Short answer
A good candidate AI policy says, stage by stage, where AI help is welcome and where you need the candidate's own unaided work. The Australian Public Service Commission's 2026 principles are a useful model. They allow AI for refining a CV and practising interviews, but not for generating answers to real interview questions or written assessments.
Key takeaways
- About four in ten job candidates said they used AI during the application process (Gartner, 2025).
- The APSC now welcomes candidate use of AI in public service recruitment unless an agency excludes it, as long as AI supports rather than replaces the candidate's own experience (APSC, 2026).
- Generating answers to actual interview questions, or generating a written assessment, is listed as unacceptable AI use (APSC, 2026).
- Only 26% of candidates trust AI will fairly evaluate them, so a fair policy also explains how the employer uses AI (Gartner, 2025).
- A rule is easiest to keep fair when every candidate sits the unaided stages under the same conditions.
Why do employers need an AI use policy for candidates?
Because most candidates now use AI somewhere in their search, and without a clear rule each one has to guess what you'll accept. In a Gartner survey of 3,290 job candidates in the fourth quarter of 2024, 39% said they used AI during the application process, and 29% of those used it to write answers to assessment questions.
Australian figures look similar. In Capterra's Job Seeker AI Survey of 250 Australian job seekers in July 2024, 73% were using AI tools in their job search. Of all Australian respondents, 22% had used AI to generate answers to interview questions and 27% had used it on a test assignment or skills assessment.
Guessing creates its own pressure. In the same survey, 88% of respondents thought it likely that other applicants were using AI to embellish their applications. A written policy takes that guesswork away and tells every candidate the same thing. You can see more of the numbers in How common is AI use in remote job interviews?
What should a candidate AI policy include?
It should name the stages where AI help is welcome and the stages that need the candidate's own unaided work. It should also say what happens if a concern comes up, and be open about how you use AI yourself.
- Where AI help is welcome, such as polishing a CV or practising for an interview
- Which stages need unaided work, such as live interviews and timed assessments
- Whether you'll ask candidates to disclose AI use, and that saying yes won't count against them
- That every claim must be true and something the candidate can explain in their own words
- A reminder not to paste confidential information from a current or former employer into AI tools
- How your organisation uses AI to screen or assess applications
- What happens if there's a concern, and how a candidate can respond
- Who to contact about adjustments, including assistive technology
The point about your own AI use matters more than it looks. Gartner found only 26% of 2,918 candidates surveyed in the first quarter of 2025 trust AI will fairly evaluate them, and 25% said they trust employers less if they use AI to evaluate their information. Being open in both directions makes the policy feel fair rather than one-sided.
What AI use is acceptable at each stage?
A sensible line is that AI can help a candidate present their own experience, but shouldn't produce the answers you're assessing. The Australian Public Service Commission's Principles for candidate use of AI in recruitment, published in April 2026, set this out stage by stage.
| Stage | AI help that's acceptable | AI use that isn't |
|---|---|---|
| Application and CV | Refining drafting, checking spelling and grammar, researching the employer | Inflating or inventing experience, pasting in generic AI text, entering confidential workplace information |
| Interview preparation | Practising with mock interviews | Generating answers to the actual interview questions |
| Written assessment or work sample | Checking spelling and grammar, researching public information | Generating the assessment or work sample itself |
Adjust the table for the job. If the role uses AI every day, you might allow it in a work sample and assess how well the candidate uses it. The APSC's guide for agencies suggests deciding what level of AI capability a role needs and how it will be assessed. Whatever you choose, say so before the stage starts.
What does fair policy wording look like?
Plain and welcoming. Candidates read a policy as a signal of how they'll be treated, so lead with what's allowed before what isn't. Here is sample wording you can adapt.
- You're welcome to use AI tools to help write and check your CV and application, and to practise for interviews.
- Please make sure everything you tell us is true, and that you can talk us through any example in your own words.
- Our live interviews and the timed skills assessment are designed to show your own thinking, so please complete them without AI or outside help.
- We may ask whether you used AI in your application. Saying yes won't count against you.
- We use [describe any AI screening tools] and a person reviews every shortlisting decision.
- If you need an adjustment, including assistive technology, contact [name] and we'll work it out with you.
Public service employers can link to the APSC principles directly. The APSC says agencies must give candidates a link to them in the candidate pack or job advertisement. See our notes on public sector hiring.
Should candidates have to disclose AI use?
Asking is reasonable, as long as the answer is treated neutrally. The APSC guide includes a sample transparency question that asks whether AI was used in any part of the application and, if so, which parts and which tools.
The same guide says a disclosed use of AI "shouldn't be viewed as a negative". What matters is whether the candidate can show relevant experience through real examples. It suggests looking past the polish and checking claims through targeted interview questions and specific reference checks, rather than judging how well an application reads.
What should happen if a candidate doesn't follow the policy?
Respond in proportion, and write it down. The APSC guide suggests telling the candidate about the concern and giving them a chance to respond before any decision is made. Their progress can be paused while that happens.
Removal from the process is for cases where the candidate gave false or misleading information. Everything else can usually be settled with follow-up questions or another short assessment. A fair process protects the candidate as much as it protects you, and it's far easier to defend later.
How do you keep an unaided stage fair for everyone?
Give every candidate the same conditions for that stage. A policy alone relies on trust, and software can't reliably spot AI help, as we explain in Can software detect AI use in a remote interview? Honest candidates then compete with anyone who quietly ignores the rule.
An in-person interview solves this, but it's costly when candidates are spread out. A supervised venue is the middle path. At SmartVue, staff check government photo ID and personal devices go into lockers before the session. The pod computer runs only your interview platform while your panel joins remotely, and each session ends with an integrity report.
Our venue is 1Smart.Space in Docklands, Melbourne, a short walk from Southern Cross Station. If you hire elsewhere, tell us where you hire and we'll confirm availability. Read how a supervised interview pod works or see pricing.
Frequently asked questions
Can job candidates use AI to write their CV?
Many employers allow it. The Australian Public Service Commission's 2026 principles accept AI for refining the drafting of a CV and checking spelling and grammar, as long as every claim is true and reflects the candidate's own experience.
Is it acceptable to use AI during a job interview?
Usually not, unless the employer says so. The APSC lists generating answers to actual interview questions as unacceptable AI use, while practising with AI mock interviews beforehand is acceptable.
Should employers ban candidates from using AI altogether?
A blanket ban is rarely needed and is hard to apply fairly. A clearer approach is to welcome AI for preparation and name the specific stages, such as live interviews or timed assessments, that need the candidate's own unaided work.
Should a candidate who discloses AI use be marked down?
No. The APSC guide for agencies says disclosed AI use shouldn't be viewed as a negative. What matters is whether the candidate can demonstrate relevant experience through real examples.
Where can candidates sit an unaided interview under supervision?
SmartVue's supervised pods are at 1Smart.Space, Level 2, 757 Bourke Street, Docklands, Melbourne, a short walk from Southern Cross Station, open Monday to Friday from 7am to 7pm. For other locations, employers can tell SmartVue where they hire to confirm availability.
Sources
- Australian Public Service Commission, "Principles for candidate use of AI in recruitment", April 2026
- Australian Public Service Commission, "Guide for managing candidate use of AI in recruitment", April 2026
- Australian Public Service Commission, "AI in recruitment", published 20 April 2026, updated 29 April 2026
- Gartner, "Gartner Survey Shows Just 26% of Job Applicants Trust AI Will Fairly Evaluate Them", 31 July 2025
- Capterra Australia, "How recruiters can identify AI use in the recruitment process", Job Seeker AI Survey, July 2024
