OpenAI and Microsoft Knew They Were Starting a 'Doom Loop' for the Web

 

OpenAI and Microsoft Knew They Were Starting a 'Doom Loop' for the Web

## The Companies Knew They Were Driving Us Toward Google Zero — And Did It Anyway

---

### The Smoking Gun That Changes Everything

Let me tell you about a document that should be plastered on the front page of every newspaper in America. It's not a leaked memo from a rogue employee. It's not a conspiracy theory cooked up on a fringe message board. It's an internal presentation written by a senior Microsoft executive in January 2024 — and it describes, in the company's own words, a **"doom loop"** that would **"hurt the performance of our models and the entire web at the same time"** .

Read that again. The end-product — AI chatbots like ChatGPT and Copilot — threatens the economic foundations of its essential suppliers. And the company **knew it**.

The document was buried in a sealed court filing until September 17, 2026, when it was unsealed as part of The New York Times' three-year copyright infringement lawsuit against OpenAI and Microsoft. And what it reveals is not just damning. It's a confession.

Microsoft's Director of Applied Science, Brent Hecht, described the scraping of news content to train AI models as **"the largest theft of labor in human history"** and said it made a **"complete mockery of the idea of 'fair use'"** . OpenAI's head of ChatGPT, Nick Turley, wrote that publishers face an **"existential threat"** from chatbots that are **"largely substitutive"** and **"will get more and more substitutive as they get better"** .

They knew. They all knew.

And they did it anyway.

This isn't just a story about copyright law or corporate greed — although it's both of those things. It's a story about the future of the internet itself. It's a story about whether the web as we know it — a web of independent publishers, journalists, creators, and small businesses — can survive the AI revolution. It's a story about Google Zero.

Grab your coffee. This one's important.

---

## What the Court Documents Actually Reveal

### The "Doom Loop" in Microsoft's Own Words

The most explosive revelation in the unsealed filing is an internal Microsoft document authored by Hecht in January 2024, shortly after the New York Times sued the company. The document describes a feedback loop that would devastate the web:

> **"Our AI content strategy has started a 'doom loop' that will hurt the performance of our models and the entire web at the same time: It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its 'content supply chain.'"** 

Let's unpack that. AI models are trained on web content — news articles, blog posts, research papers, Wikipedia entries. That content is created by human beings who need to get paid. They get paid through advertising revenue, subscriptions, and traffic. Traffic comes from search engines, primarily Google.

But when AI chatbots answer questions directly, users don't need to click through to the source. They get the answer in the chat window. The publisher gets nothing. No ad impression. No subscription. No page view. No revenue.

And then the publisher can't afford to create the content that the AI was trained on in the first place.

That's the doom loop.

### The "Largest Theft of Labor"

Hecht didn't mince words in his internal communications. He called the mass scraping of news content **"the largest theft of labor in human history"** . He said Microsoft's legal defense — that the use was "fair use" — made a **"complete mockery"** of the concept .

He wasn't alone in his concerns. A Microsoft researcher told colleagues that compensating content creators was **"in the best interest of my employer, of my country, and of many other groups I belong to"** .

Microsoft, of course, is trying to distance itself from Hecht's comments. A company spokesperson said the remarks reflect **"one employee's individual perspective, are not a legal analysis, and do not represent the company's views"** . Another Microsoft executive characterized Hecht's role as **"adversarial"** and said he **"holds divergent, academic, and forward-looking views"** .

But here's the thing: the data backs Hecht up.

### The 93% Drop That Nobody Can Ignore

Microsoft's own internal data shows that its Copilot **"answer engine"** caused click-through rates for The New York Times' domain to drop as much as **93%** compared to traditional Bing search . For some publishers, the decline was as high as **94%** .

Read that number again. **93%.**

That's not a decline. That's a demolition.

When users search for news on Bing, they used to click through to the New York Times website. Now they get an AI-generated summary and never leave the search results page. The Times gets nothing. The journalist who wrote the story gets nothing. The advertising department that funds the newsroom gets nothing.

And this isn't just happening to the Times. It's happening across the entire web.

---

## Google Zero: The Term That Defines Our Moment

### What Is Google Zero?

The term **"Google Zero"** was coined by The Verge's Nilay Patel. He defined it as **"that moment when Google Search simply stops sending traffic outside of its search engine to third-party websites"** .

