Zero Clicks #35: The Gateway to the Agentic Internet

Every two weeks in Zero Clicks, we explore the interplay of AI, media and commerce. Here’s your TL;DR for this week: 

• The big picture: How to monetize media in the era of AI agents

• Job posts: OpenAI breaks LinkedIn

• One great read: The peak AI anti-hype take from Apple

• AI tool of the week: The next era of vibe coding

Last night, I attended a fascinating dinner hosted by Cloudflare with leaders across media, tech, finance and crypto on emerging trends in agentic commerce. It was far and away the liveliest corporate dinner I’ve ever been to with hot takes flying across the table. One thing was crystal clear— nobody is quite sure how the next era of the web can, should or will be monetized. And look at that, we have a perfect segue to this week’s interview.

The gateway to the Agentic Internet
I’ve made my living on the internet for 15 years and this is the first time I genuinely feel like I’m living through the weeks when decades happen. Three years ago, the tech zeitgeist was dickbutts and the metaverse. Sam Altman restored tech’s mojo with a tweet. These are bolder and better days, but with unanswered existential questions for the information ecosystem. 

As AI upends the fundamental business model of media, few startups have quite as intriguing a role as TollBit, which is reimagining how content should be monetized on a web that may soon move beyond websites. 

We’re still in the infancy of figuring out how quality media can be sustainable in the age of AI. Last year at this time, Perplexity was scraping most of God’s green earth and promising a robust ad-sharing program from sponsored queries in search to sustain journalism. Now its ad chief has departed with negligible revenue to his name and the company has rejiggered its publisher program around rewarding media entities when their content is surfaced in Comet, the company’s AI-native browser. Meanwhile, the lawsuits and media layoffs continue.

TollBit, in contrast, offers a much more optimistic vision: one where media companies producing high-quality and content geared towards agents can be a lucrative and desirable enterprise. They are the most compelling answer I’ve found to my fear that LLMs are a Malthusian trap.

In lieu of my own essay, here’s a perspective that is bolder and far more horizon-expanding than my own. Trust me, TollBit Founder and COO Olivia Joslin does not disappoint.

MM: Your sign-in page bills TollBit as “Infrastructure for an internet without websites.” What does that mean? How does an internet without websites even function?

OJ: As humans increasingly get answers and complete tasks via agents and chatbots, there won’t be a need for humans to visit sites directly. In order for this to happen, of course, agents and bots will need to get more reliable at accessing and navigating sites, which is the area we focus on. After this is unlocked, websites as we know them will no longer be the terminal nodes of the Internet.

At that point, the vast majority of site visits will come from autonomous visitors rather than people. Bots are already surpassing humans in overall web traffic, and this imbalance will only accelerate as agentic AI becomes the primary interface to information online.

New infrastructure is needed to make sure a value exchange can continue to happen between these autonomous visitors and content or data owners.

MM: Ok so let’s think first principles about the media business model for a second. In the simplest sense, there have historically been basically two ways to monetize media:

1) Get readers to directly pay for content
2) Get advertisers to pay for attention and access to an audience

With agentic AI, a third way to monetize content creation is emerging.

Will this third monetization vector become bigger than the other two? If so, when?

OJ: We see agentic AI creating a new economic layer for content monetization that has the potential to rival, and eventually surpass, traditional models such as advertising and subscriptions. As users increasingly interact with content through AI interfaces such as ChatGPT, Perplexity, or other agentic assistants, monetization will naturally shift toward the “application layer,” where those interactions happen.

We believe that the third monetization vector is charging a toll to AI visitors that access and use your content. Instead of relying only on subscriptions or ads shown to humans, this model monetizes agentic traffic.

That’s what TollBit enables today. Publishers open a sanctioned front door for agents: bots are identified and routed away from the human site experience to our paywall for bots. Legitimate AI agents pay for authorized use and website owners are compensated using the terms they have set.

Things continue to evolve at lighting speed. This transition will likely accelerate over the next three to five years as AI interfaces become a primary gateway to information and services. While advertising and subscriptions will remain important, the third vector will grow fastest because it sits where the engagement is moving. Long term, it has the potential to become the dominant revenue stream for content because the autonomous visitors on the Internet surpass human traffic.

MM: Since the internet shattered publishers’ distribution monopoly, digital media has always been beholden to the whims of and form factors of the prevailing tech platforms, namely Google and Meta. Will it now just be OpenAI as the tail that wags the dog in terms of how media companies have to operate? Or will we see AI liberate media companies from such rigid form factors dictated by tech behemoths?

OJ: Search is no longer exclusively conducted via traditional search engines — it’s getting distributed across a massive longtail of specialized agents and AI applications. These apps & agents don’t search like humans; they don’t get bored after the third link. Instead they often look at dozens to provide an answer. They also enable us to get more done in less time. This shift means there will be more searches happening than ever before, as AI agents automate everything from research to procurement to customer service

If we democratize agentic website access by ensuring it’s scalable and open—not always free— but open, for any good acting developer, this will help ensure a long tail of agents and bots AI applications can flourish. The power and demand won’t just be from the main AI companies, but instead, this healthy layer of diverse apps built on top of core models. 

