Podcast Guest Outreach With AI: Research, Pitch and Follow-Up

A podcast producer can use AI to research each guest from sources you supply, draft a pitch built around one specific, verifiable reason for inviting them, and run a three-email sequence over about two weeks, logged in a tracker. To avoid spamming, send small batches, stop after the third message, honour every opt-out and keep an accurate sender and subject line.

The research is where AI saves the most time, and personalisation is where it does the most damage if you let it guess. A model asked to "personalise this pitch" will invent a podcast appearance or a quote. A model asked to summarise three pages you pasted will not. The whole workflow below separates those two jobs: AI reads and summarises what you give it, and you decide what goes in the email.

Follow me on Instagram@sagnikteaches

Build a target list before opening any AI tool

The quality of outreach is set by who you pick, not by how well you write. Decide three things first: the episode theme or series you are booking for, the kind of guest who could teach your listeners something they cannot get elsewhere, and what you can offer. Write them in three lines at the top of a sheet.

Connect on LinkedInSagnik Bhattacharya

Then list candidates with a reason each. A good entry reads: "Runs a four-person bookkeeping firm; wrote a detailed post on why clients send documents late; matches the episode on year-end admin." A poor entry reads: "Accountant, 10k followers". If you cannot write a reason without checking the profile, the candidate is not ready. Aim for a list of 20 to 30 for a batch. Bigger lists produce shallow research and tempt you to send in bulk.

Subscribe on YouTube@sagnikteaches

A research prompt that stays inside the evidence

For each candidate, collect two or three sources yourself: their website, a recent article or episode transcript, and a professional profile. Paste excerpts into the prompt. Do not ask the AI to browse and remember them, because that is where invented details creep in.

You are a research assistant for a podcast producer. Use ONLY the text
pasted below. If something is not in the text, write "not found".

Guest: [name]
Our show: [one-line description]
Episode theme we are booking for: [theme]

Return:
1. Who they are, in two plain sentences.
2. Three specific things they have said or written that fit the theme,
   each with a short quote or paraphrase and the source name.
3. One angle only they could speak to, and why.
4. Anything that suggests they may not want to appear (for example a note
   saying they do not accept guest requests).
5. A suggested one-sentence reason to invite them. Do not flatter.

SOURCES:
[paste excerpts, each labelled with its source name]

A realistic output for a bookkeeping-firm owner (illustrative) would list two paraphrases from a published post, a note that their website says they are not taking interviews until spring, and a suggested reason: "Your post on why clients send year-end documents late described a checklist we haven't heard on air." The fourth item is the one most pitch templates miss. A note saying "no guest requests" is a reason to pause and come back, not a reason to write more cleverly.

Check one thing before moving on: every quote in the output must appear in the sources you pasted. Search for two of them. If the model produced a line that is not in the text, tighten the prompt and rerun.

The three-email sequence

Three emails over about two weeks is enough. The first makes the case, the second adds something new, the third gives a clean exit. None should be longer than 120 words.

Email one: the reason and the ask

Subject: Your post on year-end paperwork (a guest idea)

Hi [first name],

I produce [show name], a weekly show for [audience, in six words]. Your post
on why clients send year-end documents late described a checklist we have
not heard discussed on air.

Would you be open to a 30-minute conversation about it? Recording is remote,
we send questions in advance, and you review the edit before it goes out.

If it is not a fit, no problem at all.

[your name], [show name], [postal address]
Reply "no thanks" and I will not contact you again.

Notice what is in it: one specific reason, a plain description of the ask, the time cost, and a clear exit. The AI can draft this from your research output, but you write the first line yourself. Keep your sender name and address accurate and leave out phrases such as "Re:" that imply an earlier conversation.

Email two: add value, not pressure

Send it five to seven days later only if there was no reply. It should contain something new: the proposed questions, a link to one relevant past episode, or the date range you are recording. "Just bumping this to the top of your inbox" adds nothing and is the most common reason follow-ups irritate people. Draft with AI by asking for "one new useful piece of information, three sentences, no pressure language".

Email three: the graceful close

Another week later. State that this is the last message, restate the offer in one line, and say you will leave it there. Some of your best bookings arrive in response to this email, because it removes the sense of being chased. Then stop. Mark the contact "closed, no reply" in the tracker with the date.

Personalisation checks before any email is sent

Run every draft through the same five questions. Each takes seconds, and together they catch most failures.

1. Is the personal line true, and could I show where I found it?
2. Is it specific enough that it could not be sent to anyone else?
3. Does it describe what the guest gets, not just what the show wants?
4. Would I be comfortable if the guest read my research notes about them?
5. Is there a clear and easy way to decline?

The fourth question is the useful one. Research should be professional: published work, public talks, stated interests in a professional context. Do not reference family details, location of a home, health, or anything the person has not published for professional purposes. If the AI surfaces such material, delete it from your notes.