For two decades, the web was built on a simple bargain: you create content, Google indexes it, Google sends you traffic, you monetize that traffic through ads or subscriptions, and everyone wins. It wasn't a perfect system. Google took a massive cut. Publishers became dependent on Google's algorithm. But it worked.

Google Zero is the end of that bargain.

Google's AI Overviews — the AI-generated summaries that now appear at the top of search results — answer users' questions directly, without requiring them to click on any links. According to research published in June 2026, **68% of Google searches now end without a click**, with AI Overviews appearing in more than 20% of searches and reducing click-through rates by nearly 60% when present .

A separate study by Growth Memo found that users did not click through to any external websites in about **75% of the sessions analyzed** .

Google Zero isn't a future scenario. It's here. And it's devastating publishers.

### The Traffic Numbers Are Catastrophic

Let's look at the data.

Chartbeat data reported by Axios shows page views from Google Search fell **34%** across its publisher network between December 2024 and December 2025 . Smaller publishers — those with 10,000 daily pageviews or less — have lost roughly **60% of their search referral traffic over two years** .

The numbers are even more brutal for individual publications:

- **Politico** saw organic Google search traffic from U.S. users fall **23%** between June 2025 and June 2026

- **CNN** saw a decline of about **25%**

- **USA Today's national paper** saw traffic cut **nearly in half** 

The Verge itself — the publication that coined the term Google Zero — went from **5.3 million organic visits in February 2024 to just 790,000 visits in January 2026** . That's an **85% decline**.

And it's not just tech publishers. It's everyone. IAB Tech Lab estimates that AI-powered search summaries reduce publisher traffic by **20% to 60% on average**, translating to approximately **$2 billion in annual advertising revenue losses** across the publishing sector .

### The Human Cost

Behind every traffic chart is a human story. Phillip Swan is a journalist who covered the television industry for decades. He launched TV Answer Man in 2017 to help viewers understand emerging technologies like 4K and streaming. By September 2023, his site was getting about a million page views a month and generating more than **$20,000 a month** in advertising revenue. About **70% of his traffic came from Google**.

Then Google changed its algorithm. Then it introduced AI Overviews. And Swan watched his livelihood disappear.

**"It became obvious to me that my previous traffic numbers would never return unless Google decided to change its algorithm, which it did not,"** Swan said. **"In fact, it leaned in even more so by popping AI answers atop of site links. It was just a matter of time before I would have to stop publishing, which I basically did a month ago"** .

Today, traffic to TV Answer Man has dropped below **50,000 pageviews a month**. Swan, like many publishers, hasn't even been able to speak to someone at Google directly about the loss of his livelihood. **"Google helped him build his business, and now it has quietly destroyed it"** .

---

## The Legal Battle: Fair Use or Theft?

### The New York Times Lawsuit

The New York Times sued OpenAI and Microsoft in December 2023, alleging that ChatGPT and Copilot trained on **millions of pieces of Times content** and drew on that material to serve up answers to users' queries .

The Chicago Tribune, New York Daily News, and other large newspapers owned by Alden Global Capital sued both companies in April 2024. Those cases were later consolidated .

The tech companies have argued that their actions constituted **"fair use"** — a legal doctrine that allows for some use of copyrighted material without explicit permission. They claim their use of the material was **"transformative"** .

The Trump administration has sided with OpenAI on this issue, filing a brief in September 2026 defending the company's unlicensed use of copyrighted material to train its LLMs .

But the unsealed documents undermine the fair use defense in significant ways.

### The Paywall Problem

One of the most damning allegations in the filing is that OpenAI deployed tools specifically designed to **evade the paywalls** set up by publishers . The filing also alleges that the companies **"deliberately stripped copyright notices from training data"** and built training datasets through **mass scraping** .

Microsoft CEO Satya Nadella testified in a deposition that **"anything that is paywalled should be licensed by anyone who wants to use it…for grounding or training"** and said that if he had known OpenAI had scraped paywalled content, he would have **"invoked [Microsoft's right to] require OpenAI to retrain its models"** .

That testimony is significant. Nadella is essentially admitting that scraping paywalled content without a license is wrong — and that Microsoft would have stopped it if the company had known.

### The Existential Threat

OpenAI's own leadership acknowledged the threat their products pose to publishers. OpenAI's head of ChatGPT, Nick Turley, wrote in internal communications that publishers face an **"existential threat"** from products like ChatGPT, which are **"largely substitutive"** and **"will get more and more substitutive as they get better"** .

OpenAI President Greg Brockman described the models as **"excellent at news"** .

