To find everything a person has said across podcast interviews, you need an index where their voice is attributed to their name in every episode, not just the ones where they were the headline guest. On Audioscrape, search the person’s name, open their page, and you get every appearance the classifier has attributed to them across the public corpus; add speaker: "Their Name" plus a topic to the search bar and you get their exact words on that topic, each with a timestamp and a link to the audio. AI assistants can do the same over MCP with search_speakers, get_speaker and search_audio.
Who needs this
Investors checking what a founder said about churn a year ago versus last month. Journalists tracing when an executive first denied a rumor. Recruiters and researchers profiling an expert. Comms teams tracking their own executives. In all of these the question is the same: not “which episodes featured this person” but “what did they actually say, and when”.
How to do it with Audioscrape
1. Start from the person page. Type the name in search. If the person is in the index, the suggestions include a person entry that opens /person/{slug}. The page lists their attributed appearances with the show, the episode and the date. This is the inventory.
2. Search their words. In the search bar, combine the speaker filter with your topic:
speaker: "Jane Founder" churn
speaker: "Jane Founder" "product-market fit"
speaker: "Jane Founder" from: 2025 hiring OR headcount
Quoted phrases match exactly. Add podcast: "Show Name" to restrict to one show. Switch to semantic search if you want the idea rather than the wording, for example “how she thinks about pricing”; semantic search is on Pro and Enterprise.
3. Sort by date. Results can be ordered by relevance or by date. Date order lets you read a person’s position on a topic as it changed, which is often the whole point.
4. Cite. Each result has a timestamp that opens the episode at that second. Click the share button next to a segment for a link or an embeddable clip, so a reader can hear the quote in context.
With an AI assistant. Connect Claude, ChatGPT or Cursor to https://mcp.audioscrape.com and ask “What has Jane Founder said about churn on podcasts, ordered by date, with quotes and timestamps?” The assistant calls search_speakers to resolve the name to a person_slug, get_speaker for the appearance list, then search_audio with sort_by: "date" for the quotes. For one appearance in depth it calls get_episode_overview and get_transcript.
From code. POST /api/search with "filters": {"speakers": ["Jane Founder"]} and "sort": {"field": "date", "order": "asc"}. Paginate with limit (max 100) and offset. Reference: Search API.
What you get
- A list of attributed appearances across shows, not just the ones with the person’s name in the title.
- Speaker-labeled segments with start and end timestamps and the surrounding context on request (
include_context). - Entities mentioned in each segment, so you can see which companies or people came up alongside the quote.
- Shareable links and embeds for any moment.
- You can also set an alert on the person’s page so you get a digest when they appear in newly transcribed audio.
Limits and honest caveats
- Attribution is automatic. The classifier only names a voice when it has evidence, so some appearances will be missing (the person is present but shown under a generic label) and occasionally one will be wrong. Treat the person page as a strong starting inventory, not a legally complete one. Report mis-attributions from the person page; corrections are reviewed.
- People with common names, or who share a name with someone more famous, can be harder to pin down. Add a show or topic filter.
- The index covers podcasts and public meetings that Audioscrape has ingested. Shows that are missing can be submitted at /submit. Private recordings you hold can be uploaded to your workspace and searched the same way.
- Free accounts get 100 searches per month; paid plans raise the limit.
Doing it without Audioscrape
Search podcast directories for episodes that list the person as a guest, then read or transcribe each one. This finds headline appearances and misses the panel where they spoke for six minutes. It also gives you no way to search the words themselves without transcribing every episode yourself.
Plans and limits: /pricing. Facts for AI agents: /docs/for-agents.