Podcast Guest Pitch Drafter

When to use this prompt

When you have a list of podcasts you’d like to be on and need to draft pitches that don’t sound like every other pitch the host gets. Most guest pitches fail for the same reason: they describe the guest’s expertise without ever proving the writer has listened to the show. Hosts can spot a templated pitch in three seconds and they delete those.

This prompt forces the pitch to ground itself in the host’s recent episodes — naming the show, naming guests the host has had, and proposing topics that build on what the host has already covered. That single discipline takes a pitch from a 5% reply rate to something far higher.

The prompt

<role>Outreach writer specializing in podcast guest pitches grounded in the host's actual recent episodes.</role>

<task>Draft a pitch from the guest below to the podcast below. The pitch must reference at least two specific recent episodes by name and propose three episode topics that build on, complement, or respectfully push back against what the host has already covered.</task>

<inputs>
<podcast_name>[NAME OF SHOW]</podcast_name>
<host_name>[HOST NAME]</host_name>
<recent_episodes>
[3 to 5 recent episode titles with one-sentence summaries each. Pull these from the show's feed before running the prompt; do not let the model invent episodes.]
</recent_episodes>
<guest>
<name>[GUEST NAME]</name>
<role>[GUEST'S CURRENT ROLE AND COMPANY]</role>
<expertise>[2-3 specific areas the guest can talk about, with named work or named achievements]</expertise>
<unique_angles>[1-2 things the guest has done or said that no other guest in the category has]</unique_angles>
</guest>
</inputs>

<instructions>
1. Open with one sentence naming a specific recent episode the guest genuinely engaged with and one specific takeaway from it. The takeaway must be from the supplied <recent_episodes>; do not invent.
2. Connect that takeaway to the guest's expertise in one sentence. Make the connection specific. Not "I work in this space" — name the specific overlap.
3. Propose three potential episode topics. Each topic must:
   - Build on or respectfully push against a theme from the supplied recent episodes
   - Have a working title (≤60 characters)
   - Include a one-sentence "why this would land with your audience" rationale
4. Close with: a one-line offer (when the guest is available, or what specific takeaway the audience would get), the guest's relevant link (LinkedIn or personal site), and a sign-off. Do not use "Best regards" or other generic closers.
5. Constraints:
   - Total length: 180 to 260 words. Hosts skim; pitches over 300 words get archived without reading.
   - Do not say "I'm a huge fan of the show." Hosts hear that 50 times a day. Prove it instead.
   - Do not pitch a topic that the host has already covered with a similar angle. If two of the recent episodes overlap with a proposed topic, drop that topic and propose a different one.
   - Do not invent episodes, statistics, or details about the host. If the inputs lack a detail you'd want, omit rather than fabricate.
6. Subject line: separate, ≤55 characters, names the proposed topic angle. Not "Guest pitch for [show name]."
</instructions>

<output_format>
**Subject line:** [≤55 characters]

**Body:**
[180-260 word pitch following the structure above]

**Topic ideas (in order of strength):**
1. [Working title] — [why-it-lands rationale]
2. [Working title] — [rationale]
3. [Working title] — [rationale]

**Quality check:**
- Specific recent episode referenced: [Yes / No, episode name]
- Topics built on host's recent themes: [Yes / No, brief justification]
- Word count: XXX
- Did the pitch use any forbidden phrases ("huge fan", "guest pitch")? [Yes / No]
</output_format>

How it works

The constraint to reference specific recent episodes is the entire pitch’s credibility. Hosts can tell within one sentence whether a pitch was written by someone who listens or someone who scraped a directory. Forcing the model to ground every claim in supplied episodes prevents the templated-pitch failure mode.

The “respectfully push against” option in step 3 is intentional. Hosts get pitched too many “I agree with everything you’ve said” angles. A guest who can offer a respectful counter-take is more interesting and gets booked more often.

The forbidden phrases list (“huge fan”, “guest pitch”) catches the most common dead-on-arrival language. Listing them explicitly is a 2026 best practice — frontier models actually obey blacklists when they’re stated literally.

The quality check at the end is a forcing function. It catches pitches the model fabricated to look complete. If the “specific recent episode referenced” field is empty, the pitch did not do the work and should be regenerated.

Example output

Subject line: A counter-take on the Episode 47 measurement frame

Body:

Episode 47 with Ana Pirvu landed on the right diagnosis but the wrong metric — that the funnel is breaking at the awareness layer, not the conversion layer. The metric Ana proposed (qualified pipeline) is lagging. The leading indicator is mention-share inside AI engines.

I run SEO and GEO at Leah and have spent the last 18 months building the measurement system Ana described as “what we need but don’t have yet.” Three of our enterprise clients run a weekly mention-share dashboard now. The differences in pipeline reliability between teams that measure it and teams that don’t are stark.

Three angles for an episode:

1. The metric that broke first. Why CTR is now a lagging indicator and what replaces it.
2. A live audit. I’d run a 50-prompt mention-share audit on a brand of your choice on the show.
3. Disagreeing with Episode 23. Joel claimed AI Overviews would reduce conversion volume. The data suggests the opposite. I’d argue why.

Available the next two Tuesdays.

Arthur Dosik https://www.linkedin.com/in/arthurdosik

Topic ideas (in order of strength):

1. Disagreeing with Episode 23 — Joel’s claim has a measurable counter, audience loves a real disagreement.
2. The metric that broke first — direct extension of Ana’s framing in Episode 47.
3. A live audit — high production value, novel format for the show.

Quality check:

• Specific recent episode referenced: Yes, Episode 47 with Ana Pirvu, and Episode 23.
• Topics built on host’s recent themes: Yes, all three reference supplied episodes.
• Word count: 234
• Forbidden phrases: No.

Variations

• Cold outreach version: Adapt for a generic LinkedIn DM or email to someone not running a podcast — replace “recent episodes” with “recent posts” or “recent speaking” and ground the pitch in those instead.
• Resurrect-an-old-pitch mode: Paste a pitch you sent that didn’t get a reply. Ask the model to identify the three weakest sentences and rewrite each one.
• Show-feed-first mode: Instead of supplying recent episodes manually, run this in a model with live web access (Perplexity, Claude with browsing) and ask it to fetch the last 5 episodes from the show’s feed before drafting.