And Nadella agreed under oath that conversing with chatbots **"has substituted … giving you the information right there on the website on the AI platform versus needing to go to the underlying website"** .

They knew. They all knew.

---

## The Broader Implications: What Happens When the Web Dies?

### The Content Supply Chain

Here's the paradox at the heart of the AI revolution. AI models need content to train on. That content comes from the web. But the more AI models succeed, the more they destroy the economic foundations of the web that feeds them.

Hecht's "doom loop" isn't just a theoretical concern. It's a description of a system that is actively destroying itself. **"It is highly unusual that an end-product threatens the economic foundations of its essential suppliers,"** he wrote. **"But that is the situation we have created"** .

What happens when the content supply chain breaks? What happens when there are no more journalists to write the articles, no more bloggers to share their expertise, no more independent creators to provide the diversity of perspectives that make the web valuable?

AI models don't create new information. They remix existing information. If the well runs dry, the models get dumber. And so do we.

### The $2 Billion Question

The IAB Tech Lab estimates that AI-powered search summaries are already costing publishers approximately **$2 billion in annual advertising revenue**. That's money that would have funded newsrooms, paid journalists, and supported independent media.

Some publishers are trying to adapt. They're signing content-licensing agreements with AI firms. They're experimenting with new revenue models. They're investing in direct relationships with readers.

But the fundamental problem remains: when AI answers questions directly, users don't need to visit websites. And when users don't visit websites, websites can't survive.

### The Regulatory Response

Governments are starting to pay attention. The European Union fined Google **$1 billion** for violating the Digital Markets Act, and regulators are signaling that AI Overviews and AI Mode may face additional scrutiny .

Some publishers are even considering **blocking Google from crawling their sites altogether** — a drastic step that would cut off their largest source of traffic in a desperate bid to preserve their leverage .

The question is whether regulation can move fast enough to save the web before it's too late.

---

## Frequently Asked Questions (FAQs)

### Q1: What is a "doom loop"?

A "doom loop" is a feedback cycle where AI models scrape content from the web to train themselves, then answer users' questions directly without sending them to the source, depriving publishers of traffic and revenue, which forces them to cut back on content creation, which reduces the quality and quantity of content available for future AI training. It's a self-reinforcing cycle of destruction.

### Q2: What is "Google Zero"?

"Google Zero" is a term coined by The Verge's Nilay Patel to describe a future where Google Search stops sending traffic to third-party websites. It refers to the moment when search engines answer all queries directly through AI-generated summaries, eliminating the need for users to click on external links.

### Q3: Who knew about the doom loop?

According to unsealed court documents from the New York Times' copyright lawsuit against OpenAI and Microsoft, senior executives and employees at both companies were aware of the threat their products posed to publishers. Microsoft's Director of Applied Science, Brent Hecht, wrote an internal document describing the "doom loop." OpenAI's head of ChatGPT, Nick Turley, acknowledged an "existential threat" to publishers.

### Q4: What did Microsoft's executive say about AI scraping?

Brent Hecht described the mass scraping of news content to train AI models as "the largest theft of labor in human history" and said it made a "complete mockery of the idea of 'fair use.'" Microsoft has tried to distance itself from his comments, saying they reflect one employee's individual perspective.

### Q5: How much has traffic declined for publishers?

The declines are dramatic. Chartbeat data shows page views from Google Search fell 34% across its publisher network between December 2024 and December 2025. Smaller publishers have lost roughly 60% of their search referral traffic over two years. Politico's organic Google traffic fell 23%, CNN's fell 25%, and USA Today's national paper lost nearly half its traffic.

### Q6: What is the New York Times lawsuit about?

The New York Times sued OpenAI and Microsoft in December 2023, alleging that ChatGPT and Copilot trained on millions of pieces of Times content and used that material to serve up answers to users. The Times argues this constitutes copyright infringement. The tech companies argue their use was "fair use."

### Q7: What is "fair use"?

"Fair use" is a legal doctrine that allows for some use of copyrighted material without permission in certain cases, such as parody, news reporting, or criticism. The tech companies argue that training AI models on copyrighted content is "transformative" and therefore protected by fair use.

### Q8: What do the unsealed documents reveal about fair use?

The unsealed documents undermine the fair use defense. They show that OpenAI allegedly deployed tools to evade paywalls, stripped copyright notices from training data, and built datasets through mass scraping. Microsoft's own data shows Copilot reduced click-through rates for the New York Times' domain by as much as 93%.

