Notes From an AI-Written Webinar

Just today I sat through a marketing webinar with multiple sections. Free, 60 minutes, a topic and product I actually wanted to learn about. Fifteen minutes in, I noticed something:

Probably every section was AI-generated. The intros. The "personal stories" from the speakers. The case-study transitions. The product demo. I can see it by now (the same way I can see that I had AI help me polish the final version of this post). The pacing, the structure, the wording, the predictable beats — all GPT-shaped.

The event was well-made and the speakers themselves were real. But what they were doing was largely reading. They were teleprompter operators pretending to be presenters. And they weren't even hiding it that well.

I kept watching, because I wanted to see how it landed. It landed okay. People were in the chat, asking questions, buying the upsell at the end. The webinar worked – for me as well.

But the realization sat there: if AI wrote everything that mattered — the structure, the stories, the call-to-actions — then what were the speakers actually for?

The first wrong conclusion

The next step writes itself. If AI generates the content, why do you need a human to deliver it? Replace the speaker with a generated avatar. Now the avatar can do something the human can't: answer questions in real time. Tweak the pitch for each attendee. Run twenty webinars in parallel. Never get tired. Never charge $5,000 a day.

The technology is already adjacent. Synthesia. HeyGen. Tavus. Evergreen platforms with chatbot Q&A grafted on. The pieces exist. Within two years, everyone will be running their own digital twin webinars. Right?

That was my conclusion for about a day.

It is wrong.

What webinars are actually for

The flawed assumption is that webinars deliver information. They don't, not really. If information were the point, no one would attend live. People would read the PDF. Watch the recording at 2x. Skim a blog post.

People attend live webinars for reasons that have nothing to do with information:

• Parasocial trust. They want to see that this person — the actual human — is real enough to buy from. The trust transfer requires a perceived human on the other end.
• Social proof in the room. They watch the chat scroll. They see 800 other people. The cohort is part of the conversion mechanism.
• Scarcity pressure. "Offer expires when this call ends" only works because there is a perceived live human making a real offer that closes at a real time.
• The unscripted moment. The one moment where the speaker says something off-script — answers a question with actual experience, admits a failure, makes a joke that lands — is worth more than the rest of the hour combined. That's the moment people remember. That's the moment that converts.

An avatar can do none of these things. The moment an attendee figures out it isn't a real human — and they will — the entire trust transfer collapses. The "buy now" button starts to look like a pop-up ad, because that's effectively what it is.

The corporate webinar segment — feature launches, software demos, internal training — may well go avatar. Those events are already pretending to be human. Switching to an actual avatar is just more honest. But that is not where the money in webinars lives. The money is with coaches, consultants, experts, and B2B operators running high-ticket pitches. Those people are selling themselves. The avatar version cannot.

What does get eaten

So the human survives. But almost everything around the human does not.

• The script-writing? AI eats this. Including the personal anecdotes, if the operator is shameless enough to outsource those.
• The slide design? AI eats this.
• The follow-up email sequences? AI already eats this.
• The lead qualification? Eaten.
• Transcript handling, highlight clipping, social cuts, FAQ writing, recap pages? Eaten.
• Even the delivery of the script, in the corporate segment? Eaten.

The webinar of 2028 is, mechanically, a real human showing up for an hour with everything around that hour fully automated. Prep, audience warm-up, registration ops, no-show chasing, follow-up sequences, lead scoring, CRM sync, qualified-lead routing — none of it touched by the host.

That is not a sci-fi prediction. That is a roadmap.

The expensive human problem

Here is the part most product people miss.

If the human is the only irreplaceable piece, the human becomes more valuable. But also: more expensive, in the only currency that matters — finite attention. A coach can do four live webinars a month, maybe six. Each one is worth a lot. Each one is also a hard ceiling on the business.

This creates a specific kind of tool category. Not "tools that replace the human." Not even "tools that help the human work faster." Tools that maximize the value of every live appearance.

The metric isn't engagement, or attendance, or even conversion in isolation. The metric is: how much pipeline does one hour of this person's live time produce? And how do you raise that number without asking them to be live more often?

You raise it from both sides.

Before the live moment: the audience shows up warmer. They've already seen the offer. They know the host's positioning. They've self-qualified through the registration flow. The host walks into a room of 200 people who are 80% ready to buy.

After the live moment: the host walks away and nothing important is lost. Every attendee gets the recording with synced transcript. Every no-show gets a contextual follow-up that references what they missed. Every signal — duration, return-attendance, engagement — flows into the CRM. Qualified leads are flagged for direct outreach. The rest enter the long-tail nurture sequence.

The host's hour is now bracketed by infrastructure that did all the work, while the host did the only thing only they could do: be present, be real, and close.

Where ManyMeet sits in this

This is the bet I am making with ManyMeet.

Not "build another corporate webinar platform." There are enough of those, and they are built for broadcast — exactly the segment that loses to avatars first.

Not "build an AI presenter." That is a race I have no interest in winning.

The bet is: build the layer that makes a human-delivered Zoom webinar produce several times more pipeline per appearance, by automating everything that wraps around the live hour. Registration. Reminders that actually work. Recording playback with synced transcript. Funnel and retention analytics that show which signals predict revenue. Brevo sync with event-driven follow-up. An MCP server so AI assistants can act directly on the data. Every feature judged by one question: does this raise the value of the host's one live hour?

The hosts I am building for — coaches, consultants, B2B operators running webinars on regular Zoom Meetings — are the worst possible target for avatar replacement. Their entire business is trust transfer. Their entire revenue model rests on a buyer thinking "I can trust this human with my money." The more AI floods everything else, the rarer and more valuable that trust becomes.

The corporate webinar platforms can keep optimizing for scale. The avatar startups can keep raising rounds. The coaches who use ManyMeet will keep showing up live, and they will keep closing more from each appearance than they did the month before.

That is what survival of the webinar looks like. The format does not die. The format around the format does.

If you want to see what "everything around the live hour" looks like as actual data: here is an anonymized, real sample report I generated for a partner's webinar series — funnel, retention, lead-time, and the one reminder email that decides 84% vs. 13% attendance.

Now back to optimizing a live client's ManyMeet-based webinar funnel.