Home · Business · HR & People · Job ads & screening
As of 13 August 2026, AI can only partly check candidate references.
This still needs a person who signs their name to it.
Can you do it?
15 minutesto a draft.
30 minutesto something you’d act on.
Cost, all in£0
Skill neededchat-fluent
Who has to check ita colleague
What the alternative costsA purpose-built recruitment platform such as Manatal provides candidate scoring, enrichment and pipeline automation, but it does not transfer responsibility for checking references.
If this goes wrong, you may reject a suitable candidate or hire someone on the strength of an inaccurate or incomplete reference summary.
What to actually do
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.
Use a tool built for this
Second choiceDo it yourself
The distant thirdA chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.
How to actually do it
- Open your recruitment record and confirm the reference-checking process, the candidate's consent and the referee's identity before sharing any information with an AI tool.
- Gather the complete reference responses, the questions asked, the candidate's role details and any relevant employment dates, keeping unrelated personal data out of the material.
- Paste the material into the prompt, replacing each bracketed slot, and ask the chatbot to produce the factual summary, evidence table and follow-up questions.
- Compare every item in the AI summary with the original reference responses, correcting wording that adds an opinion, implication or fact not present in the source.
- Ask a colleague to check the flagged gaps and inconsistencies against the recruitment record and decide which follow-up questions a human should put to the referee.
- Contact the referee through the agreed recruitment process, record the answers and update the reference record without allowing the AI summary to replace the original evidence.
- Make and document the hiring decision separately from the AI output, recording the human reasoning and retaining the consent and reference-checking record under your normal HR process.
Prompt
You are helping with a UK recruitment process. Analyse the candidate reference material below, but do not make the hiring decision and do not invent facts. Candidate: [candidate name or identifier] Role: [job title] Consent and process notes: [what consent was obtained, from whom, and when] Referee details: [name, job title, organisation and relationship to the candidate] Reference questions asked: [paste the questions] Reference responses: [paste the complete responses] Produce: 1. A factual summary separated into employment dates, role and responsibilities, performance, strengths, development areas, conduct, attendance if supplied, and reason for leaving if supplied. 2. A table with each material claim, the exact supporting wording from the reference, and whether it is supported, unclear or absent. 3. Any inconsistencies, unexplained gaps or answers that need a human follow-up, without treating an omission as evidence of wrongdoing. 4. Six neutral follow-up questions for the referee, including questions about any specific discrepancy you identify. 5. A short record of what this reference does and does not establish. Keep the candidate's and referee's information confidential. Do not infer protected characteristics, diagnose behaviour, score personality, or recommend reject or hire. Do not present a reference as verified unless the supplied material shows how it was verified. Use plain UK English.
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
What it gets wrong
- AI cannot prove that a referee is who they claim to be or that a reference was written by them.
- AI cannot contact a referee, interpret hesitation or resolve an ambiguous answer through a real conversation.
- AI cannot decide whether a negative comment is relevant, fair and proportionate to the role.
- AI cannot take responsibility for a discriminatory or unlawful hiring decision.
- AI can make a polished summary look more certain than the source material supports.
What caps this at PARTLY: private data access, verification cost and legal accountability.
How we scored this
Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.
| Axis | Score (0–2) |
|---|---|
| Output | 1 |
| Inputs | 1 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 1 |
| Total | 5 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can AI check candidate references?
- Partly. AI can summarise written references, compare answers with the questions asked and draft follow-up questions, but a person still has to obtain consent, contact the referee and check disputed claims.
- Can ChatGPT contact a candidate's referees?
- Not by itself in a reliable recruitment process. You need to contact the referee through your approved process, confirm consent and identity, and then use AI to organise the written responses if appropriate.
- Is it legal to use AI for reference checks in the UK?
- It depends on how you collect, use and retain the personal data and how the hiring decision is made. Use a consistent process, obtain the required consent and get HR or employment-law advice for a serious case.
- What can AI do in a reference check?
- It can extract factual claims, show which answers are supported by the supplied text, flag gaps and draft neutral follow-up questions. It cannot verify the referee's truthfulness or make the final hiring decision.
Nearby answers
Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.
The newsletter
AI news, new answers and product picks, straight to your inbox.