How to Vet a Podcast Guest

How do you vet a podcast guest?

Vet the person, not the paperwork. Listen to five minutes of them speaking unscripted, confirm the one or two claims that are checkable in public, ask for the angle in their own words rather than a topic, and read how they handle a specific question before you book. A guest who cannot do those four things well will not become a good episode because their bio was accurate.

The checks are ordered by what they cost you and what they rule out. The first one is the cheapest and it eliminates most of what you would regret.

The appearance list is deliberately not on that table. It is worth a minute if it is easy to check, and the next section is the measurement of how often it is.

Why checking previous appearances mostly does not work

A guest says they have been on a dozen podcasts. You pick two, open them, and look for their name. This is where the check breaks, and the reason is not the guest: it is that most shows do not publish their guest history where a stranger can see it in the time a stranger will spend looking.

We can put a number on this because we ran the same read twice at two depths over the same feeds. Reading each show's most recent 5 episodes finds 91 of the 197 shows that demonstrably do book guests, which is 46.2% of them. Getting to half of them takes 7 episodes. Getting to nine in ten takes 30, which is deeper than anybody checks by hand for a booking they have not made yet.

Two more findings push the same way. Of the shows whose own description claims a guest format, only 16% have recent episode titles that show it, so a show's blurb is not a shortcut. And show notes are only a partial rescue: 29% of the 360 feeds we could read notes for credit a guest by name in a recent episode at all.

Then there is the trace an appearance leaves. Among the 197 shows in our read that publish guest episodes, 64% link out from every guest episode and 16% never link out from any of them. So "there is no link back to them anywhere" is a normal outcome on roughly one guest-booking show in six, and treating it as a red flag rejects real guests for their hosts' habits.

The honest conclusion is not that guests lie about their appearances. It is that you cannot resolve the claim quickly, so it should not be the check your booking decision rests on. Spend the minute if the link is right there in the pitch. Do not spend twenty minutes proving a negative.

• The pitch names a specific episode with a link: open it. That is a confirmed appearance in ten seconds.
• The pitch names shows but no episodes: ask for one link. A guest who has been on a dozen shows can produce one in a sentence.
• You cannot find them and they cannot produce a link: that is a reason to lean harder on the listening check, not an accusation.

Check one: listen to five minutes of them speaking

This is the whole job in one step and almost nobody does it first. You are not booking a bio, you are booking a voice for an hour, and the only thing that predicts an hour of good audio is some minutes of existing audio.

Find anything unscripted: a conference talk, a webinar recording, a livestream, a video with them answering questions, a previous appearance if there is one you can open. A polished keynote is worth less than a shaky Q and A, because the Q and A shows you what happens when they are not driving.

• Do they finish a thought, or do they restart the sentence three times?
• Do they answer the question that was asked, or the one they prepared for?
• Do they tell a story with people and specifics in it, or do they speak entirely in categories?
• Would you interrupt them? A guest you cannot interrupt makes a monologue, not an interview.

If there is genuinely no audio or video of this person anywhere, that is not a rejection. It is a reason to do a ten-minute call before you book, which is cheaper than a bad recording and much cheaper than an unpublishable one.

Check two: confirm the claims that are actually checkable

Some claims resolve in seconds and some do not resolve at all. Sorting them is most of the work, and it stops you burning the afternoon on the unresolvable ones.

The rows at the bottom are the ones to handle deliberately. A guest whose value is a lived experience cannot document it and should not be asked to, and a guest whose pitch rests on a growth number should be asked what the number counts, in the pitch thread, before you book. If the answer is vague there, it will be vague on tape, where you cannot edit around it without editing out the episode.

The other thing worth confirming is boring and saves more recordings than any of the above: that the person who pitched you is the person who will show up. Ask for the name and email of whoever will be on the call. Pitches sent by an agency or a publicist are perfectly legitimate, and the only failure mode is discovering on the day that the booking was made for a different person than the one in the bio.

Check three: ask for the angle, not the topic

A topic is a category. An angle is a claim somebody could disagree with. Most weak episodes are booked on a topic and discovered to have no angle in the first ten minutes of the recording, which is the most expensive place to find out.

So ask for it in writing before you book, in the guest's own words, and read the answer as the audition it is. One question does the work.

