It is tempting to paste a stack of resumes into a general AI chatbot and ask for the best five. The tool is already open, it answers quickly, and the output sounds confident. The problem is what is missing: no bias testing for that use, no consistent criteria, and no record of why one person ranked above another.
What the Research Found
In October 2024, University of Washington researchers reported testing large language models on more than 550 real resumes while varying only the names. They "found the LLMs favored white-associated names 85% of the time, female-associated names only 11% of the time, and never favored Black male-associated names over white male-associated names." One of the researchers noted that "outside of a New York City law, there's no regulatory, independent audit of these systems."
What OpenAI's Own Policy Says
OpenAI's usage policies, updated October 29, 2025, list among the uses it does not allow "automation of high-stakes decisions in sensitive areas without human review," and name employment among those areas. The policy does not ban using the tool in hiring. It does put the burden of human review on the employer.
How the Two Compare
| Question | General chatbot | TrueScan HR |
|---|---|---|
| Built for hiring | No | Yes |
| Consistent criteria | Depends on each prompt | The same job-based criteria for every resume |
| Bias testing for screening | Not designed or documented for this use | Certified weekly bias testing and a 4/5ths adverse impact report on Enterprise plans |
| Record of each decision | Only what you save from the chat | A scored record for each resume |
| Applicant data handling | Governed by the chatbot's general terms | Handled under TrueScan HR's privacy policy for screening |
If Your Team Is Already Doing This
Stop pasting applicant resumes into consumer chat tools, write down the criteria for each role, and make sure a person reviews every decision. Under California's Civil Rights Council regulations, screening resumes for terms or patterns is an automated-decision system task, and employers must keep automated-decision system data for four years. A chat history is not a compliance record.
This page is general information, not legal advice, and TrueScan HR is not a law firm. Statements about other products come from each vendor's public pages as of October 19, 2026, and may have changed since. Product names are trademarks of their owners.
Sources: