Guide

How to transcribe a podcast episode (and why you should publish it)

A transcript is the only version of your episode that search engines can read. It is also the cheapest accessibility work available to a podcast.

30 August 2026

Podcast audio is invisible to search. Google cannot hear your episode, and neither can anyone searching for the exact topic you spent forty minutes on. A published transcript is the version of the episode that is readable, searchable, quotable and accessible — and producing one now takes minutes rather than an afternoon.

Why publish it, not just make it #

Search. Every episode becomes a page of real text about a specific subject. A show with a hundred published transcripts is a hundred indexable pages; a show with a hundred MP3s is a directory listing. This is the single highest-return SEO action available to most podcasts, and almost nobody does it.

Accessibility. Deaf and hard-of-hearing listeners cannot use your show at all without one. So do people in a loud environment, people whose second language is yours, and people who simply prefer to read.

Reuse. A transcript is the raw material for show notes, pull quotes, social clips and newsletter copy. Finding a good thirty-second clip in a text file takes seconds; finding it by scrubbing a waveform takes an evening.

Your own archive. "Which episode did we talk about that in?" is answerable by searching a folder of transcripts, and unanswerable otherwise.

Producing the transcript #

The workflow is short:

1. Export the finished episode — the same MP3 or WAV you publish. Use the mastered version, not the raw multitrack: it is the cleanest audio you have.
2. Drop it into Offline Transcription and transcribe. A forty-minute episode takes a couple of minutes on Apple Silicon, and there is no per-minute charge, which matters when you are doing this every week for years.
3. Export in two formats: HTML for the version you publish, and CSV or SRT for the timestamped version you will actually work from.

If your show has a back catalogue, transcribe the whole folder in one go. The economics of local transcription make retroactively transcribing two hundred old episodes a free afternoon rather than a budget line.

The editing pass that makes it publishable #

An unedited automatic transcript reads badly, and publishing one unedited is worse than not publishing at all — it makes your show look sloppy to the first-time reader who arrived from search.

Fix these four things:

Names and jargon. Your guest's surname, their company, the technical terms specific to your subject. Automatic transcription renders unfamiliar proper nouns as common words that sound similar, confidently. Build a list of the terms your show uses often; you will fix the same ones every week until you learn to check them first.

Filler. Every "um", "you know" and false start is faithfully transcribed. Cutting them is not dishonest — it is the same convention every published interview uses. Do not change what people said; remove the noise around it.

Paragraphs. Whisper produces segments, not paragraphs. Group them by topic so the page is readable.

Speaker labels. Add them by hand. No automatic tool does this reliably from a mixed file, and for a two-person show you know exactly who is speaking.

Budget ten to fifteen minutes an episode. That is the real cost of publishing transcripts, and it is much lower than most shows assume.

Putting it on the page #

• Publish it on your own site, on the episode page, as real HTML text — not as a PDF download, and not inside an image. It must be text a crawler can read.
• Put a short summary at the top, then the transcript below. The first hundred words are what appears in search results.
• Timestamps as headings for long episodes give readers a way to navigate and give you anchor links to share.
• Do not hide it behind a click if you can avoid it. Content loaded only after interaction is less reliably indexed.

Using it for clips and notes #

The timestamped export is the tool for this, and it is underused.

Open the CSV in a spreadsheet and you can sort segments, filter for a keyword across the whole episode, and read a candidate quote with its exact start and end time. Pull the ten best lines, take the timecodes to your editor, and you have your social clips without listening to the episode again.

For show notes, the transcript is your outline: skim the headings you made, write four sentences per section, done.

Chapters and captions #

If you publish episodes to YouTube, upload an SRT rather than relying on YouTube's automatic captions. Auto-captions have no punctuation, mangle names, and are treated differently from a real subtitle track. Making an SRT covers the export and upload.

The short version #

Transcribe the mastered episode, spend ten minutes fixing names, filler and paragraphs, publish the result as HTML text on the episode page, and keep a timestamped export for pulling clips. Local transcription makes doing this every week — and retroactively for your whole archive — cost nothing but the editing time.