AI Hiring Tools in 2026: Ten Questions to Ask Before You Buy
Short answer: Before using an AI tool to screen, rank or interview candidates, ask the vendor what it measures, how it was validated, what its latest bias audit found, what notice candidates get, how to request an alternative process, and who is liable. Under laws such as New York City's Local Law 144, the employer carries the duty, not only the vendor.
AI now sits inside most applicant tracking systems: résumé ranking, chat screening, one-way video interviews, skills tests scored by a model. For a growing company these features are tempting, because they promise to handle hundreds of applicants with a two-person recruiting team.
The rules have caught up. New York City requires a yearly independent bias audit and candidate notice for automated employment decision tools (NYC DCWP). Illinois amended its Human Rights Act to cover discriminatory use of AI in employment decisions and to require notice (Illinois General Assembly). California's Civil Rights Council adopted rules on automated decision systems in employment (California Civil Rights Council), and Colorado's AI law has been delayed and revised more than once. Our AI hiring law updates page tracks the current status.

Ten Questions for Any AI Hiring Vendor
- What exactly does the tool decide or score? Ranking, screening out, scheduling and note-taking carry very different risks.
- What job-related skills does it measure, and how? Ask for the link between each score and the job.
- How was it validated? Ask for a validation study showing scores predict performance in jobs like yours.
- What did the latest bias audit find? Ask for selection or scoring rates by sex and race or ethnicity, and the date and auditor.
- Can we audit our own data? You need exports of applicants, scores and outcomes by stage.
- What notice do candidates receive, and when? Several laws require notice before the tool is used.
- How can a candidate request another process or an accommodation? People with disabilities must have a path that works for them.
- What data does it collect, and does it train on our candidates? Video, voice and inferred traits raise privacy risks.
- Can a person review and override every decision? Avoid fully automated rejections.
- What does the contract say about liability and audits? Employers usually remain responsible for discrimination even when a vendor's tool caused it.
Red Flags in Vendor Answers
- "Our AI is unbiased by design." No tool is; ask for results.
- Scores based on facial expressions, tone of voice or word choice with no clear link to the job.
- No way to export your own data for an audit.
- A bias audit that is more than a year old or run on another customer's data only.
- No documented process for accommodations.
Check Your Own Results
Even a well-built tool can screen out one group more than another in your applicant pool. Run the four-fifths rule on each stage the tool touches, at least quarterly. Hiring Fairness has a free calculator and an AI hiring law checker that shows which rules likely apply where you hire. If a stage is flagged, pause automatic rejections at that stage while you investigate.
Insight Studio can hold the stage-by-stage export so you can track rates over time, and Organization Viewer keeps the hiring managers and teams current so results roll up by department.
When to Bring in an Independent Auditor
If you hire in New York City with a covered tool, you need an independent bias audit every year. Elsewhere, an outside review is sensible before a large rollout or after a flagged result. Our Independent Bias Audit meets the New York City requirements, and the AI Hiring Risk Assessment reviews every AI tool in your hiring process against the laws where you hire.
Questions
Is the employer or the vendor responsible for AI hiring bias?
Usually the employer. Laws such as New York City's Local Law 144 place duties on the employer using the tool, and anti-discrimination law generally holds the employer responsible for its hiring decisions.
Do small companies need a bias audit?
If you use a covered tool for candidates or employees in New York City, yes, regardless of size. Elsewhere it is not always required, but checking your own pass rates with the four-fifths rule is good practice.
What is the four-fifths rule?
A rule of thumb from the federal Uniform Guidelines: if one group's selection rate is below 80% of the highest group's rate, the step may have adverse impact and deserves a closer look.
Do I have to tell candidates that AI is used?
In several places, yes. New York City, Illinois and California have notice rules. Even where not required, telling candidates and offering another process builds trust.
Is AI note-taking in interviews covered by these laws?
Often not, if the tool only records and summarizes. Once it scores, ranks or recommends candidates, it is more likely to be covered. Check the definitions in each law where you hire.
Sources
- NYC Department of Consumer and Worker Protection: Automated Employment Decision Tools
- Illinois General Assembly: HB 3773 (AI in employment decisions)
- California Civil Rights Council
- Colorado General Assembly: SB 24-205, Consumer Protections for Artificial Intelligence
- eCFR: Uniform Guidelines on Employee Selection Procedures (29 CFR 1607)