Home · Business · HR & People · Interviewing & assessment

PARTLY

As of 13 August 2026, AI can only partly analyse a candidate's sentiment in an interview.

Most people should hand this to a purpose-built tool.

Can you do it?

5 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 costsThe available tool data gives no price for a comparable human interview-assessment service.

If this goes wrong, a model treats nerves, accent, disability, culture or a communication style as negative sentiment and a suitable candidate is unfairly disadvantaged.

What to actually do

  1. Use a tool built for this

    The route this page recommends

  2. Do it yourself

    Second choice

    A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

    How to actually do it

    1. Get the candidate's informed consent for recording or transcription, and check your organisation's retention, access and equality procedures before processing the interview.
    2. Open a transcription tool such as Otter, or use an existing accurate transcript, and label each speaker and interview question clearly.
    3. Gather the role description, the questions asked and the published assessment criteria, removing personal information that is not needed for this analysis.
    4. Paste the context and consented transcript into a chatbot with the prompt above, and ask it to return evidence, alternative explanations and confidence levels rather than a hiring recommendation.
    5. Compare every claimed sentiment signal with the candidate's actual words and the surrounding question, correcting any claim that relies on accent, appearance, disability, nerves or an inferred private feeling.
    6. Ask a colleague trained in fair interviewing to review the transcript-based notes, then keep the sentiment analysis separate from the structured evidence used for the hiring decision.

    Prompt

    Analyse the candidate's sentiment in this interview transcript. Treat sentiment as the language and tone expressed in the interview, not as proof of the candidate's private feelings, personality, honesty, suitability or protected characteristics. Do not infer age, race, sex, disability, religion, health, pregnancy, sexual orientation or any other protected characteristic. Do not make or recommend a hiring decision. Separate observable evidence from interpretation. For each conclusion, quote the relevant words, identify the interview question or topic, give a confidence level, and list plausible alternative explanations such as nerves, unfamiliarity with the format, accent, communication style or accessibility needs. Note where the transcript is insufficient and where audio or video would be needed, without treating those signals as reliable evidence of suitability. Finish with a neutral list of follow-up questions that would clarify the candidate's answers. Interview context: [ROLE, QUESTIONS AND SCORING CRITERIA]. Transcript: [PASTE CONSENTED TRANSCRIPT HERE].

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

  3. Hand it to a person

    The distant third

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

What it gets wrong

  • AI cannot know whether a candidate is genuinely enthusiastic, anxious, tired or masking another feeling.
  • AI cannot reliably separate nerves, accent, disability, culture and communication style from sentiment.
  • AI cannot establish that sentiment is a fair or relevant criterion for the role.
  • AI cannot obtain valid consent, decide appropriate retention or take responsibility for discriminatory use.
  • AI cannot replace a structured interview assessment owned by a trained human panel.

What caps this at PARTLY: judgement under ambiguity, consent and privacy and stakes of error.

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.

AxisScore (0–2)
Output2
Inputs1
Verification1
Liability1
Effort delta1
Total6 / 10

The methodology and its thresholds are published in full.

FAQ

Can AI tell if a candidate is nervous in an interview?
It can describe words and patterns that may look uncertain or hesitant, but it cannot establish that the candidate is nervous. Treating those patterns as evidence against the candidate can disadvantage people because of disability, culture, accent or interview conditions.
Can employers use AI to analyse interview sentiment?
They can use it for a limited, evidence-based summary if they have a lawful and transparent process, appropriate consent and human oversight. It should not be used as a hidden score or as the basis for an automated hiring decision.
Is AI sentiment analysis fair in recruitment?
Not by itself. Sentiment is ambiguous and can reflect communication style or the interview setting rather than job-related ability, so a trained human must decide whether any evidence is relevant and apply the same criteria to every candidate.
Can AI decide whether a candidate is suitable from an interview?
It can organise answers against criteria that you define, but it should not decide suitability from sentiment or replace a human hiring decision. This is not professional advice, and a serious employment case needs an employment solicitor or qualified HR professional.

Nearby answers

All Interviewing & assessment rulings

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.