As of 13 August 2026, AI can only partly compare candidates' interview performance.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open the job description and the agreed interview scorecard, then copy only the objective essential criteria and competency definitions into a working document.
   2. Gather the same questions, interviewer notes and authorised transcripts for every candidate, and remove irrelevant personal data and protected-characteristic information before sharing anything with an AI tool.
   3. Check your organisation's recording, monitoring and recruitment privacy process, including whether candidates were informed and whether the material can be used for this comparison.
   4. Paste the prepared criteria, rubric, questions and labelled candidate evidence into the prompt, then ask the AI to produce the evidence table and bias checks.
   5. Compare every claimed piece of evidence in the output with the relevant transcript or note, correcting scores where the quotation is inaccurate, incomplete or taken out of context.
   6. Give the comparison and the source evidence to the interview panel, ask each member to record an independent decision against the job criteria, and retain the reasons and any adjustments under your normal recruitment process.

   Prompt

   Compare these candidates' interview performance for the same role using only the job criteria and interview evidence supplied below. Do not infer personality, protected characteristics, health, disability, age, ethnicity, religion, sex, pregnancy, sexual orientation, gender reassignment or other personal characteristics. Do not treat accent, fluency, confidence, eye contact, facial expression or communication style as evidence of ability unless it is directly and objectively relevant to a stated job requirement. Do not make the hiring decision.
   
   First, create a table with one row per candidate and one column for each competency. For every score, quote or closely paraphrase the specific evidence from that candidate's transcript or notes, and mark the evidence as missing where there is none. Use this scale: 0 = no relevant evidence, 1 = weak or unclear evidence, 2 = adequate evidence, 3 = strong evidence. Do not score a competency where the interview evidence does not address it.
   
   Then identify any material differences in the questions asked, probing, time allowed or evidence available between candidates. Flag possible inconsistencies, assumptions, missing evidence and wording that could create bias. State which parts require review by the interview panel. End with a neutral evidence summary for each candidate, not a recommendation or ranking.
   
   Job description and essential criteria:
   [PASTE JOB DESCRIPTION AND CRITERIA]
   
   Agreed competencies and scoring definitions:
   [PASTE RUBRIC]
   
   Interview questions asked to every candidate:
   [PASTE QUESTIONS]
   
   Candidate interview evidence, labelled by candidate and question:
   [PASTE REDACTED TRANSCRIPTS OR NOTES]
   
   Before using this comparison, confirm that every candidate was informed about recording or transcription where required, that the material is authorised for this use, and that irrelevant personal data has been removed.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.