Overview of The Impact of AI on Podcasting (Harvard Data Science Review Cross-Post)
This episode is a cross-post from the Harvard Data Science Review podcast featuring a conversation with Katie Malone (Linear Digressions) and John Krohn (Super Data Science), moderated by HDSR’s Shali Meng. The discussion explores how AI is reshaping podcasting behind the scenes, influencing research, editing, promotion, and even audience expectations—while raising bigger questions about authenticity, transparency, and the future of human-made media.
Key Themes
AI as a production sidekick
Both hosts described using AI heavily for the “unsexy” parts of podcasting:
- drafting show notes and episode summaries
- speeding up research and background prep
- generating newsletter content and transcripts
- helping with promotion and content repurposing
- creating animated video shorts from long interviews
John emphasized that AI is being used broadly, but with a clear goal: augment the human host, not replace them. Katie echoed this, saying AI helps make solo podcast production sustainable.
Human voice still matters most
A central point throughout the conversation was that podcasting depends on personality, judgment, and editorial taste.
- Hosts want their own voice to remain front and center.
- AI is useful for support tasks, but not for replacing the host’s unique perspective.
- Both speakers were cautious about letting AI shape editorial direction or script content too much.
Katie in particular said AI becomes “jagged” when it moves beyond research support into scripting or suggesting lines of discussion.
How AI Is Changing Podcasting Workflows
Where AI is already useful
The guests identified several high-value uses of AI in podcast workflows:
- Research acceleration: finding background material faster and avoiding dead ends
- Summarization: producing digestible notes from transcripts or interviews
- Content repurposing: turning long-form conversations into clips, shorts, and newsletters
- Operations automation: reducing repetitive admin work
John described AI workflows as deeply integrated into his team’s production process, while Katie said AI helps with the parts she used to dread, especially writing descriptions and doing prep.
What still works best with humans
Despite the efficiency gains, both argued that humans are still essential for:
- setting the editorial vision
- deciding what is worth covering
- asking good follow-up questions
- bringing nuance and authenticity to conversations
Katie said the best use of AI is as a research partner and scout, not as a full content generator.
Transparency, Trust, and Authorship
The disclosure problem
The conversation spent significant time on how to think about transparency when AI is involved.
Shali Meng raised an important question: if AI helps with research, writing, or editing, how much of that should be disclosed?
Katie noted that current AI-detection labels can be misleading. For example, a newsletter might be labeled “90% AI-generated” even though the human did the high-level thinking, curation, and editing.
A growing gray area
Both guests agreed that current tools don’t fully capture the nuance of authorship:
- AI may draft the prose, but the human may still do most of the intellectual work.
- Watermarking and detection tools can oversimplify complex creative workflows.
- Attribution and liability become tricky when AI is involved in the chain of creation.
John and Katie both suggested that the industry is still figuring out where the line should be.
AI, Education, and the Future of Podcasting
Podcasting as a learning medium
The guests see a major opportunity for AI in education and learning through podcasts.
Katie noted that podcasting already works well for teaching because it can explain concepts at varying levels of depth. AI can make that format even more flexible by:
- personalizing explanations
- filling in gaps for learners
- adjusting technical depth on demand
NotebookLM as a standout example
Both guests called NotebookLM one of the most impressive AI products they’ve seen for knowledge work and educational audio.
They highlighted its ability to generate realistic-sounding conversations from source material, opening up possibilities for:
- article-specific podcast summaries
- personalized educational audio
- company-wide internal briefings
- AI-assisted explainer content
John mentioned that some organizations are already using NotebookLM-generated podcasts to explain their work internally.
Impacts on Organizations and Jobs
Better output, not necessarily fewer people
John said AI has mostly improved quality and efficiency rather than reduced headcount on his team.
AI allows people to:
- spend less time on rote work
- focus more on strategy and quality
- produce better research and better episodes
Katie similarly said she hasn’t seen large-scale job loss in her own work, but she does worry about the broader social effects as AI becomes more embedded in workflows.
Mixed signals on labor impact
Both speakers acknowledged that the labor market effects of AI are still unclear. They noted that the real-world impact often lags behind the technology’s raw capabilities.
Notable Insights
- “Use AI as much as possible without losing authenticity.” — John’s guiding principle for production
- AI is strongest as a research assistant, weakest as a replacement for editorial judgment.
- Detection tools may misrepresent human contribution by focusing only on generated text rather than the full creative process.
- NotebookLM may represent a major shift in how people consume educational content.
- AI can improve podcast quality today, but both hosts worry about a future where AI-generated shows could crowd out human-made ones.
Takeaways for Podcasters and Creators
Practical recommendations
- Use AI for repetitive, time-consuming production tasks.
- Keep human control over editorial direction and voice.
- Treat AI as a research aid, not a substitute for critical thinking.
- Be thoughtful about disclosure, but don’t assume current labels capture the full picture.
- Experiment with AI-generated derivatives like newsletters, clips, and summaries.
- Watch tools like NotebookLM closely for educational and enterprise podcasting use cases.
Closing Thought
The episode’s core message is that AI is already transforming podcasting—but mostly by making human creators more efficient, not obsolete. The real challenge ahead is preserving authenticity, trust, and creative judgment as AI becomes more capable and more common.