AI-Generated Interview Questions: How to Use AI to Design Better Assessments

The Problem with How Most Teams Design Interviews

In most organisations, interview questions are invented on the spot. A hiring manager opens the CV ten minutes before the interview, reads it quickly, and asks whatever comes to mind. The questions are based on pattern recognition from past hires, personal preferences, and whatever the interviewer finds interesting — not on a systematic assessment of what the role actually requires.

This approach is not just inefficient. It is expensive. Unstructured interviews predict job performance at about half the rate of structured interviews. Organisations that rely on unstructured, improvised interviewing make more bad hires, and bad hires are costly — estimates range from 30% to 200% of annual salary in total cost including recruitment, training, productivity loss, and eventual replacement.

The root cause is not that hiring managers are bad interviewers. It is that designing good interview questions is a skill that most people have never been taught, and the process of doing it well — defining competencies, mapping questions to each competency, calibrating difficulty and depth, ensuring legal compliance — is time-consuming enough that most organisations skip it.

AI solves the time problem. Whether it solves the quality problem depends on how you use it.

What AI Does Well in Interview Question Design

Understanding AI's genuine strengths — and its real limitations — in interview design prevents two opposite mistakes: rejecting AI as a gimmick, and treating its output as production-ready without review.

Generating Role-Specific Questions at Scale

Given a job description or a list of required competencies, a well-prompted AI model can generate twenty to thirty relevant interview questions in seconds. This replaces the blank-page problem that paralyses many hiring managers. Even if you discard half the generated questions and rewrite several others, you have compressed two hours of interview design into fifteen minutes of review and refinement.

Covering Question Type Diversity

Good interview question sets include multiple types: behavioural (past behaviour as predictor), situational (hypothetical scenarios), technical (domain knowledge), and motivational (what drives this person). Most hiring managers naturally gravitate toward one or two of these types and neglect the others. AI generates across all types systematically, producing a more balanced assessment.

Prompting Competency-Level Thinking

When you ask an AI to generate interview questions for a "senior product manager" role, the output naturally covers the competencies typically required at that level: stakeholder management, data-driven decision-making, cross-functional influence, prioritisation under constraint. This prompts hiring managers to think about competency coverage in a way that improvised interviewing never does.

Reducing Individual Bias in Question Design

Questions invented by a single hiring manager often reflect that manager's own background and preferences. An engineer who came from a startup will design questions that favour candidates from startups. AI generates questions based on role requirements rather than interviewer experience, reducing this source of structural bias in the assessment design itself.

What AI Does Not Do Well: The Limitations to Know

AI-generated interview questions are a starting point, not a finished product. Several limitations require human oversight:

Generic Rather Than Context-Specific

AI generates questions for "a senior marketing manager role" in general, not for your specific company's product, market position, team dynamics, or strategic challenges. Questions that work for a B2C consumer goods brand may be irrelevant for a B2B SaaS company. Human review is required to contextualise generic output.

No Understanding of Organisation Culture

Culture fit questions — assessing how a candidate's working style aligns with your team's way of operating — require understanding of your actual team. AI cannot generate accurate culture questions without detailed information about your organisation, and even then, the output needs careful validation.

Legal Compliance Risk Varies by Jurisdiction

Employment law varies significantly across European countries. Questions that are legally permissible in one jurisdiction may be prohibited in another. AI does not automatically enforce local legal requirements, and questions about topics that could indirectly reveal protected characteristics (family plans, health, nationality) can appear in AI output. Every question set generated by AI needs legal-compliance review before use.

The Practical Workflow: From AI Draft to Assessment Tool

The following workflow integrates AI into interview design while preserving the quality controls that make the output genuinely useful:

Step 1: Define Your Competency Framework for the Role

Before engaging AI, identify the four to six core competencies that the role requires. For a senior engineer, these might be: technical depth in specific domains, architectural thinking, cross-functional communication, mentoring capability, delivery under pressure, and self-direction. For a sales director: pipeline management, executive relationship building, team leadership and coaching, commercial negotiation, and market strategy.

This step is not replaceable by AI. It requires human understanding of what the role actually demands and what distinguishes high performers from average performers in your specific context.

Step 2: Prompt AI with Role and Competency Context

Provide the AI with: the job title, a summary of the role's key responsibilities, the specific competencies you want to assess, the seniority level, and any context about your industry or company type. The more specific the input, the more targeted the output. A generic "generate interview questions for a product manager" prompt will produce generic output. A prompt that includes the specific product domain, team structure, and competency priorities will produce substantially better questions.

Step 3: Review and Edit for Relevance and Depth

Review all generated questions against three criteria: Is this question genuinely discriminating between strong and weak candidates? Is it specific enough to the role to elicit useful information? Is it legally compliant? Remove questions that fail any of these criteria. Rewrite questions that are conceptually right but too vague or too leading. Add questions that the AI missed based on your knowledge of the role.

Step 4: Build a Scoring Rubric for Each Question

The question is only half the structured interview. The other half is the scoring rubric — the description of what a strong, adequate, and weak answer looks like. Without rubrics, interviewers default to subjective impressions rather than consistent assessment. AI can help generate draft rubrics as well, though these require even more human editing than the questions themselves to reflect what "good" actually looks like in your context.

