As of 13 August 2026, AI can only partly screen CVs for your job vacancy.

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 vacancy and write a short, numbered list of essential and desirable criteria that are genuinely necessary for the role.
   2. Remove photographs, names, addresses, dates of birth, school dates and other unnecessary identifying details from each CV, and keep a separate secure record linking each anonymised label to the candidate.
   3. Paste the job description, criteria and anonymised CVs into a chatbot using the prompt above, with each CV labelled Candidate A, Candidate B and so on.
   4. Check the output table against every CV, confirming that each score has supporting evidence and changing any score that relies on an inference or an invented fact.
   5. Compare the same criterion across all candidates, then record any missing evidence as a fair interview question rather than treating it as a rejection.
   6. Ask a hiring manager or HR colleague to check the criteria, anonymisation and proposed shortlist for consistency and potential discrimination before contacting candidates.
   7. Make the progression decision yourself, record the job-related reasons, and send only the appropriate next-step communication to candidates.

   Prompt

   You are assisting with a UK recruitment process. Screen the CVs below against the vacancy criteria, but do not make the hiring decision or recommend rejection automatically.
   
   Vacancy and assessment criteria:
   [Paste the job description and a numbered list of essential and desirable criteria here]
   
   Candidate CVs:
   [Paste anonymised CVs here, labelled Candidate A, Candidate B and so on]
   
   Instructions:
   1. Assess only evidence that is relevant to the stated criteria. Do not infer ability, personality, health, disability, age, sex, race, religion or belief, sexual orientation, pregnancy or maternity, gender reassignment, marital status, nationality, family circumstances, salary history, or any other protected or irrelevant characteristic.
   2. Do not use names, addresses, photographs, dates of birth, school dates, or other identifying details in the assessment. Flag any such details that should be removed before a human review.
   3. For each candidate, produce a table with: evidence for each criterion, evidence missing or unclear, reasonable questions for a fair interview, and a provisional evidence score from 0 to 2 for each criterion. Use “not evidenced” where the CV does not support a conclusion.
   4. Keep essential criteria separate from desirable criteria. Do not compensate for a missing essential criterion with strengths in unrelated areas.
   5. Quote or closely identify the CV evidence supporting every score. Do not invent qualifications, experience or achievements.
   6. Identify ambiguous wording, inconsistent treatment between candidates, and any criterion that is vague, unnecessary or potentially discriminatory. Do not decide whether a criterion is lawful.
   7. End with a review list for the hiring manager. State that the output is a screening aid only and that a human must apply the same criteria consistently, consider reasonable adjustments where relevant, and make the progression decision.
   
   Return the assessment in plain text with one section per candidate, followed by the review list.

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