A guest told me my interview felt scripted, and she was right: how to interview podcast guests without sounding scripted

Sunny Ray · Jun 15, 2026 · 5 min read

Volume made me a worse interviewer, and a guest had to tell me.

To interview podcast guests without sounding scripted, stop reading your questions. Do your research with the guest, not just about them, so you know their story from the source instead of from a scraped summary. Pick one thread for the episode and go deep on it. Then listen, and let your next question come from their last answer. A list of questions is a safety net. Most hosts, me included, end up performing from it.

What she said, word for word

On June 15 I got on a debrief call with a founder I'd interviewed on the show. I asked her for honest feedback, and she gave it. Her first point:

I think the questions and the intent could have been better organized. We were definitely reading, which was very clear.

She didn't stop there. The flow jumped back and forth between her personal story and her business without settling into either one. She could tell when a question didn't follow from what she'd just said, which meant I was going to the next item on the list instead of listening. And the pre-show research had mistakes in it, the kind you get when a model scrapes the web and fills gaps with confident guesses.

It stung a bit. It was also completely fair.

Reading questions kills the conversation

A prepared question list is useful. Reading from it is the problem. When a host reads, three things happen. The guest hears the shift in tone and knows they're being processed. The host stops listening during the answer, because part of their attention is on the next question. And the best moment in any interview, the follow-up to something unexpected, never happens, because the list doesn't have room for it.

She noticed exactly that. A question would land that had nothing to do with what she'd just said. To a guest, that reads as: you weren't listening. To an audience, it reads as stiff.

The fix she suggested was simple. Read less, converse more. Follow the guest's answers in real time. Keep the list as a backstop for when the conversation stalls, not as the script.

AI research gets guests wrong, so ask them

The research point is the one I'd underline for anyone running a show with AI prep. The pre-show research on her had inaccuracies, and her read was that they came from LLM-scraped data. That's believable. A model pulls from old profiles, press releases and other people's summaries, then writes it all up in the same confident voice whether it's right or not.

Getting a guest's own story wrong on air is worse than not knowing it. It tells them you didn't care enough to check, in front of an audience.

Her recommendation: do more preliminary research directly with the guest. That's what a pre-interview is for. On a short call before the recording, ask them for their story in their own words, what they want to talk about, what they're tired of being asked. You get accurate facts, the guest gets comfortable, and you find the thread worth pulling before the cameras are on. Use AI to prepare for that call, never to replace it.

One thread, deep, beats two threads, shallow

Her structural note was about time. We didn't have enough of it to cover both her personal story and her business properly. So we did both badly instead of one well.

Every guest has more than one good episode in them. The mistake is trying to record all of them in one session. Pick the thread: the founder's journey, or the business problem, or one hard decision. Tell the guest which one before you start. Then give it the whole episode. If the other thread is good, it's a reason to have them back. She suggested a part two in about six months, once her story has moved on, which is a better outcome than a rushed part one.

The volume trade-off when the show is your pipeline

Her last suggestion was the hardest one to hear: do fewer shows.

At the time I was running about 15 to 20 meetings a day across five live streams. That's the engine. Every guest is a relationship, many are buyers or investors, and the show is how I meet them. Volume is the point.

But volume has a cost, and she named it. At that pace, prep gets compressed, research gets outsourced to a model, and the host reads because there wasn't time to internalize anything. The guest feels it. If enough guests feel it, the show stops being a gift and starts being a process they went through.

The answer isn't obvious when volume is the point. What I can do is protect quality inside the volume. The changes I took from that call:

• Less scripting. The question list becomes a short set of themes, not a script.
• More real-time conversation. The next question comes from the last answer.
• Better guest research, done with the guest in the pre-interview, not just by a model before it.

She also mentioned audience feedback during live streams as something that would help, though live viewership is hard to read in the moment. That one is still open.

Ask your guests

The real lesson isn't any single technique. It's that I asked. Most hosts never debrief their guests, so they never hear that the interview felt scripted. They just notice that fewer people say yes over time.

After your next recording, get on a short call with the guest and ask what you could have done better. Then fix it.

If you want a show where the research, prep and pre-interviews are handled so you can just listen, see how hosts get prepared for each taping.

Frequently asked questions