Use AI where it makes hiring more consistent: job posts that list only job-related requirements, interview scorecards against the same must-haves for every candidate, and summaries that quote evidence. Keep it away from the final call, never let it reject anyone automatically, and check what it writes for inflated requirements and off-topic remarks. Start with A Fair Job Posting with the Interview Prompt, a free 7-minute lesson.
AI is not fair or unfair on its own. It copies patterns. Ask it for a posting with no context and it copies the average posting on the internet, degree requirements and all. Ask it to summarize a debrief and it repeats whatever the panel said, including the comment that should never have been made. Fairness comes from what you tell it to include, what you tell it to leave out, and who makes the decision.
Where does AI make hiring fairer, and where does it make it worse?
Where it helps:
- Job posts. It can separate true must-haves from nice-to-haves you would train, add a real pay range and describe the work instead of a personality.
- Structured interviews. It can draft the same questions for every candidate, tied to each must-have.
- Scorecards. It can turn a debrief into a score per must-have with a quote as evidence, and mark a must-have "Not discussed" instead of guessing.
- Candidate communication. It can write honest, timely status emails so nobody is left waiting.
Where it hurts:
- Ranking or rejecting people with no human review
- Scoring on things like school names, employment gaps or zip codes, which can stand in for protected traits
- Writing "culture fit" language that rewards sameness
- Summaries that carry a remark about family, age or accent into the record
How do I set up a fairer process with AI?
- Write the must-haves first. Four or five, each tied to the work. Everything else is a nice-to-have.
- Make AI interview you before it writes the posting, so it asks for the pay range, schedule and real requirements instead of inventing them.
- Use the same scorecard for every candidate, built from those must-haves.
- Tell AI what to ignore in every summary or screen: name, age, gender, race, photos, address, graduation years and anything that hints at them.
- Ask for evidence, not adjectives. A quote from the resume or the interview for each score.
- Keep the decision with people, and write down why, in job terms.
- Spot-check the low scores regularly, since that is where a pattern would hurt someone.
What does this look like in a real lesson?
In A Fair Job Posting with the Interview Prompt, a physical therapy clinic needs a Front Desk Lead. First you ask ChatGPT the quick way and get a posting that asks for a degree, a certification and years of supervising the job does not need, promises weekend hours the clinic does not have and hides the pay. Then you use an interview prompt with one extra rule, keep every requirement job-related. ChatGPT asks four questions one at a time (what the job involves, must-haves versus trainable skills, pay and schedule, what success looks like), sums up what it understood, and only then writes. You save the prompt as a template with blanks for the job title and company.
The next lesson, Score Every Candidate Fairly After the Interview, turns a panel debrief into a scorecard against the must-haves, with quotes as evidence, and leaves out an off-topic comment about the candidate's kids. In An AI Resume Screener, with a Fairness Step, the AI ranks resumes in a Google Sheet but has no email step, so nobody gets an automatic rejection, and you decide to phone screen a candidate who scored low.
What mistakes should I avoid?
- Copying the old job description into AI. It keeps every inflated requirement you meant to remove.
- Trusting a score without reading the evidence. A number with no quote behind it is a guess.
- Pasting full resumes into a personal account. Candidate details are personal data. Use a business plan.
- Letting interviewers skip the scorecard because AI will "write it up later". The notes need to exist first.
- Ignoring the law where you hire. Some cities and states now regulate AI in hiring. See Is it legal to use AI to screen resumes?
Which tools help?
- ChatGPT or Claude for postings, interview guides, scorecards and candidate emails.
- Textio if you want a tool built for inclusive job post and feedback language.
- Your applicant tracking system's structured scorecards, so the AI output has a home.
For the basics of everyday prompts, see How to Use ChatGPT for Recruiting. For the lesson path, see the recruiter start guide, and for automations, AI automation for recruiters.
Key takeaways
- Fairness starts with job-related must-haves written before AI writes anything.
- Same scorecard, same questions, evidence quoted for every score.
- Tell AI exactly which traits and proxies to ignore.
- AI can rank and draft. People decide, and nobody is rejected automatically.
Frequently asked questions
- Can AI remove bias from hiring?
No tool removes bias by itself. AI can make the process more consistent, which helps, but it can also repeat bias from its training or your old data, so people still need to check the output.
- Is it fair to use AI to write interview questions?
Yes, if every candidate for the role gets the same questions and each question ties to a must-have. Review the list for anything that touches on family, health, age or other personal topics.
- Should candidates be told we use AI?
Telling them is good practice, and in some places it is required for certain tools. Check the rules where you hire and where your candidates live.