When a customer's chief executive, head of engineering, head of product, head of marketing, head of finance, head of operations, or independent operator appears as a guest on a podcast episode and names your product during the conversation, mentions your product during an audio interview recorded for an industry publication, references your product in a panel discussion broadcast as an audio recording, or describes your product as part of their daily workflow during a spoken-format conversation, they are delivering a category of endorsement that no marketing-elicited testimonial can replicate. The mention has been delivered with the customer's own voice, captured in audio, transcribed for indexing, archived on multiple distribution platforms, and made under the social pressure of a spoken conversation where retraction is impossible and where the surrounding context establishes the operator's identity and role. The audio recording carries acoustic, paralinguistic, and timing signals that text cannot encode, and the transcript carries the verbatim record that supports automated extraction at scale.
Almost no B2B software-tooling, developer-platform, or operator-facing-product marketing team systematically extracts product mentions from podcast episodes and audio interview transcripts. The omission is the natural extension of the same blind spots we documented in our SEC filing extraction guide, our quarterly earnings call extraction guide, our academic paper extraction guide, our patent filing extraction guide, our YouTube content extraction guide, our Reddit content extraction guide, our open-source repository extraction guide, and our Stack Overflow extraction guide. Financial disclosures cover business-context mentions made in writing under legal liability. Earnings calls cover spoken executive mentions made under analyst scrutiny. Academic papers cover research-context written mentions. Patent filings cover engineering-context mentions made under legal duress. YouTube content covers face-attached video mentions. Reddit content covers peer-scrutinized text mentions. Open-source content covers cryptographically signed engineering mentions. Stack Overflow content covers reputation-attached Q&A mentions. Podcast and audio interview content covers voice-attached, transcribable, multi-platform-archived spoken mentions made under the social pressure of an unscripted conversation that cannot be retracted after the recording is released — the ninth pillar of the structurally durable public corpus, and the only one where the customer's testimony carries acoustic and paralinguistic signals that supplement the verbatim text.
This guide describes the extraction workflow for the podcast and audio interview corpus.
Why a podcast mention beats almost every marketing-elicited testimonial
A podcast mention is a category of endorsement that has passed through filters no marketing-elicited testimonial encounters. Six properties stack to make it one of the most adversarially credible spoken-format endorsement formats in modern B2B operator-facing marketing.
First, the mention has been delivered with the customer's own voice in an unscripted conversational format. Most podcast episodes are recorded with minimal pre-scripting beyond an outline of topics. The guest's spoken statements carry the unrehearsed quality of conversational testimony, and the surrounding host questions and follow-up probes establish that the guest's statements are responses to genuine inquiry rather than prepared talking points. The unscripted-conversation property is materially stronger than the equivalent on any pre-written-testimonial format, where the customer has had time to revise the language for marketing-acceptable phrasing.
Second, the mention is permanently archived on multiple distribution platforms simultaneously. A single podcast episode is typically published to Apple Podcasts, Spotify, Google Podcasts, the show's own RSS feed, and frequently to YouTube as an audio-with-static-image upload. The episode is indexed by Listen Notes, Podchaser, and Castbox. Removing the episode from one platform does not remove it from the others, and the RSS feed publishes the episode to thousands of independent podcast clients that cache the audio locally. The multi-platform-archival property is materially stronger than the archival guarantee on any single-platform-hosted spoken content.
Third, the mention is contextualized by host introductions, episode descriptions, and show notes that establish the speaker's identity and role. Every podcast episode opens with a host introduction that names the guest, states the guest's role and company, and frequently summarizes the guest's relevant background. The episode description and show notes published on the platform pages include the same metadata in written form, and frequently include time-stamped chapter markers that index specific segments of the conversation. The contextual-metadata property is what makes podcast mentions more deployable than mentions from any format without comparable speaker-identification anchoring.
Fourth, the mention can be transcribed with high accuracy using modern speech-to-text systems. Modern speech-to-text systems (OpenAI Whisper, AssemblyAI, Deepgram, Rev) produce transcripts of podcast audio with word-error rates below five percent for clean studio recordings, and many podcast networks publish official transcripts directly on the episode page. The transcribability property is what makes the podcast corpus one of the most automation-friendly spoken-format corpora for testimonial extraction at scale.
