Can You Use AI in a Job Interview? Rules and Exceptions

Before using AI during a job interview, check whether that specific use is permitted in that specific round. Private practice, live answer generation, and an explicitly AI-enabled assessment are different activities. Permission for one does not automatically cover the others.

This distinction matters whether you are applying in Europe or North America. A company's location, its enthusiasm for AI, or the software used for the interview is not a reliable substitute for its candidate instructions.

The aim is to demonstrate your experience and judgement within the agreed assessment conditions. Prepare to answer independently, clarify ambiguous rules, and keep accessibility support separate from tools that supply answers for you.

What does “using AI in an interview” actually mean?

The phrase can describe three very different situations: an employer using an automated interviewer, a candidate practising beforehand with a simulator, or a candidate receiving assistance while being assessed.

An automated interviewer does not, by itself, authorise you to generate responses with another tool. Equally, rehearsing your own experience before the meeting is not the same as having software answer a live question.

For the employer-led format, read what to expect from an AI screening interview. This guide focuses on what you, the candidate, may use.

What published employer policies show

The policies below were checked on 18 October 2026. They are employer examples, not rules for every workplace in the countries where those employers recruit.

SAP: its candidate guidance distinguishes preparation from assessed performance. It prohibits real-time AI assistance in video, phone, and recorded interviews, and describes AI support during online assessments as cheating. Its guidance also rejects invented qualifications and replacing your own input with generated responses.

Instacart: its guidance allows preparation but prohibits AI in live interviews unless explicitly requested as part of a skills presentation. It identifies a specific exception in its senior software engineer process: the Karat Next Gen round, where AI use is allowed and encouraged. That exception does not extend to every round.

St. James's Place: its early-careers guidance accepts support with ideas, structure and preparation, while warning against fictional experience, generated whole responses and formulaic scripts.

The practical conclusion is narrower than “AI is banned” or “AI is now expected.” Check the policy and invitation for the role, location and stage you are actually attending. Where they appear to conflict, ask the recruiter to resolve the difference before proceeding.

Use a stage-by-stage checklist

“Open book” does not necessarily mean unrestricted AI use. Permission to research a problem is not automatically permission to submit a generated solution.

Ask a precise question, not “Is AI okay?”

Start with the candidate guide, invitation, assessment brief, and any confidentiality instructions. Then ask about the tool's actual function. Brainstorming before the interview, transcribing speech, and generating answers are not interchangeable.

“Before the [round name], could you confirm which tools are permitted? I would like to clarify the rules for [specific function, such as coding suggestions or internet research]. Are there restrictions on the tool or on sharing the assessment material, and is any disclosure required?”

Keep the answer with your preparation notes. Ask again when the instructions for a later round differ; a recruiter allowing research for a take-home exercise has not necessarily authorised assistance during the follow-up interview.

If clarification has not arrived by the start of the session, raise the question before the assessed work begins. Do not switch on unapproved help and try to settle the rules afterwards.

Use practice to make yourself less dependent on assistance

Begin with a question relevant to the job and answer it yourself. Reviewing an AI-generated model answer before attempting your own can leave you practising someone else's wording rather than retrieving your experience.

Use a simple rehearsal loop:

1. Answer unaided. Explain the situation, your responsibility, your decisions, and the result.

2. Find the weak point. Was the answer vague, too long, or unclear about what you personally did?

3. Improve the structure without changing the facts. Reject invented metrics, responsibilities, or expertise.

4. Try an unfamiliar follow-up. Explain a trade-off, a mistake, or what you would do differently without reading a script.

The last step tests whether you know the experience, not just the paragraph. The STAR framework can help organise the story while leaving the wording flexible.

When interviewing in a second language, rehearse in that language too. Simplify sentences until you can explain the same point naturally. Improving your preparation is different from using a live translation or answer-generation tool in an assessed language task; clarify those rules separately.