A tracker that prevents double-sending

A spreadsheet is enough. The columns that matter: guest name, organisation, source of the contact address, research date, reason for invitation, email 1 date, email 2 date, email 3 date, status, opt-out flag, and notes. Three statuses cover almost everything: "in sequence", "replied" and "closed". Add a separate "do not contact" tab that every batch is checked against before sending. If a guest opts out, they go on it at once, and you never remove them.

Recording the source of each address matters. If a guest asks where you got their email, you can answer in one line. If you cannot, that is a sign you should not have used it.

Consent and opt-out basics

This section is general information, not legal advice; confirm with a qualified adviser for your situation. The rules differ between countries, and several may apply if your guests are in different places.

In the US, the Federal Trade Commission's guide to CAN-SPAM says the law covers commercial messages, meaning email whose primary purpose is advertising or promoting a commercial product or service. It requires accurate header information (From, To, Reply-To), no deceptive subject lines, a clear disclosure when the message is an advertisement, your valid physical postal address, and a clear way to opt out. Opt-outs must be honoured within 10 business days, and you cannot charge a fee, ask for personal information beyond an email address, or require any step other than a reply email or a visit to a single web page. The FTC also notes that the law makes no exception for business-to-business email. Whether a guest invitation is "primarily commercial" depends on the facts. If your show promotes your company's services and the email leans on that, treat it as covered. The cheap and safe approach is to follow the requirements for every outreach email.

In the UK, the Privacy and Electronic Communications Regulations (PECR) govern electronic marketing. The regulator's guidance says marketing emails to individuals need their specific consent, apart from a limited "soft opt-in" for your own previous customers that never covers new contacts, and it treats sole traders and some partnerships as individuals. Companies, limited liability partnerships and government bodies can be emailed without prior consent, though the guidance calls it good practice to keep a "do not email" list of any that object, and data-protection law may still apply when the address belongs to a named employee. The guidance itself says it is under review after recent changes to the law, so read the current version. A short, relevant, non-promotional invitation is a different thing from marketing, but the boundary is not always obvious, so check the regulator's guidance or ask an adviser before running a large campaign. In all cases: identify yourself, say why you are writing, keep the data you hold to what you need, and delete it when the sequence ends without a reply.

Keeping the volume honest

Small batches protect both reputation and deliverability. Send 10 to 15 emails a day from a real mailbox, not a bulk platform, and stop if you see bounces or spam complaints. Use a separate sending address on your show's domain rather than a personal one, and make sure the domain's basic email authentication is set up, which your host or email provider can walk you through. Do not buy lists. If an address came from a database that the guest never joined, a polite email can still come across as spam, however well it was written. For the technical and filtering side, see how to write cold emails with AI that stay out of spam.

Before and after: the same guest, two pitches

Here is the difference research makes, using an invented guest and an invented show for illustration. The first pitch is the kind that fills inboxes: "Hi, I love your work! I host a podcast with thousands of listeners and think you'd be a great guest. Let me know if you're interested!" It names nothing, offers nothing and could have gone to anyone. It also contains an unverifiable claim about listeners.

The second is built from the research prompt: "Hi [first name], your post on why clients send year-end documents late described a checklist we have not heard on air. I produce a weekly show for owners of small firms. Would a 30-minute chat about it suit you? You would see the questions in advance and approve the edit." The only extra work was five minutes of reading and one sentence of writing, but it changes the answer from "who is this?" to "that is a fair question".

A realistic mistake and how it showed up

A producer asks the AI, without pasting any sources, to "write a personalised pitch for the founder of a gift shop who has been on podcasts before". The model writes: "I loved your recent interview about building the brand from your kitchen table." The founder has never given such an interview and started the shop in a unit on a high street. The email lands with an owner who now believes the producer did not do the work, and the reply is one line asking to be removed. The failure is traceable: no sources went in, so the model produced plausible filler. The fix is the rule in the research prompt, "use only the text pasted below", plus the check that every quote appears in the sources. Treat any personal detail that you cannot point to as an error.

Dealing with publicists, agents and assistants

Many guests worth booking are reached through someone else. If a contact page says to write to a manager or a publicist, write to them and follow their format. Expect a shorter pitch to be better: the gatekeeper needs the show description, the audience, the topic, the time ask and the dates. Include a link to one episode that resembles the planned conversation, since an agent will check it. Ask the AI to draft a "gatekeeper version" of your pitch at 80 words with the guest's reason removed and the logistics emphasised. Keep the sequence to two follow-ups here as well. A publicist who finds you polite and organised will often come back with another client next quarter.

Measuring whether the process is working

Track four numbers for each batch: how many pitches you sent, how many replies you received, how many bookings you made, and how many guests actually recorded. Divide to get a reply rate and a booking rate. With small batches the numbers will be noisy, so compare across three or four batches before changing anything. If replies are low, the usual culprits are the guest list and the first line. If replies are fine but bookings are low, the ask is too large, so offer a shorter recording or send questions in advance. If bookings are fine but recordings fall through, the confirmation and reminder steps need work.

