---
type: "post"
id: "introducing-exa-podcasts-search-3-2m-episodes-with-1e4847"
slug: "introducing-exa-podcasts-search-3-2m-episodes-with-1e4847"
title: "Introducing Exa Podcasts: Search 3.2M Episodes with Full Transcripts"
description: "Exa has added podcast search to its existing Search API: 3.2 million episodes from 119,000 shows, indexed with full transcripts."
retrieval_nugget: "Exa has added podcast search to its existing Search API: 3.2 million episodes from 119,000 shows, indexed with full transcripts. The capability is available across Exa endpoints at no additional charge, giving builders a directly queryable source of long-form spoken material for research, briefing, monitoring, and agent workflows."
published_at: "2026-10-30"
updated_at: "2026-10-30"
record_date: "2026-10-30"
date_kind: "published_at"
topics: ["agent-workflows","mcp"]
entities: ["Exa","Exa Podcasts","Exa Search API","Particle"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "Exa has added podcast search, indexing 3.2 million episodes with full transcripts; the archive’s related Exa Places release does not cover this new capability."
prior_refs: ["memory:7bef9c60-53bf-8d8a-a995-e61c021a7a3e"]
source_url: "https://exa.ai/blog/podcast-search"
source_urls: ["https://exa.ai/blog/podcast-search"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:introducing-exa-podcasts-search-3-2m-episodes-with-1e4847"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/introducing-exa-podcasts-search-3-2m-episodes-with-1e4847/","markdown":"https://newruntime.com/posts/introducing-exa-podcasts-search-3-2m-episodes-with-1e4847.md","json":"https://newruntime.com/posts/introducing-exa-podcasts-search-3-2m-episodes-with-1e4847.json"}
---

# Introducing Exa Podcasts: Search 3.2M Episodes with Full Transcripts

## Retrieval answer

Exa has added podcast search to its existing Search API: 3.2 million episodes from 119,000 shows, indexed with full transcripts. The capability is available across Exa endpoints at no additional charge, giving builders a directly queryable source of long-form spoken material for research, briefing, monitoring, and agent workflows.

## Summary
Exa has added podcast search to its existing Search API: 3.2 million episodes from 119,000 shows, indexed with full transcripts. The capability is available across Exa endpoints at no additional charge, giving builders a directly queryable source of long-form spoken material for research, briefing, monitoring, and agent workflows.

## What happened
On October 29, 2026, Exa announced Exa Podcasts, a podcast corpus added to the company’s search index. The company says the index contains 3.2 million episodes from 119,000 shows. It describes the collection as roughly 23.7 billion words, or about 2.2 million hours of audio.

The important implementation detail is that the material is not limited to podcast titles, descriptions, or episode metadata. Exa says each episode is transcribed, allowing searches over what speakers actually said. Search results can include the show and episode title, publisher and description, hosts and guests, publication date, duration, and a full transcript with speaker labels. The corpus spans technology, business, finance, news, science, culture, sports, and other categories.

Exa positions the release as a data source for AI agents. Its examples include finding the newest episode of a named show, locating recent comments by Jensen Huang about inference demand, and finding episodes in which founders discuss pricing mistakes. In practical terms, an application can use the search layer to identify relevant appearances, then use transcript content to summarize a discussion or retrieve a statement in context. Exa says the index is updated continuously and that new episodes appear soon after they air.

The company says Exa Podcasts is live in the Search API for all users and is available across all endpoints at no additional charge. Particle contributes data to the index; Exa points users needing deeper podcast analytics to that partner. No separate podcast price, usage limit, retention policy, geographic coverage policy, or transcript-correction process is specified in the announcement.

Exa also reports results from its own comparison across seven APIs. Using each service’s best tested mode, it checked whether an API returned the newest episode from each of 250 active shows in the top five results. Exa reports a 98.4% result. It also says it asked 667 questions about specific episodes and measured whether result sets led to the correct answer, reporting 85.0%. These are vendor-reported benchmark results rather than an independently supplied evaluation, but they identify freshness and answerability as the properties Exa is optimizing for.

## What the archive adds
Earlier Telegram coverage tracked Exa’s improved search quality, its Company Search Benchmark, and its packaging of web search for MCP and Vercel AI Gateway. This is a new product event on that broader Exa retrieval story: instead of another web-search interface or evaluation, it adds a distinct, transcript-level source corpus to the existing API surface.

## Trend signal
No supplied New Runtime pattern is a sufficiently direct match. The announcement is primarily an expansion of a search corpus and retrieval interface, rather than evidence about a broader agent architecture, protocol, security, or organizational shift.

## Why now
The timing is explained only operationally. Exa says podcasts contain extended comments from founders, executives, investors, and experts that may appear before written material, while conventional search has difficulty reading audio files. Continuous updates matter because the value of that material falls if an agent retrieves an older episode when a newer one exists.

## Connections to other events
This release follows Exa’s earlier work around web retrieval for agent workflows, including an MCP and Vercel AI Gateway packaging mentioned in the archive. The connection is functional: those surfaces make search available as an agent action, while Exa Podcasts enlarges the material that action can retrieve. The supplied source does not establish a technical dependency between those releases.

## Why it matters
Teams often treat podcast research as a manual step: find appearances, listen or skim transcripts, extract claims, and prepare a briefing. A searchable transcript corpus can turn much of that sequence into retrieval work. An agent preparing for a sales call could find a prospect’s recent appearances; a market-research workflow could look for statements about pricing or demand; an executive briefing tool could summarize a current episode and retain speaker attribution.

The caution is that availability is not the same as reliability. Builders should preserve the episode date, speaker label, and transcript passage alongside any generated conclusion. They should also test whether their queries retrieve the intended episode rather than merely a similarly titled show or an old discussion. Exa’s stated freshness results are useful directionally, but they do not replace evaluation on a team’s own tasks.

## Limits and uncertainty
The announcement is Exa’s product description and its benchmark report. It does not provide independent validation of the 98.4% newest-episode result or the 85.0% question-answer result, nor does it disclose the other APIs, their configurations, or the full question set. It also does not establish transcript accuracy, speaker-label accuracy, update latency, licensing terms for downstream use, or coverage gaps across shows and languages. “Soon after they air” is not a defined service-level commitment.

## Skill to build now
Add a podcast-specific retrieval check to one existing research workflow. Query for a person, company, or topic; require the workflow to return the episode date, show title, speaker name, and the transcript excerpt supporting each claim. Then compare the generated briefing with a manual review of several episodes. This small evaluation will reveal whether transcript search improves coverage and freshness for your work, and whether attribution survives the handoff from retrieval to generation.

## Reader takeaway
Exa Podcasts makes a large body of spoken, long-form material available through the same Search API surface agents already use. The immediate opportunity is not “podcast summaries” in general, but evidence-backed retrieval of recent remarks with episode and speaker context. Treat Exa’s quality claims as vendor claims, then test the corpus on your own research tasks.

## Open question
How quickly new episodes enter the index, and how accurately Exa’s transcripts and speaker labels hold up on the shows most relevant to a given workflow, remain the practical questions for adopters.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Agent Workflows](https://newruntime.com/topics/agent-workflows.md): Explore the agent-workflows topic hub.
- topic hub: [Mcp](https://newruntime.com/topics/mcp.md): Explore the mcp topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)