Illinois' HB 3773, in force since 1 January 2026, bars AI use that has the effect of discriminating on protected characteristics, requires notice, and bans ZIP codes as a proxy for a protected class. The older AI Video Interview Act covers AI analysis of recorded video interviews: notice, an explanation, consent, and deletion within 30 days of a candidate's request.
California's automated-decision system regulations took effect on 1 October 2025 for employers with five or more employees, with a four-year record-keeping requirement.
Colorado's SB26-189, effective 1 January 2027, requires pre-use notice, an explanation after an adverse decision, and an opportunity for human review.
Read together, nobody is banning automation. Regulators want notice, records, and a person involved in adverse decisions. For the full picture, see Is It Legal to Use AI to Interview Candidates in 2026?
When is automatic rejection reasonable?
The safe zone is narrow. Automatic rejection holds up when the criterion is objective, stated in the job post and clearly job-related: a licence the role legally requires, a "no" to a work authorisation question for a role you cannot sponsor, or an incomplete application. A person applying these rules would reach the same result.
It gets risky when the tool infers something. Scores based on tone of voice, facial expression, word choice, personality or "culture fit" are the category most likely to raise bias and disability concerns. A speech difference or an accommodation can move those signals without saying anything about ability to do the job.
Are AI scores reliable enough to draw a hard line?
Usually not. Ranking models are good at ordering candidates and much weaker at drawing a cutoff. Two candidates one point apart are not meaningfully different, but a hard cutoff treats them as opposites. Edit the job description and the shortlist shifts.
That is normal for a scoring system, and a good reason to treat the score as a queue, not a verdict.
How should you configure the gate?
Disclose it. Say in the job post and application flow that AI is used to review applications.
Separate rules from scores. Knockout rules such as licence, authorisation and location can reject automatically. Scores should route candidates to human review, not reject them.
Keep a person on adverse decisions. A recruiter reviewing a ranked list and confirming rejections in bulk is human review. An automated email nobody read is not.
Keep records: criteria, thresholds, scores, outputs and audits. Four years covers California's requirement.
Test for impact. Compare pass rates across groups at each stage, and look into any stage that narrows the pipeline more for one group.
How GetHirePlus handles this
In GetHirePlus, AI resume screening gives each applicant a relevancy score, and you set a threshold per role. Candidates above it are surfaced for review or moved to the next stage. Candidates below it are not advanced automatically, and they are not rejected automatically either. They stay in the ATS, where a recruiter can review the score and move them forward.
AI phone interviews produce a full transcript alongside the scores, so any rejection traces back to something a person can read.
Let the machine decide who gets looked at first, and let a person decide who gets told no.
Sources
K&L Gates, federal AI guidance reversal (January 2025)
NYC DCWP, Automated Employment Decision Tools
Illinois Public Act 103-0804 (HB 3773)
Illinois Artificial Intelligence Video Interview Act (820 ILCS 42)
California Civil Rights Council, automated-decision systems regulations (June 2025)