Keep a note of what each reply said. "Not now" and "wrong fit" are different answers, and they point to different fixes. After twenty replies, paste the notes into an AI tool and ask it to group them by theme. A line such as "five of eight declines mention the time commitment" is a reason to shorten your ask, and you could not have seen it without gathering the text. The time this takes is small, and it turns outreach from a guessing game into something you can improve.

What to automate and what to keep by hand

Automate the dull parts: reading sources, summarising them, drafting a first version, logging dates and prompting you when a follow-up is due. Keep by hand the parts that carry your name: choosing guests, writing the personal line, deciding when to stop and answering replies. A scheduler that sends email two on its own is fine if it checks the tracker for a reply first and the sequence stops the moment someone answers or opts out. If you cannot guarantee that stop, do not automate the sending. A guest who replies on Tuesday and receives a "just following up" on Wednesday will not forget it.

Replies: what to do with yes, no and maybe

A yes needs a quick, clear next step: two or three time slots, a recording link, a one-page briefing, and a note on how the edit will be reviewed. A no needs a thank-you and nothing else; add them to the do-not-contact tab. A maybe is the most common reply, such as "not now, maybe in spring". Put a date in the tracker, set a reminder, and when it arrives write a fresh email that references something new rather than the old thread. AI can help draft the briefing and propose a reminder cadence, and automating follow-ups without being pushy covers the cadence logic in more detail.

Keep reply handling out of your general inbox so it does not get lost. A saved search or a label for outreach replies makes this easy, and managing your inbox with AI shows a triage routine you can borrow.

A week of outreach for a small business show

Say a producer runs a weekly show for small-business owners and wants three guests for next month. On Monday she builds a list of 24 candidates, each with a reason, in 90 minutes. Tuesday she pastes sources for 12 of them into the research prompt and checks the quotes, about two minutes each. Wednesday she writes the personal line for each and sends the first ten emails. The following Wednesday she sends email two to non-responders with the proposed questions. A week later she sends the last email to those still silent and closes them. Over the cycle, she might expect a handful of replies from 24 pitches (illustrative; response rates vary widely by show, guest and timing), enough to fill three slots. If it is lower, the usual fix is a better reason or a different guest list, not a higher send volume.

After the booking: reduce no-shows and make the guest look good

Outreach does not end at "yes". Send a confirmation with the recording time in the guest's time zone, a link, the questions, and a short note on how they can promote the episode. A day before, send a brief reminder. After publishing, send the guest the links and a clip or two, which they are more likely to share if the captions are right, as in AI for podcast clips. A guest who is treated well is also your best source of introductions, and a warm introduction beats any cold sequence.

Questions producers ask about guest outreach

How many follow-ups are too many?

Two follow-ups after the first email, spaced about a week apart, is a sensible ceiling for a cold guest pitch. After the third message with no reply, stop and note it in your tracker. Revisit after a real reason to write again, such as a new episode relevant to them.

Does CAN-SPAM apply to guest invitations?

It applies to commercial messages whose primary purpose is advertising or promoting a product or service. A genuine invitation to be a guest may not be primarily commercial, but the line is not always clear. Use an accurate sender, honest subject line, your postal address and an easy opt-out anyway, and ask an adviser if unsure.

Can the AI write the whole email?

It can draft one, but the personal line has to come from something you read or heard. Use AI to find and summarise evidence, then write or edit the opening sentence yourself. Emails that read as generated get ignored, and an invented detail about a guest can end the relationship before it begins.

Is it fine to use a guest's work email?

It depends on who the guest is and where they are. Rules differ for individuals and for staff of organisations, and data-protection law may apply. Prefer addresses the person publishes for professional contact, record where you found it, and offer a clear way to say no.

Further reads

• How to Write Podcast Show Notes With AI: A Template for Producers — Turn the finished episode into show notes the guest can share.
• Podcast Sponsorship Media Kit: Build One With AI and Real Numbers — Use a strong guest line-up as evidence in a sponsor pitch.
• Can AI Edit a Podcast? What It Does Well and Where It Fails — Know what AI editing can do before you promise a guest a quick edit.
• How Staffing Agencies Automate Candidate Follow-Up With AI — Another follow-up cadence that avoids nagging.
• How Wedding Venues Use AI for Viewings, Enquiries and Follow-Ups — See how enquiry follow-ups are timed in another trade.
• Descript vs Riverside for Podcast Producers: Which to Standardise On — How to choose between Descript and Riverside for client shows: a timed test, plan limits, multi-client set-up, AI features and data terms.
• AI Link Building Outreach: Personalised Without Spam — A compliant outreach routine for SEO agencies: research prospects, pick one angle each, send three emails at most, and check Google's spam policy and email law.
• AI That Creates Podcasts: Can It Replace a Production Company? — AI can turn documents into a podcast episode in minutes. Here is how to test whether it is good enough to publish for clients, and when to hire humans.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: US Federal Trade Commission, CAN-SPAM Act compliance guide for business; the UK regulator's guidance on electronic mail marketing under PECR (archived copy, October 2026). Confirm both with an adviser before a large campaign.