Ever since the start of this year, I've been flooded with emails of people trying to get on Pretrained. Sometimes people pitch themselves and sometimes it's marketing agencies pitching talent that they work with. Here's one email:

Hey Pierce,

Your recent conversation with Richard around the agent swarm era had me hooked, especially the debate about moving from simple ChatGPT queries to orchestrating hundreds of parallel agents. The parallel you made to corporate org structures as a model for agent design struck a chord with what we've built at Tuio. We rebuilt insurance from the ground up as an AI-native company, where a team of proprietary agents like Leia, Watson, and George run over 80 percent of customer interactions and 85 percent of claims end to end.

I think your listeners would appreciate a transparent, business-grounded take on what happens when a heavily regulated world lets AI handle the "machinery" behind risk, compliance, and human experience. We've proven that if you design the swarms right, humans and machines actually debate policy decisions (sometimes literally in Slack), and a century-old sector suddenly gets fast, fair, and accessible. Our model is operational, not theory, and has translated into much stronger retention and lower costs than legacy brands.

I'd love to jam on topics like autonomous operations, ethics and human oversight, or how agent swarms can create profitable businesses, not just tech demos. My audience of 5,000+ business leaders keeps pushing for clarity without hype, and I'd bring that same approach to Pretrained.

This pitch is just about what I'd do if I were in a similar position: flattery, establish legitimacy, throw out some ideas that you could talk about together. Aristotle would be proud. Pathos, logos, and ethos all baked into one email. They even listened to our episode!

But obviously - it's complete bullshit.

If you've spent enough time with LLMs you can smell the slop from a mile away. But as far as slop goes, this one is actually decent. They clearly read some transcript of our discussion to reference the themes¹. The main tell is how verbose it is. No real founder has time to write that much. I figured someone hacked together some Claude Code or OpenClaw workflow and called it a day.

But after enough of these emails, I started seeing a pattern:

• Most of them have 3 emails total: one original, two follow ups.
• Most offer some network access to distribute your podcast to their community and increase your reach.
• It's mostly founders in AI-related industries but a handful of academics.

After maybe the twentieth email I finally just asked what was up:

Hey Eden, Thanks for the message! Out of curiosity what service are you using to send these generated emails? See a lot of them in my inbox.

I tried to avoid a more explicit dig along the lines of why the hell are you filling my inbox with AI generated garbage??. And their response:

Hi Pierce, Thank you for your response. Regarding your question, the software we use is Podpitch.

Credit for the transparency. But what on earth is Podpitch? And why on earth are you not embarrassed by being called out that your pitch is so obviously AI? I shudder at the thought of letting an autonomous agent fire off emails to the world with my name attached. Your email signature is supposed to carry some weight - it's a proxy for your judgment, your taste, your ability to communicate.

Sigh, let's look at PodPitch

PodPitch bills itself as an automated podcast booking service. You sign up, it scrapes podcast directories to build a database of shows and transcripts, then generates and sends personalized cold emails to hosts on your behalf. I suppose I left my podcast email public at some point in the RSS feed - although it's pretty easy to guess it once you know my name.

Their homepage stats sent me, mostly in a bad way:

• 537K Pitches Sent
• 67% Open Rate
• 23% Reply Rate
• 0.04 Seconds to send an email

The open rate I can understand. Most podcasts don't get much inbound traffic, so when something shows up that looks like a real human pitch you're going to read it. But the 23% reply rate is a surprise. Nearly a quarter of podcast hosts are engaging with these automated pitches. Are they believing the flattery? Is it the promise of reciprocity, where they figure they'll promote each other's audiences?

Or maybe they just can't tell it's AI in the first place.

The economics

Let's work backwards from those numbers. If PodPitch has sent 537K pitches at a 23% reply rate, that's 123K replies. Not all replies convert to bookings - some are rejections, some are hosts asking follow-up questions that never go anywhere - but even a conservative 10-20% booking rate from replies means somewhere between 12K and 25K podcast appearances facilitated.

That's a lot of episodes. Dare I say inauthentic episodes, geared more to selling the listener something than actually having a real conversation.

PodPitch also isn't the cheapest option at $200-500/month² depending on your plan. But the siren's call of getting booked on a podcast is obviously too strong to resist. Especially when stacked up against corporate marketing budgets. If you get booked on a single mid-sized show that reaches your audience, or a few small ones, the cost per impression is almost zero compared to alternatives. CAC rates have already skyrocketed post-generative AI as it's way easier to generate online content, trust capital is more built on reputation, and there are more new startups fighting over the same customers.

