The Web Is Disappearing: What Google's Court Admission Reveals About AI Search

I recently declared that AI search represented an "evolution” rather than an “extinction." I included cheerful statistics about ChatGPT's growth, carefully balanced observations about SEO's continued relevance, and ended with the professionally optimistic assertion that "SEO isn't dying, it's expanding."

Then I read Google's latest court filing.

It was one of those moments where you realize you've been standing too close to the French Impressionist painting to see what it actually depicts. The perspective shift was immediate and uncomfortable.

Google—the very same Google whose executives have spent months insisting the web is thriving—admitted in a legal document that "the open web is already in rapid decline." They made this admission in a Virginia courtroom where they're fighting to keep their advertising monopoly intact. Their argument, essentially: Don't break us up because the thing we're accused of monopolizing is already dying anyway.

The cynical brilliance of this legal strategy shouldn't overshadow the admission itself. When the entity that essentially is the web for billions of users admits the web is declining rapidly, we should probably listen.

The Visible Wreckage

The numbers are brutal enough to make any publisher reach for something stronger than Olipop. At MailOnline, click-through rates crater by more than half when Google throws an AI Overview at the top of search results. Even when they're ranking number one, the holy grail of SEO, their CTR drops from 13% to less than 5% on desktop when a Google AI Overview is present. I've seen specific queries where traffic that once delivered 6,000 clicks collapsed to 100. 

Not a typo. One hundred.

The bloodbath extends across the industry. Premium publishers down 10% in eight weeks. Non-news brands down 14%. News outlets watching 16% of their traffic evaporate in a single week. These are cliff drops that leave chief digital officers staring at analytics dashboards like crime scene investigators.

Jon Slade, CEO of the Financial Times, described it as "sudden and sustained." The kind of language you use when something fundamental has broken and won't be coming back.

But here's what struck me: everyone's measuring the wrong thing.

The Dark Traffic Phenomenon

I've been watching something peculiar in client analytics over the past year. Direct traffic, the visits where users type your URL directly into their browser, has been exhibiting patterns that don't match any traditional model. Increases that correlate with nothing visible. Spikes in highly qualified visitors who somehow know exactly what they want. Conversion rates that make no sense given the lack of preceding touchpoints. 

We can attribute a meaningful share of this to a new subcategory of direct traffic: AI-direct. These are visits that appear “direct” in analytics, but the decision to visit was made via a conversation in an LLM. 

I watched my own behavior with ChatGPT last week. I spent forty minutes in conversation about project management software, exploring features, comparing options, discussing use cases. The AI mentioned several platforms, provided thoughtful analysis, helped me understand what I actually needed versus what I thought I wanted. 

When I finished, I didn't click a single link. I opened a new tab and went directly to the company’s site. That visit appeared in that company’s analytics as direct traffic. The forty minutes of AI-assisted research that led to a highly qualified, ready-to-purchase visitor? Invisible. Completely dark. 

This is happening millions of times daily, and nobody's measuring it.

Think about the implications. AI search visitors convert at 4.4 times the rate of traditional search visitors. Of course they do; they've already done extensive research. They've had their questions answered, their objections addressed, their alternatives evaluated. By the time they reach your site, they're browsing to confirm a decision that's already 80% made.

The traditional purchase funnel assumed multiple touchpoints we could track: awareness through display ads, consideration through search, evaluation through review sites, decision through direct response. AI collapses all of this into a single conversation that leaves no trace in your analytics.

Your attribution models are blind to what might be your most valuable traffic source.

The Disappearing Middle

What's vanishing goes beyond traffic metrics. An entire layer of the internet is evaporating: the aggregators, comparison sites, "best of" lists, review round-ups, all the content that exists primarily to synthesize information from other sources. AI does this synthesis instantly, conversationally, without the affiliate links and display ads that funded that middle layer.

I've watched this play out in real-time with software review sites. SaaS companies used to depend on placement in Capterra, G2, GetApp. Now users ask ChatGPT to compare options based on their specific needs. The AI searches and synthesizes current information from review sites in real-time, combines this with its foundational training, provides personalized recommendations, and sends users directly to vendors, bypassing the review sites entirely. 

The review platforms that spent years building their databases and comparison frameworks are being disintermediated by their own content, recycled through large language models. This resembles atmospheric evaporation more than traditional disruption, the middle layer of the web simply dissipating into AI training data.

Two Realities, One Timeline

We're experiencing a split-screen moment in internet history. On one half: users discovering a better way to find information, getting faster answers, having their first truly conversational interactions with technology. On the other half: publishers watching their traffic metrics nosedive, ad revenues collapse, and business models evaporate.