### Q9: What happens if publishers can't survive?

If publishers can't generate revenue, they can't afford to create content. The quality and quantity of content on the web will decline. AI models, which depend on that content for training, will get dumber over time. The web as we know it — a diverse ecosystem of independent creators, journalists, and publishers — could collapse.

### Q10: Are publishers doing anything about it?

Some publishers are signing content-licensing agreements with AI firms. Others are experimenting with new revenue models, such as subscriptions and direct reader relationships. Some are considering blocking Google from crawling their sites entirely. The New York Times is suing OpenAI and Microsoft.

### Q11: What is the government doing about it?

The Trump administration has sided with OpenAI on the fair use issue, filing a brief defending the company's unlicensed use of copyrighted material. The European Union has fined Google $1 billion for violating the Digital Markets Act, and regulators are scrutinizing AI Overviews and AI Mode.

### Q12: What is the "content supply chain"?

The "content supply chain" refers to the ecosystem of content creators — journalists, bloggers, researchers, and others — whose work provides the training data for AI models. When publishers lose revenue, the supply chain breaks down.

### Q13: How much money have publishers lost?

IAB Tech Lab estimates that AI-powered search summaries are already costing publishers approximately $2 billion in annual advertising revenue losses.

### Q14: What can I do to help?

You can support independent publishers by subscribing to their publications, clicking through to their websites when you find something useful, and sharing their content. You can also stay informed about the AI threat to the web and demand accountability from tech companies.

---

## High-Value Keywords and Tags for AdSense Optimization

**Primary Keywords:**

- OpenAI Microsoft doom loop

- Google Zero publishers

- AI scraping theft of labor

- New York Times lawsuit OpenAI

- AI search traffic decline

**High-Value Financial Keywords:**

- OpenAI stock analysis

- Microsoft AI strategy

- Google search market share

- Best media stocks 2026

- AI regulation investment

**Long-Tail Keywords (Low Competition, High Intent):**

- What is the AI doom loop

- How AI is killing the web

- Google Zero explained for publishers

- New York Times vs OpenAI lawsuit explained

- Microsoft exec largest theft of labor

- How to save the open web from AI

- AI search impact on news publishers

**Tags:**

#OpenAI #Microsoft #GoogleZero #DoomLoop #AIScraping #NewYorkTimes #Copyright #FairUse #AIRevolution #TechNews #BusinessNews #MediaIndustry #Publishing #Journalism #OpenWeb #AIEthics #DataPrivacy #Antitrust #Regulation #Investing #StockMarket #MicrosoftStock #GoogleStock #AI Law #ContentCreators #DigitalMedia #NewsIndustry #ArtificialIntelligence #BigTech #AmericanConsumer

---

## Conclusion: A Confession in Their Own Words

The unsealed court documents in the New York Times' lawsuit against OpenAI and Microsoft are a confession. Not a confession extracted under duress or fabricated by accusers. A confession in the companies' own words, written by their own executives, describing a "doom loop" that they knew would damage the web — and choosing to proceed anyway.

Microsoft's Brent Hecht called it **"the largest theft of labor in human history."** OpenAI's Nick Turley called it an **"existential threat"** to publishers. Satya Nadella admitted under oath that chatbots have replaced the need to visit websites.

The result is a web that is dying.

The question now is what we do about it. The courts will decide whether the tech companies' actions constituted fair use or theft. Regulators will decide whether new laws are needed to protect publishers and preserve the open web. And consumers — all of us — will decide whether we're willing to pay for the content we consume or whether we'll continue to let the AI platforms eat for free.

Because the doom loop isn't just a problem for publishers. It's a problem for all of us. The web is a commons. It belongs to everyone. And it's being enclosed by a handful of trillion-dollar companies that have decided their business models matter more than the ecosystem they depend on.

The documents prove they knew. The data proves it's happening. And the future of the open web depends on what we do next.

---

## Disclaimer

This article is for informational and educational purposes only and does not constitute financial, investment, or legal advice. The views expressed are those of the author and do not necessarily reflect the official policy or position of any financial institution. Investing involves risk, including the possible loss of principal. Past performance does not guarantee future results. Readers should consult with a qualified financial advisor before making any investment decisions. The author is not responsible for any financial losses incurred as a result of actions taken based on the information provided in this article. All data and figures cited are sourced from publicly available reports and are subject to change. This article discusses legal proceedings and AI industry topics; readers should consult qualified professionals for specific guidance.