The Doors That Disappeared

For most of its existence, the Google search engine was a directory. You typed in a search query, and it pointed you somewhere: ten blue links, ranked by relevance, each one a door to someone else's knowledge. You chose which door to open. You read what was behind it. You decided, consciously or not, whether the person who wrote it was worth trusting. That transaction is ending.

At Google I/O 2026, the company replaced its search bar with what it calls the "intelligent search box," a dynamically expanding, multimodal AI interface powered by Gemini 3.5 Flash that accepts text, images, files, video, and open Chrome tabs.^[5] AI Mode, the conversational engine behind the new box, is now available in nearly two hundred countries across ninety-eight languages with no subscription required. It has surpassed one billion monthly users. Queries are doubling every quarter.^[6]

The shift is not from bad search to good search. The engine that once helped you navigate to someone else's knowledge now removes attribution. It speaks that knowledge back to you in its own voice, with no indication that it belongs to anyone.

Liz Reid, Google's head of search, told NPR that the company still sends "billions of clicks" to publishers daily and that clicks from AI Overviews are "higher quality," producing longer user engagement.^[7] Helen Havlak, publisher of The Verge, said her internal data contradicts that claim directly.^[8] The numbers bear her out. Ninety-three percent of AI Mode queries end without a single click to an external website.^[9] Only eight percent of users go looking for the original source.^[10] Position-one organic click-through rates, the metric that determined whether a publisher lived or died, collapsed from 7.3 percent to 1.6 percent on queries where AI Overviews appear.^[11]

When the answer has no author, and the author can no longer afford to exist, what is the answer actually worth?

Pieter Claesz, "Vanitas Still Life" (1632). Mauritshuis, The Hague. An overturned glass, a guttering wick, a skull beside the books. Everything that once held value, emptied. Public domain.

The Source Goes Quiet

One travel blogger in Ontario is a casualty. The pattern is an industry.

Barbara Peng, the CEO of Business Insider, told her staff in May 2025 that the company needed to "endure extreme traffic drops outside of our control."^[12] She then laid off twenty-one percent of the workforce. Business Insider's organic search traffic had fallen fifty-five percent between April 2022 and April 2025.^[13] The company scrapped the majority of its e-commerce operation, the division most dependent on search referrals, and announced a pivot to live journalism events and enterprise AI tools. The Insider Union called the layoffs "a brazen pivot away from journalism toward greed."^[14]

HuffPost lost half its search referrals over the same period.^[15] The New York Times saw its search traffic share fall from forty-four percent to thirty-seven percent.^[16] The Mail Online watched its organic click-through rate collapse from thirteen percent on desktop and twenty percent on mobile to five and seven percent respectively.^[17]

These are the large publishers, the ones with subscriptions, events, and brand recognition to fall back on. The small ones have nothing. Morgan McBride ran Charleston Crafted, a DIY home improvement blog. Her traffic dropped over seventy percent after March 2024. Advertising revenue fell sixty-five percent, costing tens of thousands of dollars. "You can't just sit around waiting for things to turn around," she said.^[18] Scott Lapatine, the founder of Stereogum, a music blog that had operated for nearly two decades, watched seventy percent of his ad revenue disappear in a single year. He blamed Google's AI Overviews alongside Facebook and X's deprioritization of links.^[19]

Chartbeat data tells the structural story. Small publishers (those drawing between one thousand and ten thousand daily page views) experienced a sixty percent decline in search referral traffic over two years. Medium publishers saw forty-seven percent. Large publishers, twenty-two percent.^[20] The smaller you are, the harder you fall. And the smaller publishers are the ones who write about the things nobody else covers: the regional recipe, the niche hiking trail, the local history that no newsroom has a beat reporter for. When they disappear, the knowledge disappears with them.

Danielle Coffey, the leader of the News/Media Alliance representing over two thousand outlets, put it plainly:

"Google is using our content without compensation. It's parasitic, it's unsustainable, and it poses an existential threat."^[21]

The Spiral

There is a feedback loop at work, and it has no natural exit.