Fifth, the mention carries acoustic and paralinguistic signals that text cannot encode. The recording captures the guest's tone of voice, pace, pauses, emphasis, and emotional register at the moment of the endorsement. A guest who emphasizes a product name with enthusiastic intonation, pauses before delivering the endorsement to mark its weight, or laughs during the description of a competing tool's failure carries signals that the transcript alone cannot convey. The acoustic-signal property is uniquely available to the spoken corpus and supports a deployment format (audio clip with caption) that no text-only corpus supports.
Sixth, the mention is associated with structured metadata that supports automated discovery at scale. Every podcast episode has structured metadata (host, guest, publication date, episode number, show name, RSS feed URL, episode URL on each platform, episode description) that podcast directories expose through search APIs and that the Listen Notes API exposes for programmatic discovery. The structured-metadata property is what makes the podcast corpus one of the most discoverable spoken-format corpora for testimonial extraction at scale.
The six podcast content locations where customer mentions appear
The podcast ecosystem has six primary content locations where a product mention can surface, and each carries a different credibility weight and a different downstream usability.
Location 1 — The guest interview episode where your customer is the named guest
A guest interview episode where the customer is the named guest is the highest credibility-dense location because the entire episode is structured around the guest's perspective and experience. A mention made by the guest during their own interview carries the full weight of the guest's authority and is delivered in a context where the host has invited the guest specifically to share their perspective. The named-guest format is the highest-weight format for podcast extraction.
Location 2 — The panel episode where your customer appears alongside other operators
A panel episode where the customer appears alongside other operators is the second-highest credibility-dense location because the panel format places the customer's mentions in direct comparison with the perspectives of other operators in the same category. A mention made during a panel discussion is implicitly compared against the perspectives offered by the other panelists, and the survival of the mention without contradiction by the other panelists is itself a signal of consensus.
Location 3 — The host-narrated episode where your customer is referenced as a case study
A host-narrated episode where the customer is referenced as a case study is the third-highest credibility-dense location because the host's narration carries the editorial weight of the show's brand. A host who describes the customer's deployment of a product is delivering a third-party endorsement that supplements the customer's direct testimony.
Location 4 — The Q&A-format episode where your customer asks or answers a question naming your product
A Q&A-format episode where the customer asks or answers a question naming the product is a moderate credibility-dense location because the Q&A format encourages brief, contextual product references in the course of answering a specific operational question. The Q&A format is lower-weight than the named-guest format but adds peer-validation in the same conversational register.
Location 5 — The roundup episode where your customer is named among a set of recommended tools
A roundup episode where the customer is named among a set of recommended tools is a moderate credibility-dense location because the roundup format places the customer's mention in the context of an editorial selection of category-defining tools. The roundup-episode-mention is less time-bound than an individual interview mention but carries the weight of being a curated, editorially vetted reference.
Location 6 — The advertisement or sponsorship read where your customer testifies to a product on behalf of the show
An advertisement or sponsorship read where the customer testifies to a product on behalf of the show is the lowest credibility-dense location for organic testimonial extraction because the format is paid. However, advertisement reads that quote a verbatim customer statement extracted from an unpaid context are deployable as long as the original unpaid context is the canonical source rather than the advertisement read itself.
The extraction workflow — eight steps from query to deployable testimonial
The podcast corpus rewards a workflow that distinguishes between the platform search (which uses episode metadata) and the transcript search (which uses verbatim spoken content). The eight-step workflow below converts a query into a deployable testimonial in a way that survives downstream review and remains attributable to the original customer.
Step 1 — Construct the product-name query and the spoken-form variant set
The first step is to construct the product-name query and the spoken-form variant set the workflow will use across all six content locations. A query for ProofShow would include the product name in its written form, the product name in its likely spoken pronunciations (proof-show, proofshow, proof show), the company name in spoken form, and common mispronunciations the speech-to-text systems may produce. The variant set should be saved as a structured artifact for reuse across all subsequent extraction sessions.
Step 2 — Run the Listen Notes search and the platform-specific transcript searches
The second step is to run the Listen Notes search for the product-name query and the platform-specific transcript searches that index transcribed content (Spotify transcripts, Apple Podcasts transcripts where available, Podscribe, and the show-specific transcript pages where networks publish official transcripts). The output is a corpus of candidate episode-level mentions that include the spoken context where the product is named.