AI-Powered Podcast Guest Research That Creates Better Questions

Most podcast hosts spend thirty minutes researching a guest and still ask the same questions everyone else asks.

I know because I used to be one of them. I’d scroll LinkedIn, skim a couple of recent blog posts, maybe check the company About page. Then I’d walk into the interview with generic questions about their background, their company’s origin story, and their thoughts on industry trends.

The conversations were fine. Polite. Forgettable.

The breakthrough came when I stopped treating research as collecting biographical data and started treating it as hunting for the one angle nobody else has explored.

Why Most Podcast Guest Research Misses the Mark

Standard guest research focuses on credentials and recent wins. Where did they work? What did they build? What’s their latest launch? That produces interviews that sound like a LinkedIn profile read aloud.

Real research uncovers contrarian takes, current frustrations, and the frameworks behind the decisions. It finds the story behind the success story: the failed experiment that led to the breakthrough, the unconventional belief that drives how someone actually thinks.

Most hosts collect facts. The goal is to collect insights.

The Difference Between Facts and Insights

The most memorable B2B podcast moments tend to come from guests explaining a decision-making framework or pushing back on a popular industry belief. Yet most interview questions still aim at biographical information and company milestones.

That’s the gap. The guest isn’t the problem. The preparation is.

Last month I interviewed a VP of Marketing at a Series B SaaS company. Surface research showed typical growth-trajectory material. But deeper digging turned up something better: a buried blog post from eighteen months earlier where he’d written about deliberately killing his highest-traffic content to focus on pipeline quality over vanity metrics.

That became the entire interview. We spent forty minutes dissecting why he nuked 200k monthly visits, how he sold leadership on it, and what happened to revenue afterward. It’s the most-shared episode from the past six months.

None of that surfaces from skimming an About page.

The 10-Minute AI Research System

I built a workflow that turns scattered information into interview gold in about ten minutes. Two phases, working together to surface the insights that create quotable moments.

Step 1: Data Collection (5 minutes)

Gather raw inputs from five sources. Don’t summarize. Just copy relevant text into one document.

• LinkedIn activity. Recent posts, comments, and articles. Look for opinions, debates they’ve engaged in, and topics they return to.
• Company blog. Their bylined content, plus announcements they likely influenced. What initiatives are they driving?
• Previous podcast appearances. Search “guest name” + podcast on YouTube. Where did they seem most passionate or most frustrated?
• Twitter/X. More unfiltered than LinkedIn. What do they complain about? What takes do they push back on?
• Industry publications. Trade quotes, conference talks, panels. What are they known for?

Raw input only. The analysis comes next.

Step 2: AI Analysis and Question Generation (5 minutes)

Feed the collected data through three structured prompts.

Prompt 1 — Pattern Analysis: “Analyze this research on [guest name]. Identify: 1) topics they return to repeatedly, 2) contrarian or unconventional views they hold, 3) challenges or frustrations they mention, 4) frameworks or mental models they reference. Output in bullets.”

Prompt 2 — Gap Analysis: “Based on this research, what angles haven’t been explored in their previous interviews? What questions would surprise them? What topics do they care about but rarely discuss publicly?”

Prompt 3 — Question Generation: “Generate 15 interview questions based on this research. Focus on: 1) contrarian takes, 2) decision-making frameworks, 3) failures or pivots, 4) future predictions. Avoid generic background questions.”

When I ran this on a fintech founder, the analysis showed he’d referenced “revenue-based financing” in three different contexts but had never been asked to explain his framework for evaluating it. That became a fifteen-minute segment and his most-quoted clip.

The Four Types of AI-Generated Questions That Actually Work

The best AI-generated questions fall into four categories. Each does a different job.

The Contrarian Question

These challenge conventional wisdom in the guest’s industry. The AI finds where they’ve disagreed with popular opinion, then builds tension around it.

“You’ve written that customer success teams are becoming bloated. Most SaaS leaders are doubling CS headcount. What are they getting wrong?”

Instead of asking about best practices, you’re exploring why best practices might be wrong.

The Story Behind the Story

These dig into failures, pivots, and decisions that didn’t work. AI often spots phrases like “didn’t go as planned” buried in old posts.

“Your LinkedIn shows you rebuilt your entire onboarding flow last year. What broke that forced your hand?”

Guests light up when asked about problems they solved, not successes they achieved.

The Framework Question

These get guests to articulate their mental models. AI catches when someone references a decision process repeatedly without ever fully explaining it.

“You mention ‘signal versus noise’ in customer feedback three times across different posts. Walk me through how you actually separate them.”

Framework questions force structured thinking in real time. That’s where quotable moments come from.

The Future State Question

These explore where the guest thinks the industry is heading, based on clues in their content or recent moves.

“Your company just hired three AI engineers. Most agencies use AI tools but don’t build AI teams. What do you see that they don’t?”

These position the guest as a strategic thinker, which is exactly what they want to be seen as.

Building the Workflow: Tools and Prompts

I’ve tested this with Claude, ChatGPT, and Perplexity. Each is better at a different job.

• Claude handles long-form pattern recognition best. It holds context across 3,000+ words of pasted research and finds subtle patterns.
• ChatGPT generates the strongest questions. It understands what makes a question provocative versus predictable.
• Perplexity is fastest for gathering recent interviews and articles without manual searching.

The Exact Question-Generation Prompt

Role: You're preparing interview questions for a B2B podcast host.
Context: [Paste all research here]
Task: Generate 15 interview questions that:
1. Avoid biographical basics (no "tell us about your background")
2. Focus on contrarian views, decision frameworks, failures, or future predictions
3. Reference specific things mentioned in the research
4. Would surprise the guest (different from questions they usually get)
Format: Number each question and include a 1-sentence explanation of why it's worth asking.

The “why it’s worth asking” line is the important part. It forces the AI to justify each question against the actual research instead of defaulting to generic interview filler.

I started with basic prompts like “write interview questions about this person.” The results were terrible. Adding structure, constraints, and reasoning requirements transformed the output. The current version produces usable questions about 80% of the time, versus maybe 20% with my original approach. That took six weeks of iteration: run the questions in real interviews, note which ones created the best moments, feed that back into the prompts.

From Research to Interview Assets

Here’s where this stops being a hack and starts being a system. The research output feeds three more assets that compound the value beyond the conversation itself.

• Pre-interview brief. A one-pager sent to the guest 24 hours out, with the topics and 3-4 sample questions. They prepare thoughtful answers instead of thinking on the spot.
• Show notes template. The research insights become the foundation for the description, key takeaways, and social clips. You highlight the specific contrarian takes and frameworks you discussed, not generic summaries.
• Content repurposing queue. The best quotes and frameworks become seed material for LinkedIn posts, newsletter content, and articles. One good interview can produce weeks of content when you’ve already identified the quotable moments.

This is the same principle behind everything I build: a single input producing outputs across the full funnel. One conversation becomes ten things instead of one. If you want the broader version of this thinking, that’s what Systems-Led Growth is about.

The research workflow saves time on the front end and multiplies output on the back end. Instead of thirty minutes gathering basic information, I invest ten minutes building a foundation for better questions, a better conversation, and better assets.

Most hosts treat guest research as a necessary evil before the “real work” of interviewing. But research is where great interviews actually begin. The conversation is just the execution.

Want to build systems like this across your whole go-to-market motion? Book a call.

Related reading: The Content Marketing Workflow That Lets One Person Do the Work of Five · score yourself with the matching audit · start with an audit · read the manifesto · The Content Creation Workflow That Produces Five Posts a Day (As One Person)