How to Detect AI-Generated Answers in Job Interviews

When you’ve put time and effort into building a solid recruitment process, it can feel destabilising when you get a candidate all the way to the interview stage and realise they may not be providing you with authentic information. With AI tools becoming more accessible, their influence is starting to show more in hiring conversations - recognising and adapting to this shift is key to protecting the integrity of your process. Let's see how you can spot candidates using AI and take steps to address it.

1. Understanding AI in Interviews

1.1 How do candidates use AI to generate answers?

Candidates may use AI-generated responses to appear more knowledgeable or articulate during online interviews. Typically, they leverage tools like ChatGPT to draft answers to common interview questions, or use speech-to-text systems to provide them with answers to questions live during the interview. This allows them to present responses that are well-structured and technically sound, even on the spot.

1.2 How does AI integration affect interviews?

When used by candidates, AI poses challenges by introducing answers to questions that might not truly reflect their capabilities. This can result in misalignments in skills assessments, leading you away from the best fit for your roles.

Moreover, whilst fluency in AI tools may be a skill you assess for during the interview, when candidates use it to generate answers on their behalf without disclosure, a level of transparency is lost between the candidate and the interviewer.

1.3 Why detect AI-generated answers?

Identifying AI-generated responses is crucial because it ensures authenticity in evaluating a candidate’s genuine skills and competencies. Authentic answers reveal insights about how candidates think, solve problems, and fit within your company's culture.

2. Signs of AI-Generated Answers

2.1 Unnatural perfection in responses

You're likely to encounter AI-generated answers that are too perfect, almost robotic. These responses often lack natural language quirks such as pauses or verbal fillers, making them suspiciously flawless. While smooth responses might seem appealing, they're often a giveaway that you're not getting authentic answers.

Consider the context of your conversation. Does the answer seem too polished? It's crucial to look beyond the perfection. Human conversation typically includes minor hesitations or self-corrections, and a completely seamless answer might not be genuine.

2.2 Lack of personal experiences or anecdotes

One common feature of AI-generated responses is the absence of genuine personal stories. AI struggles to create authentic anecdotes, often resulting in generic comments instead.

If the response feels generic or lacks specificity, you might be dealing with an AI-generated answer. Human candidates will often share anecdotes drawn from their experiences, whereas AI might struggle to offer such specificity.

2.3 Too technical or generic responses

AI-generated answers might lean overly technical or remain annoyingly generic. This split isn't natural for human interaction, where one expects detailed, nuanced discussions on complex topics without being sterile or too brief.

In technical domains, humans often express personal insights or use analogies to explain their understanding. On the contrary, AI might simply regurgitate facts, showing no real connection or deeper understanding.

2.4 Inability to elaborate or go deeper when asked follow-up questions

AI-generated content tends to falter under scrutiny when follow-up questions are introduced. If a candidate can't expand on an initial answer or connect with the topic beyond surface level, this suggests reliance on AI.

Humans excel at breaking down thoughts and drawing from diverse experiences to enrich their answers. A lack of depth or an inability to build on an initial response typically points to AI-generated inputs.

2.5 A delay in answering whilst the AI writes responses

Lag in response can be a red flag. If you notice pauses where a candidate seems to be waiting suspiciously or typing only to produce a coherent paragraph after a conspicuous delay, this hints at AI usage.

Humans inadvertently glance or pause for brief moments as they gather thoughts, but consistent delays followed by fully constructed answers might mean an AI is in play.

3. How to combat AI generated answers

3.1 Ask questions without black and white answers

Questions without clear right or wrong answers encourage genuine candidate engagement. They prompt candidates to draw upon their experiences and personal judgement, making it more difficult to rely solely on AI for responses. These questions also promote a richer, more layered evaluation of the candidate.

Examples of effective questions include:

• What makes a team work well together, even under pressure?

• How do you determine if a risk is worth taking?

• What’s your approach when there’s no clear guidance or precedent?

3.2 Keep recordings and transcripts of interviews

Maintaining recordings and transcripts allows you to revisit interview content if there's a suspicion of AI-generated content. These records provide a reference for consistency and can highlight any incongruent answers upon review. Implement these practices with platforms that provide integrated recording and transcription services for review purposes, such as Evidenced.

3.3 Ask questions that involve a visual aid

Utilising questions requiring a visual aid, such as diagrams or flowcharts, challenges candidates to present and discuss their ideas spontaneously. It also means that speech-to-text software that may be listening to the interview cannot interpret the question, and will be unable to provide an AI generated answer. If you want to be even more comprehensive you can request that the candidate does not read the question out loud.

Some tasks you could provide could be to:

• Sketch an idea solution on a virtual whiteboard.

• Interpret data from a provided graph.

• Walk through a process using flow diagrams.

3.4 Run structured interviews

Structured interviewing creates a consistent framework where every candidate is asked the same questions, allowing you to identify AI generated patterns or discrepancies in responses more easily. This method can also help ensure that interviewers are asking the right questions and follow ups, making it harder for AI-generated content to slip through.

3.5 Design questions that require personal experience

Questions designed to elicit personal experiences deter candidates from using AI tools, as these demand authentic anecdotes and reflections. By asking more personal questions you gain insights into personal challenges and achievements, whilst making it hard for AI-generated content to provide real value.

Create questions like:

• Can you share a story where you demonstrated leadership?

• What personal project are you proud of?

• How have you adapted to a significant change?

Make sure to look for specificity in answers and gently probe candidates for underlying motivations and context.

3.6 Raise Your Concerns During the Interview

If you suspect a candidate is using AI-generated responses during an interview, it’s important to address it in the moment without sounding accusatory. Rather than confronting the candidate directly, try to steer the conversation in a way that encourages genuine, unscripted dialogue. For example: 'That’s an interesting answer - could you tell me about a time you actually applied that in your work?'

If concerns persist, you could gently introduce the topic by saying: 'We’ve been noticing that AI tools are increasingly being used in interviews - just to confirm, are you using any tools or assistance right now?'

Framing it as a broader industry observation rather than a direct accusation helps keep the tone neutral and gives the candidate a chance to respond transparently.

3.7 How Evidenced can help maintain a human touch

Maintaining a human touch at interview is increasingly crucial amidst advancements in AI. Even when interviews need to be run online rather than in person, Evidenced's platform offers the tools for interviewers to run efficient, effective interviews that still include human-to-human interaction.

Evidenced also help you protect your interview process from AI generated answers by allowing you to:

• Ensure compliance: provide one-click policy changes to your companies' interview process, including guidance on AI generated answers.

• Target interview training: provide real time guidance to new interviewers on what warning signs to look out for.

• Easily review interviews: get instant insight into every interview and every decision.

• Run structured interviews: run every interview to a template, ensuring the right types of questions are being asked.

Ultimately, consistency and authenticity should remain at the forefront of interview processes to ensure that your hiring decisions reflect true candidate potential. By leaning into practices that uncover genuine insights and utilising tools like Evidenced for structure and guidance, the risk of hiring based on inauthentic responses reduces significantly.