Editorial policy
Every episode on PodThis is made with AI. This page says exactly what that means, who is responsible for what, and how to get something corrected. Last updated 2026-10-10.
How an episode is made
A topic goes through a seven-stage pipeline: research, script, voices, a quality check, stitching and mastering, cover art, and metadata. Google Gemini does the research and writes the script, summary and chapters. Gemini's image model draws the cover art. Voices come from Gemini text-to-speech or, for some styles and languages, Cartesia. Before any line is voiced, an automated check finds stock phrasing in the script and has it rewritten. No human voice is cloned or impersonated. Sources the model consulted are listed as "Sources & further reading" on the episode page when the pipeline could name them.
How AI audio is marked
Episodes are labelled as AI-made where people see them, and marked for software too. The MP3 file has a tag in its metadata that says it is AI-generated, which voice provider and model made it, when, and the episode link. Files made from October 2026 on have it; older files are being tagged. In our RSS feeds, every episode description ends with "AI-generated with PodThis". Google puts its SynthID watermark into the sound of Gemini voices, and into the cover art from Gemini's image model. Cartesia does not document a watermark, so for its voices the tag and the labels are the marking for now.
PodThis Featured
Featured series (the shows under Series) are produced by PodThis. Before release, every featured episode is listened to, edited and fact-checked by Martin Andersen, founder and editor of PodThis. Episodes that fail that review are regenerated or dropped; the "How this episode was made" note on each page names the editor.
Creator episodes
Episodes under a creator's /@handle are made by that creator from their own topic or
text. PodThis runs automated moderation before synthesis but does not editorially review creator
episodes. They are labelled "AI generated" and, on the episode page, say who created them. Creators
choose whether an episode is public, unlisted or private.
Accuracy and corrections
Generated audio can contain mistakes. If you find an error in a featured episode, e-mail support@podthis.co with the episode link and timestamp; we correct the transcript and, where it matters, re-record the segment. Errors in creator episodes are the creator's to fix; rights or safety concerns go through our DMCA and reporting page.
Pages and blog posts
The text on podthis.co outside the episodes, including this page and the blog, is written by Martin Andersen. He reads every sentence and checks every number against the product before it goes live.
What we do not do
- No pages or episodes generated to target search keywords.
- No training of models on user-generated content.
- No voice cloning or impersonation of real people.
- No undisclosed sponsorship inside generated scripts.