As of 13 August 2026, AI can only partly rank applicants for a UK job.

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 30 minutes until you can act on the result.

   How to actually do it

   1. Open the final UK job description and write down the essential and desirable criteria that genuinely relate to the role.
   2. Gather the same type of CV or application information for every applicant, then remove names, photographs, addresses, dates of birth and other irrelevant personal details.
   3. Paste the job description and anonymised applications into the prompt, labelling each applicant consistently as Candidate A, Candidate B and so on.
   4. Ask the chatbot to score evidence against each criterion, keep tied candidates where the evidence does not justify a distinction, and flag vague criteria or possible protected-characteristic proxies.
   5. Compare every score and explanation against the original application, correcting invented evidence, missing evidence and inconsistent treatment of candidates.
   6. Ask a colleague involved in the recruitment process to check the criteria, the ranking and the flagged risks before anyone is rejected or invited to interview.
   7. Record the final human decision and the evidence used, and apply the same selection process to all applicants.

   Prompt

   You are helping with a UK recruitment process, not making the hiring decision. Rank the candidates below against the job criteria only. Do not use or infer protected characteristics or proxies for them, including age, sex, race, disability, religion, pregnancy, marital status or any other irrelevant personal attribute. Ignore names, photographs, addresses, dates of birth, school names where they are not relevant, and employment gaps unless the criterion explicitly and lawfully requires context. Do not invent facts or treat missing information as evidence against a candidate.
   
   First, convert the job description into a short table of essential and desirable criteria with an objective scoring rule. Then score every candidate against each criterion using only the supplied information. Show the evidence for every score, mark each criterion as met, partly evidenced, not evidenced or unclear, and identify information that should be checked consistently with every candidate. Produce a provisional order with tied candidates where the evidence does not justify a distinction. Flag any criterion that could create indirect discrimination or is too vague to score fairly. Do not recommend rejection solely from this analysis. State clearly that a human decision-maker must review the evidence, apply the same process to all candidates and make the final decision.
   
   Job title: [JOB TITLE]
   Job description and selection criteria:
   [PASTE JOB DESCRIPTION]
   
   Anonymised candidate information:
   [CANDIDATE A]
   [CANDIDATE B]
   [CANDIDATE C]
   
   Return: 1. the objective criteria and scoring rule, 2. the evidence table for each candidate, 3. the provisional ranking, 4. tied or unclear cases, 5. possible bias or unlawful-proxy risks, and 6. consistent follow-up questions for the next stage.

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