Both screens are showing the truth.

When Google says the web is thriving, they're looking at query volume and user satisfaction metrics. When they say it's in rapid decline, they're looking at publisher economics and the open ecosystem. These statements describe different aspects of the same transformation.

The user experience is evolving. The publisher experience faces extinction. And Google, the architect of both outcomes, profits either way.

The Feedback Loop of Death

Here's what really makes this a problem: AI systems need fresh, high-quality content to remain accurate and useful. They're trained on the corpus of human-generated information. But they're simultaneously destroying the economic incentives that create that information.

We're witnessing a digital ouroboros: AI eating its own tail. When publishers can't monetize content because AI summarizes it without attribution, they stop producing. When they stop producing, AI training data degrades. When training data degrades, AI outputs become less reliable. The snake consumes itself.

Google knows this. In their court filing, they essentially argue that breaking up their ad monopoly would accelerate publisher collapse. They might be right. They've created a system where they're simultaneously the disease and the only medicine keeping the patient alive.

The question nobody's asking is what happens when the patient dies anyway?

Strategic Implications for the Non-Delusional

If you've read this far, you're probably wondering what to actually do with this information. The answer depends on your position in the ecosystem.

For publishers: I wish I had better news. The traditional ad-supported content model probably won't recover. The pivot options are limited: subscription models (good luck with that saturation problem), becoming a direct brand (requires capital most don't have), or finding new platforms entirely. The cruel irony is that quality matters more than ever just as the economics of quality have collapsed.

For brands: This is your moment to get ahead. While others lament dark traffic, smart brands are mastering GEO (generative engine optimization), structuring content for AI extraction, building presence where AI looks, and ensuring consistent AI recommendations. The winners in this new era will be the brands AI systems cite (and mention) the most.

For marketers: Learn to optimize for invisible influence. Your content needs to be structured for AI extraction, not human scanning. Think in terms of discrete, quotable facts and clear value propositions that AI can confidently reference. Build presence where AI systems train: Wikipedia, Reddit, YouTube, and other authoritative sources. Accept that you'll never fully track attribution again. Let it go, man.

For everyone: Start measuring differently. Watch direct traffic patterns, branded search volume, conversation analytics if you have them. The metrics that mattered in the age of links—impressions, clicks, sessions—are becoming less relevant. What matters now is mention density in AI responses, brand recall strength, and conversion quality from dark traffic.

Asking Uncomfortable Questions

Writing this piece forced me to confront several questions I don't have clean answers for:

Is this transformation sustainable, or will AI essentially strip-mine the web until there's nothing left to train on?

When five companies control how information flows to billions of people, have we fundamentally broken something that can't be repaired?

What happens to the concept of authoritative sources when AI mixes information from everywhere into unmarked synthetic responses?

Are we watching the emergence of the informational Dark Ages, where the source material becomes invisible and only the synthesis remains?

I don't know. I don't think anyone knows. We're running this experiment in real-time on the entire world's information infrastructure.

Where This Leaves Us

Until recently, I was optimistic about adaptation. Now I'm realistic about transformation. The difference matters.

We're watching the controlled demolition of the open web and its replacement with something fundamentally different. The open web, where independent creators could publish, be discovered, and build sustainable businesses, is being replaced by a mediated experience where a handful of AI platforms synthesize and summarize everything, keeping users within their walls.

Google's court admission may have appeared to be verbal oopsie, but it wasn’t. Rather, it was an acknowledgment of what's already happened. The web's collapse is underway, visible in every publisher's analytics, every creator's revenue report, every marketer's attribution model.

But crisis creates opportunity. The harsh kind that rewards those who see reality clearly and adapt ruthlessly. The publishers pretending everything's fine will go under. The brands building for the old web will become invisible. The marketers optimizing for yesterday's metrics will wonder why nothing works anymore.

The ones who'll thrive are already accepting that the web we knew is gone and are building for what's actually emerging: a post-search world where discovery happens through conversation, where dark traffic carries your best customers, and where the entire middle layer of the internet gets compressed into AI responses.

When I wrote that "SEO isn't dying, it's expanding." I wasn't wrong, exactly. I was just looking at the wrong scale. The entire framework that made SEO necessary is disappearing. We're optimizing for synthesis now, not search.

The web has entered metamorphosis. Most of us are still treating it like a caterpillar when it's already dissolving into undifferentiated goo, hoping to emerge as something with wings. Whether it becomes an iridescent butterfly or a fat, brown moth remains to be seen. But it will definitely no longer crawl.