What to do when AI is explicitly permitted

An AI-enabled round still requires you to understand the task. Ask which tools and accounts are approved, whether external services may receive the material, and how the assessor wants you to explain your contribution.

For a coding exercise, inspect generated changes, check assumptions, run suitable tests, and explain the choices you accept or reject. If a suggested implementation treats missing values as zero, for example, establish whether that behaviour fits the requirements before building on it.

The same principle applies outside software. For a business case, distinguish supplied data from assumptions. For a presentation, check every claimed fact. A confident-looking output is not evidence that the underlying answer is correct.

Do not hide where the assistance mattered. Explaining how you evaluated an output is more informative than implying you produced it unaided when the task expressly allowed tools.

Take-home assignments need their own rules

Read the brief before starting. Check whether the employer treats proofreading, brainstorming, research, and generation differently. Do not assume that a tool commonly used in your current job is allowed in another employer's assessment.

Where disclosure is requested, explain the actual contribution:

“I used [approved tool] to suggest test cases. I wrote the implementation, selected the relevant cases, checked their expected results, and removed suggestions that did not match the specification.”

Use that description only if it is true. Disclosure does not retroactively authorise a prohibited activity. Keep enough notes to explain your process, but do not retain confidential assessment materials beyond what the employer permits.

The guide to preparing for a work-sample assessment covers the wider task: clarifying scope, prioritising the work and discussing your decisions.

Handle accessibility, transcription and recording separately

Captions, screen readers and speech-to-text should not be casually equated with a tool that supplies interview answers. Contact the recruitment or accessibility team about the function you need and the available arrangements.

Describe the support rather than disclosing unnecessary personal details. For example: “I need accurate captions to follow the conversation. Does the platform provide them, or can we agree on an alternative?”

Check whether a proposed tool also records, summarises, uploads, or generates content. An arrangement for captions does not automatically settle those separate functions. Do not add an unapproved recording bot to the meeting simply because it was available in your calendar software.

Employer instructions do not replace applicable privacy or accessibility law. Those obligations depend on the jurisdiction and circumstances; obtain local guidance for a legal dispute rather than relying on a general interview checklist.

Protect confidential information during preparation

Use a public job description and an anonymised account of your experience where possible. Remove customer names, private financial details, restricted source code and other material you are not authorised to share.

For a cross-border hiring process, check the actual service and data-handling instructions rather than assuming the recruiter's location tells you where a tool processes information. Permission to use AI and permission to upload the employer's documents are different questions.

When you do not know an answer

A difficult question is not a reason to introduce unapproved help. Ask for clarification, explain what you know, and identify where your knowledge ends.

“I have not used that exact system. In a similar situation, I would first check [relevant factor]. Could I clarify whether the main constraint is [A] or [B]?”

Then give a relevant response, rather than using uncertainty to avoid the question. See how to recover when you do not know an interview answer.

Common questions

Can I use AI because the employer uses it?

That does not establish permission. Follow the candidate instructions for the specific assessment, and ask about anything unclear.

Must I disclose private interview practice?

Follow any stated disclosure requirements and answer questions about your preparation honestly. There is no need to invent a universal disclosure rule for every employer and jurisdiction.

Does using AI at work mean it is allowed in the interview?

No automatic permission follows. A team can use tools in its work while assessing particular skills without them, or run a separate round that explicitly includes them.

Practise before the real assessment

Lyrra's role-specific mock interviews let you practise in text or voice with job-description context and review feedback. Use that rehearsal to improve the answers you can give independently. It does not change an employer's rules or predict the hiring decision.

The useful boundary is straightforward: know the permitted assistance, protect the material you use, and be able to explain the work and experience you present.

Sources and policy checks

Employer guidance checked on 18 October 2026: SAP's ethical use of AI in recruitment; Instacart's candidate AI-use guidelines; and St. James's Place's early-careers guidance. These examples do not establish one policy for Europe or North America. Recheck the instructions for your own role and assessment.