This is effectively the arithmetic that has long been true for spam or phishing. The cost to send approaches zero but the cost to receive stays constant. And the upside to a single win can make your whole month.

Trust erosion

I was under the impression that podcasting had largely survived the content quality collapse that hit blogging, SEO, and social media. Part of that is because audio is harder to fake at scale.³ But the guest acquisition pipeline was also a natural quality filter. If someone took the time to write you a thoughtful pitch, referencing specific episodes and explaining why they'd be a good fit, that effort was itself a signal. It demonstrated that they cared enough to do the work.

PodPitch and services like it strip that signal entirely. When every pitch looks thoughtful because an LLM generated it, thoughtfulness stops being informative. It's Goodhart's Law applied to email: once you optimize for the appearance of genuine interest, genuine interest stops being measurable.

The result is predictable. Hosts who get burned by a few bad guests from automated pitches will start ignoring all cold outreach. The people who genuinely would be great guests - who actually did listen to your show and have something real to contribute - get drowned out by volume. The hosts retreat to their own networks, guest diversity shrinks, and shows get more insular.

We've seen this pattern play out in every channel that gets flooded with automated outreach. Cold email for sales, LinkedIn messages, Twitter DMs. The channel works well when volume is naturally limited by effort. Automation removes the effort constraint, volume explodes, trust collapses, and eventually the channel becomes useless for everyone - including the people doing the automating.

A post-publish amendment

Right as I clicked publish on this post, I got the following email:

Hi Pierce,

Really flattered that you were checking us out...

We’re great at finding emails quickly, as you may have noticed! More on that later.

Here's my guess: you were checking out PodPitch so you could directly appear on podcasts that your target audience is already listening to...

Am I close?

PodPitch is THE tool that tons of businesses like Saywhat are using today to secure more podcast appearances and boost their visibility.

I know we can help you & Saywhat effortlessly secure the spotlight. The average PodPitch user books 7+ podcasts in their first 30 days.

Have 15 minutes available next week so I can show you a completely personalized demo? Just say the word and I'll start prepping it for you.

Parker New York, New York PodPitch

Again, clearly an AI written email. They sent it roughly 30 minutes after I was browsing around their site to pull their pricing information.

There is an ecosystem of companies like RB2B that use cross-site cookies (or first-party data reporting) to unmask your identity, even if you never gave them your email address. It's the same kind of vibe of showing you targeted ads once you've browsed some ecom site. But getting this email in such rapid succession to my browsing their site felt violating in a way. You're looking at what I'm doing online and I don't like that.

Where this goes

I mourn the death of the cold email. I've met some amazing, interesting people online by sending them an email out of the blue. They'll occasionally still respond when I do - but the rates are shrinking by the month. Which makes sense if I look at my own inbox. I'm buried in a tidal wave of these emails. And it's gotten hard enough to tell what's an AI written email that it's much easier to just ignore all of them.

Because again - I can tell what's AI written. But doing so takes way longer than seeing if something is spam. With spam you can do a quick classification of the subject line and contents. Princes in Nigeria or erroneous subscriptions to Norton Antivirus I can send to trash without even thinking about it. But there are enough neurons that you need to activate to sus out whether something pattern-matches as an LLM that you basically have to read the whole email anyway.

Services like Hey even emphasize that they won't show you any non-whitelisted email accounts by default. Senders that you don't know you won't see at all; that'll catch these emails but it also might catch a new friend. It reinforces the existing networks and connections between people. It relegates new connections into a Screener pile; if it's dominated by spammy outreach anyway, that just makes it a newage Spam folder. There was something so liberating about all communication landing in the same exact place. Whether you know them or not.

If you trace the game theory here, the only equilibrium is ignoring most emails. The price of reading through spam too great; the cost of a missed legitimate connection too low. But the idealist in me still wants to hold hope: what if we could keep email a safe space?

Also I'll just note. If you're a founder considering paying for automated podcast outreach: your name is on those emails. Every host who figures out what's happening - and more will - is associating your brand with spam. That's a trade worth thinking about before you hand your reputation to a bot that can send an email in 0.04 seconds.