🎙️ ChatGPT Can Now Take Your Interview Audio. Don’t Let It Become the Source.

Here’s a genuinely useful new feature for reporters: you can now upload an interview recording directly to ChatGPT and work with the audio itself.

That means an MP3, M4A, WAV or several other common audio formats can go straight into a chat. ChatGPT can transcribe it, summarise it, find particular comments and answer questions about what was said. Audio files can be as large as 512 MB, although very long recordings may time out.

For journalists, the useful bit isn’t simply “AI can transcribe my interview.”

It’s what you can do after transcription.

Turn an interview into a reporting pack

Instead of asking for a polished article, try this:

You are helping me analyse a recorded journalistic interview.

First produce a faithful transcript. Do not deliberately clean up, improve or paraphrase the speaker’s wording.

Then identify:

1. The five most newsworthy statements.
2. Potentially strong direct quotes, with enough surrounding context to understand what the speaker meant.
3. Every factual claim that I should independently verify.
4. Any apparent contradiction, ambiguity or change of position within the interview.
5. Important follow-up questions I failed to ask.

Do not invent information that is not present in the recording. Clearly distinguish what the speaker said from your own interpretation.

That changes transcription from an administrative chore into the beginning of the reporting process.

You could also upload several interviews separately and compare what different sources said about the same issue.

But check the tape

This is the important part.

OpenAI warns that transcripts can contain errors and that speaker identification may be unreliable. So never publish a quote merely because ChatGPT says that is what somebody said. Go back to the original recording and listen to every important quote, name, number and contested claim before publication.

There’s also a source-protection issue. An interview may contain confidential identities, unpublished allegations or sensitive personal information. Before uploading that material, check your newsroom policy and the service’s current privacy and retention settings. Personal ChatGPT users can opt out of having new conversations used for model improvement; OpenAI says content from Business and Enterprise offerings is not used for model training by default.

The useful principle is simple:

Let AI interrogate the recording. But the recording remains the source.

Further reading: OpenAI — Uploading files and audio to ChatGPT

OpenAI — Data Controls FAQ

🌐 Social Weaver offers practical AI training for journalists and newsrooms, including interview, research, verification and AI-assisted reporting workflows: social-weaver.co.za/services