As of 13 August 2026, AI can only partly screen candidates against job criteria.

What to actually do

1. Hand it to a person

   The route this page recommends

   Someone with a licence or accountable authority has to sign this before it counts.

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 current job description and recruitment policy, then copy the essential and desirable criteria into a separate working document.
   2. Remove unnecessary personal data from each application, including photographs, names, addresses and information unrelated to the selection criteria, while retaining job-relevant evidence.
   3. Check that every criterion is objective, necessary for the role and capable of being evidenced before pasting it into the prompt.
   4. Paste the prompt into a chatbot, add the job title and criteria, then submit one anonymised application at a time.
   5. Compare each evidence row with the original application and mark any claim the chatbot has misread, omitted or inferred.
   6. Ask an HR adviser or employment solicitor to check the criteria, process and any proposed rejection where discrimination, reasonable adjustments or sensitive personal data may be involved.
   7. Have an authorised human recruiter make and record the shortlist decision, including the evidence and human reasoning that support it.

   Prompt

   Act as a recruitment administration assistant, not the decision-maker. Screen the candidate information below against the objective, job-related criteria only. Do not use or infer protected characteristics, health information, disability, age, sex, race, religion or belief, sexual orientation, pregnancy, marriage or civil partnership, gender reassignment, nationality, caring responsibilities, photographs, names or addresses as selection factors. Do not guess missing information.
   
   For each criterion, produce a table with: criterion, evidence found, evidence source, meets / partly meets / not evidenced, and a short explanation. Separate essential criteria from desirable criteria. Flag ambiguous wording, inconsistent evidence, possible reasonable-adjustment issues and any criterion that may be unrelated to job performance. Do not recommend rejection solely because evidence is missing. End with a provisional evidence summary, not a hiring decision, and list the points a trained human recruiter must check before contacting candidates.
   
   Job title: [JOB TITLE]
   Essential criteria: [PASTE CRITERIA]
   Desirable criteria: [PASTE CRITERIA]
   Candidate information: [PASTE ONE ANONYMISED APPLICATION AT A TIME]
   
   Use only the information supplied. Do not rank candidates against one another unless I provide the same criteria and comparable evidence for every candidate. Do not invent facts, make legal conclusions or draft an automated rejection.

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