Thanks for the pitch, this could be a fit.

Before I put a date in: what is the one thing you would want my listeners to disagree with you about, and what changed your mind about it?

A few sentences is plenty. If it lands I will send times.

You are looking for three things in the reply. A specific claim rather than a subject. A reason they hold it that involves something that happened to them. And evidence that they read your question rather than pasting a paragraph they send everyone.

• A strong answer names a position and a moment. Book it.
• A generic answer that restates the topic is a guest who will need you to carry the hour. Book it only if you want to carry the hour.
• No answer, or an answer from a different person than the one who will record, is the cheapest no you will ever get.

This check also does something the others cannot: it tells you what the episode is called. If you cannot write a title from the reply, there is no episode yet, and one more email is much cheaper than one more recording.

Check four: fit, and the small conduct screen

Fit is the check hosts skip because it feels obvious and then get wrong in the same way every time: they book an interesting person for an audience that did not ask for them. The test is not whether you find the guest interesting. It is whether you can name the episode a specific existing listener would send to a friend.

• Can you name the listener this episode is for, without describing your whole audience?
• Does the guest overlap the subjects your best-performing episodes already cover?
• Would you have gone looking for this person if they had never written to you?

Then the conduct screen, which is short and which almost every pitch passes. Search the person's name alongside your show's subject and read the first page. You are not conducting an investigation and you will not find one; you are checking that nothing about booking them would be a surprise to your audience. If something does surface, decide with your own judgement rather than a rule, because a search result is a fragment and your listeners will hold you responsible for the whole booking.

Two practical guardrails that are worth having as policy rather than as a decision made under time pressure. First, decide in advance whether you accept payment for a booking, and say so publicly if you do, because paid placements are advertising and your audience is entitled to know which is which. Second, do not let an unbooked pitch thread become a research project: our own read is that a stranger's history is often unverifiable, so a screen that demands proof will mostly select for guests with publicists.

How to say no to a guest without burning it

Most of these screens end in a no, and the no is a bigger part of the job than the yes. Guests talk to each other exactly as much as hosts do, and the person who is wrong for this episode is sometimes right for a later one or knows the person who is.

Decline on fit and on scheduling, never on quality, and never explain the screen. Nobody has ever been improved by hearing that their answer was generic.

Thanks for the pitch, and for the detail in it.

I am booked out for the season and the run I am planning is heavier on [subject] than on [their subject], so I am going to pass this time rather than sit on it.

If that changes I will come back to you. And if you know anyone working on [subject], do send them my way.

The last line is the part that pays. A declined guest who sends you a better one has turned your no into a booking, which is the version of this exchange worth optimising for.

The whole screen, in order

Fifteen minutes, before you send a date:

• Find five minutes of them speaking unscripted and listen to it. Everything else is a proxy for this.
• Confirm the one or two claims that resolve in public. Skip the ones that do not; the measurement above is why.
• Send the angle question and read the reply as an audition.
• Name the listener the episode is for. If you cannot, there is no episode.
• Search their name with your subject, read the first page, and use your judgement.

And one thing worth saying plainly at the end of a page full of failure rates: the rates above are about feeds, never about people. Most pitches come from real people who would make a decent episode, and the screen exists to make the yes cheap to give rather than to find a reason for no.

Sources and benchmark date

Every figure above is derived at build time from our own read of 362 public podcast feeds between July 16, 2026 to August 26, 2026, not typed in. The method, the limitations and the raw data are public:

• Which Podcasts Actually Book Guests? (our study)

  Original analysis of 362 public podcast RSS feeds from a 384-show directory. Source for the detection curve, the says-against-shows split, the show-notes rate and the link-out habits on this page. Method, limitations and the full dataset are published with it under CC BY 4.0.

• How many episodes should you check before you pitch?

  The detection curve in full, including the depths at which it flattens and the ceiling the deeper read still leaves. It is written for a guest checking a show, and the arithmetic is identical for a host checking whether a guest's claimed appearance is visible.

• The RSS podcast namespace (episode titles, notes and dates)

  Apple's specification for the podcast feed tags. The titles, show notes and publication dates every measurement on this page reads are standard fields a show publishes for the apps that carry it, which is why the same checks run on any show rather than only on shows listed somewhere particular.

Frequently asked questions