How to check what someone actually said in a podcast or interview.

A quote travels faster than its source. By the time a claim from a three hour podcast reaches a group chat, it has been shortened, sharpened and sometimes reversed. If your work depends on what was actually said, you need the exact words, who said them, and when.

Summaries are not enough

AI summaries are useful for deciding whether to listen. They are not evidence. A summary blends speakers, drops conditions like if and unless, and turns a guess into a statement. For anything you will repeat, cite or act on, go back to the words.

A simple method

• Get a transcript with timecodes. Most video platforms provide one, and transcription tools fill the gaps.
• Split it into claims. A claim is one statement that could be true or false, such as a number, a prediction or a statement about a person.
• Attach the speaker and the timecode to every claim, so anyone can jump straight to the moment.
• Keep the exact words next to your paraphrase. If they drift apart, the exact words win.
• Note what kind of claim it is: a fact, a forecast, an opinion or a promise. They age differently.

Why receipts matter

A receipt is the shortest proof that a claim was made: the speaker, the source, the timecode and the sentence. It lets a reader check you in ten seconds instead of ten minutes. Over months, receipts also show when someone changes their position, which is often the most valuable signal of all.

Why we are building Epistell

Epistell does this at scale. Paste a link or drop a PDF, and it returns every claim tied to its speaker, its timecode and a receipt. Its report library covers ten formats, from a daily brief to a committee memo. The brand and design are complete, and the full stack build is next.