The reader problem
Candidates should not have to infer how a recruitment process works from a generic statement such as “we use AI”. That wording leaves important questions unanswered: where is AI used, what information does it handle, who checks the output, and who makes the decision?
This guide gives a practical transparency pattern for recruiters. It is operational guidance, not legal advice. Confirm the notice, lawful basis, retention, access, accommodation, and objection process with your privacy and employment advisers for each market.
Three questions a useful notice answers
- Where and why? Name the recruitment stage and the narrow purpose, such as organising CV evidence against agreed role requirements.
- What happens next? Explain what the tool produces, what it cannot establish, and which reviewer checks the source material.
- How can I respond? Give a monitored contact route for questions, corrections, accessibility needs, and requests for human clarification.
A four-part transparency framework
1. Name the support, not a mysterious system
Use plain words: “We use software with AI-assisted features to help our recruitment team organise information in application documents.” Avoid naming a capability the product does not provide, such as automatic hiring or a guaranteed ranking. Identify the stage (for example, CV review) and the purpose (for example, finding evidence related to the published requirements).
2. Describe inputs and outputs at the right level
Tell candidates which application materials are in scope and what reviewers see: extracted facts, requirement-level evidence, gaps, or questions for follow-up. Do not imply that an output is a fact about the person. Explain that an omission may be unresolved evidence and that a reviewer can inspect the source document.
3. Make the human decision boundary explicit
State the actions that require an authorised person: assessing evidence, deciding progression or rejection, recording a reason, and sending candidate communication. If a workflow has an exception or a separate automated step, document it accurately rather than relying on a general “human in the loop” phrase.
4. Offer a usable response path
A transparency notice needs an owner and a route. Provide a monitored email or form, the information a candidate should include, and what will happen next. Route correction requests to the team that can inspect the application and route privacy or rights requests to the responsible privacy contact. Publish an accessibility or accommodation route where relevant.
Worked example
Weak: “Applications may be processed by AI.”
Clearer: “Our recruitment team uses AI-assisted software during CV review to organise information from your application against the role requirements. It may highlight evidence, gaps, or follow-up questions for a reviewer. A recruiter checks the source material and makes the decision about progression; the software does not send candidate messages or make a hiring decision on its own. If information is incorrect, you need an accommodation, or you want to ask how this stage works, contact [the named recruitment contact].”
Keep the example aligned with the actual workflow. Do not promise an appeal, explanation, or deletion outcome that the operating process cannot deliver.
Candidate-transparency checklist
- The notice names the stage and purpose of AI support.
- Inputs and the type of output are described without presenting inferences as facts.
- Human review, progression, rejection, and communication boundaries are named.
- The correction, question, privacy, and accommodation routes are monitored and owned.
- The wording is localised and understandable to the candidates who will read it.
- Product, privacy, employment, and legal owners have reviewed the current workflow.
- The notice is versioned and checked whenever the workflow or provider changes.
Skilltage guidance and limits
Skilltage guidance: make requirement-level evidence inspectable, preserve visible uncertainty, and keep candidate progression and communication behind an explicit recruiter action. This supports a more understandable review process; it does not establish legal compliance, prove fairness, or replace a privacy notice, impact assessment, accommodation process, or hiring judgement.
AI-assisted review can still reproduce incomplete source documents, inconsistent requirements, or reviewer errors. Monitor the real workflow, document material changes, and provide a human route for candidates who need clarification or correction.
Next step
Take the next role you are hiring for and draft a five-sentence notice using the four-part framework. Ask the product owner and privacy reviewer to compare every sentence with the actual workflow before publishing it. Then review the underlying evidence boundary in the evidence-first CV screening guide.
Related resources
Continue with structured hiring requirements and how to evaluate AI recruitment tools before changing a screening workflow.