The challenge here is scale. It’s not realistic for publishers to manage 1:1 deals with the millions of agents that will need access to their sites, or vice versa. That’s why TollBit exists: to provide a unified gateway where AI agents can access and license content in real time, while ensuring rights holders are compensated.

Alongside this, new standards like NLWeb are emerging to make websites more agent-friendly. While that helps improve discoverability and structure, the real economic layer comes from ensuring those agent interactions are monetized — and that’s the problem TollBit solves.

MM: How do publishers and other consumer facing technology companies navigate the awkward teenage years of this new web, where the majority of commercial traffic is still human but agents are increasing in share every day? Said another way, what’s one tactical thing that publishers and marketers can do to prepare for an agentic future today that doesn’t disrupt the human user experience?

OJ: “Awkward teenage years” is a very apt characterization. AI agents are growing fast, and we see them as a new class of user — not just a temporary nuisance. The solution isn’t to force them to mimic human behavior - that just leads to more unauthorized scraping. Websites need to provide a clear, authorized pathway designed specifically for the needs of these new visitors. Websites need to create a sanctioned “front door” for agents that makes legitimate access easier and more attractive than scraping. 

We are in the 1995 internet era; new protocols are emerging, how agents interact with sites will continue to evolve. Website owners should lean into creating the norms for this new era of the internet. 

Creating this happy path for autonomous visitors is a big focus at TollBit. We give websites the infrastructure to distinguish humans from AI traffic in real time and route them differently: humans get the seamless experience they expect, while agents are directed to structured access points where usage can be managed, and monetized.

By utilizing this front door, publishers can regain control and set the terms for access. Agents get the access they need in a scalable way and publishers capture the value of that usage rather than losing it to unauthorized scraping.

MM: As I alluded to above, many publishers are cutting massive direct deals, most notably People Inc. pocketing $16M from OpenAI and the New York Times nabbing $25M from Amazon. One of the core theses of TollBit is that, while lucrative, directly negotiated deals between AI companies and publishers cannot scale at the rate of the agentic web. Can you elaborate on why a marketplace concept is ultimately the only sustainable concept for both sides?

OJ: There is no doubt that direct licensing deals between AI companies and individual publishers can be wins in the short term. They drive revenue, validate the value of quality content, and set precedent for compensation. But the agentic web can’t be built on a handful of bespoke contracts and one-to-one deals. We’re talking about a future where millions of AI-driven interactions happen every second across thousands of content sources, from global newsrooms to niche vertical publishers to independent creators.

Direct deals can’t match that scale because they are inherently slow, high-friction, and biased toward the biggest players with the most leverage. They leave many creators out of the value chain entirely, and they require constant renegotiation as usage patterns, capabilities, and content needs evolve.

TollBit solves these problems by creating a fair and standardized opportunity for any publisher to set their terms for access. It levels the playing field, creates more competition for their content, and ensures they’re compensated fairly. For AI companies, they can pay for content in real time and reduce legal risks. Simply put, this is the best scalable, transparent, and fair way to remunerate many sources at once.

MM: The cynic in me says that given the massive power imbalance between the leading answer engines and publishers, the AI companies will continue to just shatter norms and scrape content, unless blocked at the provider level by Cloudflare. Why, from the perspective of an AI company, is it sensible business to pay the toll?

OJ: We certainly encourage websites to block bots when it aligns with their strategy. You may have seen our partnerships with DataDome, HUMAN, and Fastly. Investment in cybersecurity helps drive up the cost of scraping and may help encourage AI companies to alternatively pay sites directly. But blocking isn’t always necessary. There are increasing reasons why AI companies are willing to pay for content access. Some of these include saving them on compute costs, eliminating litigation risk, improving their product experience via speed or accuracy, etc. We call this the carrot and stick approach.

MM: Speaking of Cloudflare, I have to imagine the immediate hours after Cloudflare announced it was blocking AI crawlers by default were intense. Walk us through the range of emotions and internal discussions after the Cloudflare news broke on July 1.

OJ: First of all, we are glad more folks are getting publishers thinking about blocking bots and acting on the increase in AI traffic. We’ve always been trying to tell publishers that robots.txt is not enough; you must actively try to forward bots off, and also meet them with a paywall. Since the beginning of the year, we’ve seen over a 20x increase in publishers blocking bots and redirecting them to the TollBit Bot Paywall. So overall, this is great for the industry and us. 

Collaboration with AI companies is important, though, just getting better at blocking won’t be the solution. The harder you block bots, the better they become at evading detection. This forces the cybersecurity tools to have to get more advanced and expensive, and this creates a negative feedback loop that threatens the industry. 

