Artificial intelligence is already having an impact on how podcast creators produce their shows according to a survey of independent producers. The Podcast Host surveyed more than 600 creators and it says four in ten (42%) have started to use some form of AI to produce their show. For the majority (58%) who have yet to use AI tools, The Podcast Host says it received variety of reasons. But the one that stood out is the fact that many indie producers say they do not yet know enough about AI in order to bring it into their podcast workflow.
When it comes to those that have made the leap, the survey finds that topic idea generation is the most common use with nearly half (49%) saying they lean into AI to help generate show content. And 18% say AI helps them search out guests for their episodes.
Other indie creators are using AI tools like ChatGPT to help with script writing, which at 43% was the second-biggest use of the technology by indie producers. Another 30% said they use it to write their show notes.
The Podcast Host’s Katie Paterson says generative AI tools may not give producers the most original show ideas, but on the positive side they also never get creative block and they will work to get the human producer’s creative juices flowing. “Tools like this can help you explore angles you hadn’t thought of before,” she writes in a blog post.
Paterson points out there are some pitfalls to be careful of when using AI. First, it can be tough to sound natural when reading an AI-generated word-for-word script. She says a very thorough edit is needed on the copy before the episode is recorded to help smooth out rough edges or make the script sound more like the way a host speaks naturally.
There is also the question of the reliability of the research results, especially as reports have suggested that as more data gets plugged into AI tools, the less reliable their answers have become. Paterson says that means creators need to do their own due diligence on a topic before recording an episode, drawing on research from a combination of different sources. She also points out that ChatGPT can only provide answers from before its September 2021 update, meaning some of its answers may be outdated. Also, ChatGPT occasionally produces incorrect answers due to so-called hallucinations.
“There are also a whole load of AI-powered fact-checking tools that scan through written content to check the information is correct,” Paterson says. “While these wouldn’t be used for pre-recording research, you could run your transcript through one of these tools if you want that extra safety net before publishing.”