Playbook · Podcasts

Playbook: Get booked on podcasts with agents, then turn each episode into a content pack

Find shows that fit, pitch each one from its recent episodes, keep a person on the send button, then fan the finished episode out into every deliverable at once.

Teardowns: 3Stack: Claude skills · Custom GPT · LemlistHuman approval: every pitch
Elad Hefetz · Jasz Joseph · Gigi Kenneth Verified Oct 16, 2026 Markdown copy

Who it’s for: consultants, founders and creators who want to be a guest on more podcasts without paying a PR agency.

Tools these people use: Claude skills, parallel research agents, the Apple iTunes Search API, RSS parsing, Bright Data and Lemlist (Elad Hefetz); a custom GPT in ChatGPT, MatchMaker.fm and advanced Google search (Jasz Joseph); a Claude Code skill called pod-pack, Riverside and a show-profile.yaml (Gigi Kenneth, for the episode afterwards).

The steps

  1. Teach the agent who you are and who you don’t serve. Jasz fed her custom GPT her homepage, services page, bio, past projects, ideal clients and the industries she avoids (she doesn’t work with ecommerce or B2C). She added personal details and values by voice. Source: Jasz (teardown, original)

  2. Discover shows from podcast directories, not people lists. Elad switched to the keyless Apple iTunes Search API, ran it across dozens of search terms and deduped by RSS feed URL. The first sweep found about 600 shows. Jasz’s VA researches with MatchMaker.fm and advanced Google search. Sources: Elad (teardown, original); Jasz (teardown, original)

  3. Enrich from the RSS feed. Each feed gives the owner email, website, language, last episode date and the last five episode titles. Elad calls recent titles “the single best signal for personalizing a pitch.” About 65-70% of shows exposed a direct owner email. Source: Elad (teardown, original)

  4. Score fit and drop dead shows. Elad scores each show on fit, reachability, recency and audience size, weighted toward fit, drops shows dormant beyond 90 days, and matches each to one angle from his talking-point library. Jasz’s GPT decides whether a show fits and explains its reasoning so the VA can override it. Sources: Elad (teardown, original); Jasz (teardown, original)

  5. Pitch from the show’s recent episodes, with fixed rules. Jasz’s GPT asks for the last three to five episode titles and descriptions, then whether outreach is by form or email, and drafts accordingly. Elad’s rules: “never pitch Airfleet”, paraphrase a recent episode, 90-130 word emails, 50-70 word LinkedIn notes, the host’s real first name. Sources: Jasz (teardown, original); Elad (teardown, original)

  6. Approve before anything sends. Elad: “Every draft is previewed before anything sends.” He approves the batch. Jasz’s VA does the final review and flags form questions only she can answer, like “Who are your three biggest mentors?” Sources: Elad (teardown, original); Jasz (teardown, original)

  7. Sequence follow-ups and stop on reply. Elad’s Lemlist sequence runs on days 0, 2, 5, 9 and 14 across email and LinkedIn, stops when someone replies, and sends from a dedicated domain. He says “Several yeses came on the follow-up, not the first email.” Source: Elad (teardown, original)

  8. Track every pitch. Jasz’s GPT writes a summary of each pitch that goes into a shared doc by hand, and the VA screenshots submitted forms. Source: Jasz (teardown, original)

  9. Correct the agent, not just the output. Jasz fixed tone and word choice in the GPT itself and refines it in weekly check-ins with her VA, adding new client stories and lessons from past appearances. Source: Jasz (teardown, original)

  10. After the episode, fan it out. Gigi’s pod-pack reads a show profile and the cleaned transcript once, then spawns one background agent per deliverable at the same time: YouTube titles and description, tags, a Spotify description, a blog post with an FAQ, LinkedIn, Instagram and TikTok posts, a newsletter blurb, pull-quotes and chapters. It commits the pack to a content repo, one folder per episode. Source: Gigi (teardown, original)

Postby’s suggestion: steps 1-9 come from guests and step 10 from a host. As a guest, you can only run step 10 if the host shares the recording or transcript. The sources don’t cover that hand-off.

Human approval points

  • Elad: previews every draft and approves the batch before Lemlist sends.
  • Jasz: her VA reviews everything before it goes out; she answers personal form questions herself. The story says pitches now go out without her own review.
  • Gigi: pod-pack asks before the Riverside export, publishing or opening a pull request, and flags uncertain spellings and legal-risk claims. “You still make the calls on the final title and anything with reputation weight.”

Common failure points (as reported)

  • Reusing an event-prospecting pipeline for podcasts didn’t work: “Podcasts have no public attendee roster” (Elad).
  • Early pitches weren’t customised enough to each audience, which is why Jasz’s GPT now asks the VA questions first.

Caveats carried over

  • Elad: the source text doesn’t show the author’s name; attribution to Elad Hefetz comes from page byline metadata. The number of hosts pitched isn’t stated, so the “almost 70 percent” reply rate has an unknown denominator. Airfleet company blog. X: Unverified.
  • Jasz: a third-party story on an ActiveCampaign site, not her own write-up. The jump from 0 to 10 bookings also coincided with a new research tool (MatchMaker.fm). X: Unverified.
  • Gigi: no results stated; free MIT-licensed tool. X: Unverified.

Examples

Get one teardown a week.

Same format, a new operator every issue. Double opt-in.