As of yesterday, long live Google! They avoided the break-up, but must share data with rivals. I'm looking forward to writing about this over the weekend, stay tuned.
It doesn't change any of my thoughts below.
Agentic Search is the future
Matthew Prince, CEO Cloudflare in his recent interview spoke about the death of publishers. The race to clicks - the listicles, clickbait headlines, controversial takes (they're done for thank god).
AI answers sit above the fold (the very tippy top of Google search results), the cost of scraping subsidised by the people who make the work, and traffic is now a vanity metric.
To me, it's both great news, and bad news. We're starting to see new companies step in, to try take advantage of the opportunity.
What happens when AI has all scraped all the web, and no one is incentivised to create new content?
Introducing Comet. The new browser from Perplexity.
But wait there's more.
Aravind Srinivas, Perplexity's Founder, posted that Perplexity is launching Comet Plus to “reward publishers for quality content.”
That line stuck.
He made is sound like salvation. But. It also reads like the moment a new operator lays track across the open web.
Choo-Choo folks.
Here’s my take.
Comet is way more than just a nicer search box. It is a bid to own the on-ramp, the payout rules, and the crawl permissions in one stack.
It's giving Google 2.0 energy
I’m excited by the promise. I’m wary of the power. If this model wins, we do not simply fix referral traffic. We will be rewriting how knowledge is produced, priced, and policed.
A quick history lesson
You probably are aware, but incase you aren't, here's a small summary of what's happened so far:
AI answer surfaces are replacing clicks. The old trade was simple. Let the 'itsy-bitsy' crawler in (OpenAI, Anthropic, Google, Perplexity, etc), get a proportion of the attention back. Give an inch, take a mile. That contract has been torn up quietly, one summary at a time.
Cloud infrastructure grew teeth. Network providers now block or meter AI crawlers by default, and they’ve started talking openly about pay-per-crawl. Consent no longer lives in a polite text file. It lives at the edge. Cloudflare has been quietly taking advantage of this opportunity, blocking bots at the edge.
Perplexity launched a browser and layered a $5 monthly subscription on top that pays publishers from a shared pool using signals like citations, human visits, and agent actions. In other words, they aren’t just answering questions. They are starting to write the rules for who gets paid when the answer uses your work.
Basically, the interface has moved into the browser. Money begins to flow in a new direction.
Suddenly you can see the outline of a rail and the smell of burning coal.
Why the publisher model needed to change
The unit of value changed without the payment model following along.
A click-based world paid for promise.
An AI-answer world pays, if it pays at all, for delivery.
The content that ends up inside the model’s response is doing the heavy lifting. The rest might as well be wallpaper. Clicks are irrelevant.
If you make anything expensive, the way you get paid has to be specific, not vague. Publishers carry real costs. Reporting time. Legal review. Updating fast when facts change. That work shows up as freshness, accuracy, and context. The new model has to tie money to those properties, not to the illusion of “reach.”
Which is why Comet’s framing makes sense.
Pay for “filling knowledge gaps” instead of raw impressions. Reward sources when an answer actually uses their work and a human sees it. If done right, that finally values the thing that costs publishers the most: being right, and being first without being wrong.
Here's Comet’s play
Ok. Let's strip back the hype and buzz. Basically, there are three moving parts.
The on-ramp 'Comet': A browser that renders answers natively (Think of Chrome or Safari). If you own the front door, you decide what people see first. That is power number one.
The meter: Signals that claim to measure how much a publisher contributed to what the user consumed. Citations inside the answer. Human visits after the answer. Agent actions that trace back to a source. This is the accounting layer, where Perplexity will own the metrics that matter).
The pot. A subscription pool with a stated split for publishers. Pay from the pot according to the meter. This is the incentive layer, where Perplexity will decide who get's paid what.
Sounds very Google 2.0 right?
Put those together and you have a rail.
It starts at intent, runs through a proprietary surface, and ends in a private ledger that decides who deserves money this month.
Don't get me wrong, I like the direction.
Paying for usage is saner than paying for empty page views. But usage only becomes fair if the meter is auditable and the pot isn’t what Perplexity decides it should be.
Creator funds, payout schedules, being slaves to the algorithm. This has played out so many times.
Why this could be brilliant
Creators and publishers have been trapped between two bad choices.
Keep the paywall up and disappear from the AI-shaped internet.
Keep it down and watch your work summarised to death, with little return for you.
A licensed rail with money attached offers a third path.
Let the model read, get paid when the answer uses your content, and keep the right to say no.
Scarcity through choice.
A market for gaps also encourages depth.
Shallow rewrites don’t fill anything. How many times have you seen the 'best X for Y' products repeated over and over, which are all obviously paid.
Original reporting. Niche expertise. Local context. The open web desperately needs those things. If Comet pays properly, we get more of them.
And, we get no sponsored native advertising (I can't wait).
There is a second upside no one is talking about.
The model creates room for corrections and updates to be worth something.