Step 5: Store in Your ATS Question Bank

The question set and rubrics should be stored in your ATS, attached to the role or the competency framework. This creates reusability — the next time you hire for the same or a similar role, the question set is already available. It also ensures consistency: all interviewers conducting first-round interviews for the same role ask the same questions, enabling meaningful comparison across candidates.

Treegarden's question management module allows you to build and tag question banks by role type, competency, seniority level, and interview stage, making it easy to assemble a customised interview guide for each new role from a library of validated questions.

Behavioural vs. Situational Questions: When to Use Each

The two most valuable question types for structured interviews serve different purposes, and AI should generate both:

Behavioural Questions

Behavioural questions ask about past experience: "Tell me about a time when..." They are based on the principle that past behaviour is the best predictor of future behaviour. They require candidates to provide specific examples from their experience, making the answers verifiable and difficult to fabricate in detail.

Behavioural questions are most valuable for assessing competencies that are developed through experience: leadership, conflict management, customer relationship skills, project delivery. They work less well for assessing potential in candidates who are making a career transition and genuinely lack the specific experience being asked about.

Situational Questions

Situational questions present a hypothetical scenario: "What would you do if..." They assess judgment, problem-solving approach, and values — rather than past experience. This makes them more appropriate for roles where you are hiring for potential or where the specific technical context differs from the candidate's background.

The risk with situational questions is that candidates can give theoretically correct answers that do not reflect how they would actually behave. For this reason, experienced interviewers follow situational questions with behavioural probes: "You've described what you'd ideally do — can you tell me about a time when you actually faced a similar situation?"

AI Personalisation: Questions Based on the Candidate's Profile

Beyond generating standard question sets, more advanced AI integration in ATS platforms enables per-candidate question personalisation. The AI analyses the specific candidate's CV and generates targeted follow-up questions based on what is — or conspicuously is not — on their profile.

Examples of AI-personalised probes:

• The candidate claims to have "led" a digital transformation project. AI suggests: "What was your specific decision-making authority on the project, and what decisions required escalation?"
• The candidate has a two-year gap in their employment history. AI suggests: "Could you walk me through how you used the period between [year] and [year] and what you were working on during that time?"
• The candidate switched industry three times in five years. AI suggests: "What specifically drove each of your industry transitions, and how did you evaluate the decision at the time?"

These personalised probes are not questions the structured interview would include — they are additions specific to this candidate's profile. Combining them with a standardised base question set preserves comparability while allowing interviewers to explore individual circumstances.

Building a Question Library: Long-Term Value

The greatest return from AI-assisted interview design comes not from the first question set you generate, but from the question library you build over time. As each role's question set is refined through use — interviewers flag which questions generated the most useful answers, which were too easy to rehearse, which candidates consistently misunderstood — the library improves iteratively.

Over 12 to 24 months of structured use, a well-maintained question library in your ATS becomes a proprietary asset: a collection of validated, role-specific, legally reviewed questions that reflect your organisation's specific context and standards. This is something AI alone cannot provide — it requires the combination of AI speed and human refinement over time.

Frequently Asked Questions

Are AI-generated interview questions reliable?

AI-generated questions are a starting point, not a final product. They are highly reliable at generating a broad range of role-relevant questions quickly. However, they need human review to ensure they reflect the specific context of your organisation, align with your company's competency framework, and do not contain legally problematic content. Used correctly — as a draft that a recruiter or hiring manager reviews and customises — AI-generated questions are significantly better than questions invented on the spot during an interview.

What is a structured interview?

A structured interview uses the same set of predetermined questions for every candidate applying to the same role, assessed against the same rating scale. This consistency makes structured interviews far more predictive of job performance than unstructured conversations — research consistently shows structured interviews are twice as predictive as unstructured interviews at determining future job performance. AI accelerates the creation of the question set but the discipline of using it consistently is where the value is realised.

How does AI reduce bias in interviews?

AI reduces bias in several ways. First, it generates questions based on role requirements rather than interviewer preferences, preventing the tendency to ask questions that favour familiar backgrounds. Second, structured question banks ensure all candidates are assessed against identical criteria, rather than each interviewer making up questions that reflect their own experience. Third, AI can flag questions that may introduce bias — gender-coded language, questions about personal circumstances, or culturally specific scenarios that may disadvantage some groups.

What types of interview questions should AI generate?

The most effective AI-generated interview question sets combine: behavioural questions (Tell me about a time when...) that assess past behaviour as a predictor of future performance; situational questions (What would you do if...) that assess judgment in role-relevant scenarios; technical questions that assess specific domain knowledge; and competency questions mapped to the organisation's leadership or functional competency framework. Each question type serves a different assessment purpose and they should be used in proportion to the role's requirements.

Can AI personalise interview questions per candidate?

Yes. Advanced ATS platforms with AI integration can analyse the specific CV and application of each candidate and suggest personalised follow-up questions based on what is on their profile. For example, if a candidate claims expertise in a specific technical area, the AI might suggest probing questions about that specific area. This preserves structured assessment while allowing interviewers to explore candidate-specific details that standard questions would not surface.