We’ve gone from 5M transactions a quarter in Q4 to over 30M in Q1, and we are actively supporting 1:1 deals and monetization. A key reason why we’ve been successful is this collaboration with AI companies & the fact that we’re agnostic to whichever CDN or cyber tool publishers use. A fraction of the 2000+ sites on TollBit use Cloudflare - there’s a lot of other CDNs out there. We think there needs to be a monetization solution that works easily across all publisher tech stacks. This is why we partnered for first party integrations with Fastly, DataDome, Human Security, with more on the way. 

MM: I’m often guilty of viewing the AI and publisher dynamic as a chessboard with just a few core models ((i.e. OpenAI, Anthropic) and application players (i.e. Perplexity) but there’s a much larger world of downstream AI applications that need to source content from publishers. What is an example of a couple of longer tail AI apps that are paying publishers in your marketplace and how do those deals effectively get structured?

OJ: You’re spot on about there being a longtail. There are numerous AI applications that are paying for content access to sites via TollBit, one example we’ve announced being Particle News. These AI applications often pay for content on a per-use or per-access basis, depending on their specific application and use case. 

We also support a number of the larger 1:1 deals on our platform that have a variety of structures to them. But most often we only see these deals occurring between the biggest publishers and biggest AI companies, versus this being a scalable option for the majority of websites and AI apps. 

MM: The old Hemingway adage of going broke “gradually, then suddenly” feels prudent in thinking about the web becoming an agentic medium. What do you think will be the catalyst that puts us on the “suddenly” end of the curve? Are we already there?

OJ: The industry is changing rapidly. The important thing to realize is that this is the first inning of AI. A lot of that disruption is happening despite only innovators and early adopters using AI. Now is the time to get the new economic models sorted out. We have yet to see the first killer, agentic consumer app come about. That mass market AI adoption will be an important milestone.

MM: Let’s end ultra-tactical. What’s one thing that any entity who creates content on the web– from a lowly newsletter purveyor like me to the largest media company in the world– should be doing today to maximize discoverability on the agentic web?

OJ: This is an interesting question. We believe that until there is a fair exchange of value where websites actually benefit from the agents and bots scraping their sites, websites should not be focusing on discoverability. Today, chatbots send back negligible human visitors to sites. For example, a large sports website was visited by AI bots 13 million times in a month, and this resulted in only 600 human visitors for the site. 

Websites would be better off focusing on optimizing their referrals from LinkedIn rather than fixating on AI chatbot optimization for now. Today, we encourage websites to set rates for AI apps to access their content to make it easy for AI developers to pay for access VS continue to scrape. 

Content Strategist, chatgpt.com, OpenAI
Feels impossible to start with anything other than what has to be the most viral job post in the history of LinkedIn. The jokes and memes basically write themselves here.

What most pundits missed is that this is basically a $400K/yr SEO gig. In totally unrelated news, Google is so dead that its stock just hit an all-time high at $2.8T market cap.

VP, Growth, Beehiiv
A year ago when I chose to write this newsletter on Beehiiv, I thought I was making an obvious bet on the ESP that would win the platform wars. But as it stands right now, Substack is the verb. Beehiiv is not. 

Ultimately though, I still think this is Beehiiv’s game to lose. The company still has the much better business model that is grounded in selling picks and shovels to a much wider swathe of media operators and not locking them into a new walled garden. Poach a few marquee Substackers, and the network effect starts to unravel.

To that end, Beehiiv is also bolstering its talent and creator partnerships team in addition to the VP of Growth opening– two of the best gigs in the media business right now. 

Sr. Director of AI Partnerships, Liveramp
Does this opening mean that one of the large foundation models is far closer to rolling out a full-fledged ad network than initially assumed and is now assembling the fit and finish pieces? Let a boy dream!

Head of AI Transformation, Fermat
Fermat is a great microcosm of what the transaction layer of agentic commerce will look like and often out ahead of the curve. They’re at it again here with a really, really slick product-marketing rebrand.

Marketing Technology Director, Liquid Death
If you’re going to play with B2B SaaS all day to increase shareholder value, you might as well do it for one of the most fun brands out there!

The Illusion of Thinking, Apple Machine Learning Research

Of all the AI anti-hype posts on the web, this is the one that I predict will have the most lasting, and potentially damning, impact. It’s not exactly a crisp page turner, but if you can handle a few gnarly graphs, it is the most important research to read this year.

Anything
I’m not sure it’s been adequately appreciated how insane it is that Lovable went from $0 to $100M in true ARR in just eight months. Of course, the risk with AI companies that don’t own the foundation model is that they’ll be the quickest to $100M and then to $0. Brace yourself– scores of vibe-coding interfaces are coming.

Anything is an interesting, under the radar player worth watching. There’s a subtle panache in their branding that I haven’t seen from any of the other players in the space. I have six weeks of upcoming paternity leave and intend to finally try my hand at being a vibe “brogrammer” and this is going to be my initial weapon of choice. If anyone else is playing around with it, drop me a reply and let’s jam a bit.