How to Research a Topic Across 20 Podcasts (Without Losing the Sources)
Podcasts are where practitioners say the quiet part out loud. A five-step method for turning scattered interviews into a memo with citations you can check.
Guide5 min read
Articles are written after the fact. Podcasts are recorded while the thing is still messy — which is why a founder will say "we tried that and it failed because of X" on a show and never put it in a blog post.
That makes them useful sources and terrible archives. The evidence is scattered across 90-minute conversations, undated in your memory, and impossible to search by hand. The method below is about keeping the evidence attached to the claim.
Start with a question that can fail
"The future of AI" is not a research question. "Are companies actually replacing junior developers with agents, or just talking about it?" is. It has a yes/no shape, it implies evidence, and it can be contradicted.
Write the question on top of the note before you collect anything. Every episode either contributes to it or is a distraction.
Collect wider than you think
Search the topic, not the answer. If you search "junior developers replaced", you will find the episodes that argue yes. You want the ones that disagree.
A workable collection for one question:
- Five episodes from shows you already trust. Baseline.
- Five from adjacent fields. If you are researching hiring, product and economics shows will say things HR shows will not.
- Five from people with a stake. Founders, investors, tool vendors — they are biased, and the bias is data.
- A few older episodes. The 2023 take is often more revealing than the 2026 consensus.
Twenty is enough to see the shape of the conversation. At forty you are procrastinating.
Extract the claims, not the summaries
This is where most research dies: people save 20 summaries and learn nothing. Summaries describe episodes. You need claims across episodes.
For each episode, write three lines:
Claim: Agents replaced the bottom rung of the ladder, not the middle.
Evidence: Two teams, both said onboarding juniors got cheaper instead.
Doubt: Both were vendors selling agent tooling.The Doubt line is the one that makes the memo worth reading.
When you are short on time, ask across the corpus instead of reading episode by episode. Ask AI in Podwise answers from transcripts with citations:
What do podcasts say about agents replacing junior developer work? Who argues each side, and which episodes are they from?
You get positions with sources; open the two that surprise you. For a topic where the disagreement is the interesting part, that is faster than reading summaries one by one.
Verify like an editor
Three checks catch most bad synthesis:
- Open the timestamp. If a quote is not where the tool says it is, distrust the rest of that episode's notes.
- Check the incentive. A guest from an agent vendor has a position. That does not make them wrong; it makes their claim a data point, not a fact.
- Look for the missing side. If every source agrees, you have sampled one community, not the industry. Go find the argument against.
Podcast claims age badly too. A 2024 episode about hiring is a historical document, not a current one. Date every claim in your notes.
Write the memo with the disagreement intact
The output is not "here is what podcasts say." It is a short memo with three sections:
- What most sources agree on. Two or three sentences.
- Where they split. The strongest version of each side, with names and episodes.
- What I could not find out. The gap is often the most useful part, and it is honest.
Put the citations inline, not at the bottom. A reader who wants to check one claim should not have to hunt.
Most guests assume agents will absorb entry-level tasks first. The split is
about what happens next: [Guest A] argues the middle goes too, because
review is the bottleneck; [Guest B] says review is exactly why it stalls
(Show, 00:22:10).
Not found: any team that measured the effect on hiring over more than a year.When not to use podcasts
If you need current numbers, podcasts are the wrong source. If you need the official policy, read the policy. If you need a definition, use a reference.
Use podcasts for the parts that do not make it into documents: what people tried, what broke, what they would do differently, and who disagrees with whom. That is the material you cannot get from a search results page, and it is worth the method.