If your source is used in an answer that later proves wrong, the marketplace can claw back. If you do the follow-up that fixes the claim, you get the tail.
Value. Scarcity. Truth. Plus, some potential innovation. See below as on September 4th. Agentic commerce. Cool!
Ultimately, when there's competition. The consumer wins.
Why this could be dangerous
One company setting the default surface and the payout rules is not a friendly neighbourhood start-up. No 'Good morning, Blake' here.
It is a rail operator. Rails collect tolls. Rails prefer their own freight. Rails do not become kinder with time.
Google literally dropped 'Don't be evil' from their mission.
If I could highlight that fact any bigger, I would.
If Comet wins the browser slot and the answer slot, the negotiating table shrinks to a single chair.
Consumers, don't win. Publishers don't win. Perplexity, does.
The risks are easy to name.
It can be self-dealing: When the platform both hosts the answer and decides which sources deserve money, it can quietly favour house content or friendly partners. Publishers learn to write for the meter rather than for readers. Happened today, will happen tomorrow.
Incentive clarity: Payout formulas drift over time. Small changes in weighting turn into big changes in income.
Lock-in: If Comet becomes the easiest licensed route to audiences, publishers face a hard choice. Accept the terms and depend on one rail, or retreat behind private APIs and pray your audience follows. Spoiler alert, most times, they don't.
That’s the Google 2.0 worry.
We replace the search monopoly of the last decade with a rail monopoly of the next.
Publishers, assemble!
For the first time in years, publishers have a lever that actually bites.
Scarcity. Access. Trains (choo-choo).
The website's themselves now block unlicensed crawlers at scale, at the edge. Cloudflare wins. Publishers win.
They can meter, price, and revoke.
Consent is not a polite suggestion on a robots.txt page. That gives creators bargaining power they’ve never had with search.
Matthew Prince, Cloudflare CEO literally said he could block 20% of internet traffic overnight if he wanted to. That's real power.
If Comet wants premium sources, it needs licensed access. If licensed access costs real money, the payout pot must be large enough to attract them. If the pot is large, the platform needs more paying users and stronger default distribution. You can see the loop forming. Rails on one side. Pipes on the other. Publishers in the middle with something to sell again.
I think, that standoff is healthy. It stops any one party pretending the others don’t exist.
Only thing is... The more rails that are laid, the less this matters for the trains.
I can see only one option.
Diligence and ruthlessness in negotiation. Publishers, never give an inch.
What I see a fair rail promising
If Perplexity wants the trust of people who make things, it needs to behave like a utility before it is treated like one.
Transparent formulas: Publish the payout algorithm, the weights, and the appeals process. Let independent auditors verify the statements.
Receipt-level accounting: Every payment should carry a record of which answer used which chunk of your work and when. Real Receipts.
No house edge creep: If Perplexity publishes or commissions content, label it, treat it like third-party supply, and disclose any advantage it receives. These shouldn't be some little 'sponsored' label out of sight. Really big. Always.
Portability. Creators should be able to export their licences, usage logs, and history to negotiate with rival rails. Competition is healthy. Let's encourage it.
Do I expect a start-up to love those ideas?
Not really.
Do I think the market will demand them if this rail becomes key infrastructure?
Absolutely.
So what to do now, if you're a publisher
This is not the moment to sit back and wait for a nicer check. It is the moment to make your work legible to machines, enforceable at the edge, and valuable in the answer layer.
Let's get technical:
Turn on the gate: If your CDN offers AI-crawler controls, use them. Negotiate from consent, not from fear of invisibility.
Sign your pages: Attach cryptographic fingerprints or watermarks that survive basic copying. Provenance will become a ranking and payment factor.
Write for receipts, not clicks: Make your claims specific, citable, and updatable. Build “why this changed” posts that models can link to when facts move
Call it Rail SEO if you like. I think it's catchier than Generative SEO anyway.
It looks nothing like the old thing.
It is closer to running an wikipedia like service than crafting clickbait headlines for robots.
My stance
I think Comet’s direction is necessary.
The open web cannot survive on empty promises of “exposure” while models eat the value upstream.
Paying for usage, not clicks, is the right reset.
The browser move is clever.
The publisher pool is a start.
The positioning around knowledge gaps is honest about where models fail and where humans shine.
We have lived through one cycle of monopolistic distribution destroying the open web. No one is keen to repeat it with nicer branding and a slightly friendlier share. Well, except maybe Google.
So I’m treating Comet like a promising idea in a trial period.
I’m hopeful because the incentives line up for the first time in a decade. I’m cautious because incentives drift over time. Creators and publishers will do their part. Make fresh, authoritarian content. Update it when it changes.
Let's hope Comet Plus keeps up their end of the bargain.
And if they get their hands on Google Chrome, we'll see what happens when they get power levels over 9000!
Update and spoiler alert. Not this time, Perplexity.
Managed to get a dragon ball z reference in there. How good.
Until next time.
Blake