Transparency starts before the shortlist
Many employers think candidate transparency means a privacy notice at the bottom of the careers site. Legally, that may be a starting point. Practically, it is too late and too abstract.
A candidate should not discover only after rejection that an AI tool filtered CVs, produced matching scores or prioritised answers to knockout questions. Transparency should sit inside the recruitment flow: job ad, application form, process information, human review and correction route.
That is not only better from a legal perspective. It is also better for trust. Candidates are more likely to accept AI when they understand what the tool does and does not do, who makes the final decision and how they can ask questions.
First clarify the role of AI
Not every AI role is the same. A chatbot answering FAQs is different from a model that ranks candidates. A scheduling tool proposing interview slots is different from a system that evaluates interview answers.
Use three internal levels:
- AI as administrative support: scheduling, summarising, draft emails.
- AI as process support: CV parsing, skill extraction, matching suggestion.
- AI as decision support: ranking, shortlist advice, rejection suggestion.
The closer AI gets to the decision, the more concrete candidate communication should be.
What should a candidate understand?
Good candidate communication does not need to be a technical whitepaper. It does need to answer five questions:
- Is AI used in this process?
- What is AI used for?
- Which data may be processed?
- Does a human make the final decision?
- Where can the candidate ask questions or request correction?
Avoid vague lines such as: "We may use automated technology to improve your application." That says nothing. A better version:
"We use software that helps structure CVs and highlight relevant experience for recruiters. The software does not make final hiring or rejection decisions. A recruiter reviews the shortlist and can correct the suggestion."
That text is short, understandable and verifiable.
The four places where transparency belongs
1. Job ad or careers page
Briefly explain that AI-supported software may be used in the application process. Keep it concrete and avoid legal overload.
Example:
"In this application process, we use AI-supported software to structure applications and help recruiters find relevant experience faster. Final assessment is done by people."
2. Application form
When the candidate submits data, it should be clear what happens with it. This is the right place for short process information plus a link to privacy details.
Example:
"Your CV and answers may be automatically analysed to structure relevant information. Our recruiters use this output as support and review the assessment themselves."
3. Candidate email or status page
When AI plays a clear role in screening or matching, a short explanation in the confirmation email helps.
Example:
"After receipt, your application is first structured in our ATS. A recruiter then reviews the match with the role requirements. You can contact us if information has been processed incorrectly."
4. Rejection or feedback moment
Not every rejection needs a technical report. But if a candidate asks about the role of AI, the organisation should be able to explain which step AI supported and that a person reviewed the decision.
Transparency without exposing the model
Employers sometimes fear that transparency means publishing the model, vendor or full scoring logic. That is usually not the right approach.
The candidate primarily needs understandable process information. You do not need to publish every parameter. You do need to be honest about the role of AI and human review.
A good balance:
- explain the function of the AI;
- explain which types of data matter;
- state what the AI does not do;
- state who reviews the output;
- provide a route for questions or correction.
Connect transparency to human oversight
Transparency is weak if there is no real human control internally. You can tell candidates that a recruiter reviews the output, but then that recruiter must know how to do that.
Candidate transparency therefore belongs together with:
- recruiter training;
- review instructions;
- override logging;
- bias monitoring;
- complaint and correction process.
If the recruiter cannot explain why a candidate did or did not progress, the transparency text becomes false comfort.
What belongs in the evidence pack
For HR-AI transparency, keep at least these documents:
- candidate notice;
- privacy text;
- process description;
- vendor information about system output;
- human oversight instruction;
- example of shortlist review;
- escalation and correction route;
- recruiter training record.
These documents do not all need to be public. They do need to be available internally when legal, privacy, worker representation, a customer or a regulator asks questions.
The practical route for employers
Start small:
- inventory AI in the recruitment flow;
- determine whether each step is administrative, process-supporting or decision-supporting;
- write candidate-friendly wording for each step;
- check whether the wording matches the real workflow;
- train recruiters on candidate questions;
- record corrections and overrides;
- review the wording after every vendor or workflow change.
Embed AI helps employers, staffing firms and HR-tech vendors build this layer inside the HR-AI Risk & Evidence Sprint. For a quick first view, start with the AI Act Gap Intake.
Final note
Candidate transparency is not a legal footnote. It is part of a fair recruitment process.
When you clearly explain where AI assists, where people review and how candidates can ask questions, you reduce compliance risk and build trust in a selection process that is becoming more digital every year.