• Google's AI synthesizes information from across the web.
• The synthesis replaces the visit.
• The visit generated the revenue.
• Without the revenue, the publisher contracts or closes.
• The AI then trains on whatever remains: thinner, lower-quality, less original.
• The answers degrade.
• But the degradation is invisible, because ninety-three percent of users never click through to compare the answer against its source.

Matthew Prince, the co-founder and CEO of Cloudflare, one of the companies that routes much of the internet's traffic, framed the problem in terms even the AI companies should understand:

"If content creators can't get compensated, they'll stop creating content."^[22]

The data supports him. AI crawler-to-visit ratios have reached 250:1 and higher, compared to 2:1 for traditional search crawlers.^[23] Bots now account for fifty-one percent of all web traffic, the majority of it AI-related.^[24] The machines are reading the web more voraciously than ever. The humans those machines serve are visiting it less than ever. The economics of that equation do not balance.

Global publisher traffic from Google dropped by a third in 2025.^[25] News publishers expect search referrals to fall forty-three percent by 2029.^[26] Sixty-two percent of product discovery already happens on TikTok, Instagram, or YouTube rather than the open web.^[27] Mobile users spend eighty-eight percent of their time in apps, not browsers.^[28]

The web that Google was built to index is being hollowed out by Google. Fewer quality sources produce worse AI answers, which produce less reason to create quality content, which produces fewer quality sources. The information economy is eating its own seed corn.

Pieter Bruegel the Elder, "The Tower of Babel" (1563). Kunsthistorisches Museum, Vienna. One structure, built to reach heaven, consuming everything around it. Public domain.

The Monopoly That Writes the Answer

No company on earth mediates more questions than this one. With approximately ninety percent of the global search market,^[29] it faces no meaningful competitor in the link-based era it just ended. In August 2024, Judge Amit Mehta of the U.S. District Court for the District of Columbia ruled, in a 277-page opinion following a nine-week bench trial, that Google had violated the Sherman Antitrust Act by maintaining an illegal monopoly in internet search.^[30] It was the most significant antitrust ruling against a technology company since the Microsoft consent decree.

Thirteen months later, in September 2025, Mehta delivered his remedies decision. He rejected the Department of Justice's most aggressive proposals, declining to force the divestiture of Chrome or Android. Instead, he banned exclusive default-search contracts, ordered Google to share certain search index data with third parties, and established a six-year technological oversight committee.^[31] In his opinion, Mehta acknowledged that generative AI had "changed the course of this case" and wrote that competition in the AI space gave him reason for caution.^[32]

The caution may be misplaced. The remedies were designed for a link-based era. They address the question of who appears in search results. They do not address the question of what happens when there are no search results at all, only an answer that echoes in Google's voice using other people's work.

Penske Media Corporation, the parent company of Rolling Stone, Variety, Billboard, The Hollywood Reporter, and Deadline, filed a fifty-six-page antitrust memorandum in the D.C. District Court in February 2026 arguing that Google engages in "coercive reciprocal dealing": accept AI scraping or become invisible in search results.^[33] The European Union has designated Google a "gatekeeper" under the Digital Markets Act.^[34] Google's appeal and the DOJ's cross-appeal are both before the D.C. Circuit.^[35]

One company, already ruled an illegal monopoly, now speaks in a single voice to a billion people, using other people's work, without compensation, in a format designed to make the original source unnecessary.

What We Stop Doing

The deepest cost is not economic. It is cognitive. Betsy Sparrow, a psychologist at Columbia University, demonstrated in 2011 what she called the "Google Effect": when people know they can look something up online, they invest less effort in remembering it.^[36] The finding was unsettling but contained. The information was still out there, on the other end of a search. The human still had to find it, read it, evaluate it. The act of searching was itself a cognitive process, a series of decisions about what to trust and why.

AI Overviews remove the act of searching. AI Mode removes the act of finding. The user receives a pre-packaged synthesis without encountering the sources, the disagreements, or the context that produced it. Researchers writing in Big Think have warned that we are "offloading cognition to AI," causing critical thinking to atrophy in the same way a muscle weakens from disuse.^[37] A separate analysis in Psychology Today described AI-mediated knowledge as delivering "pre-packaged insights without asking for mental participation," producing passive consumption of machine-generated thought where active inquiry once existed.^[38]

The homogeneity problem compounds the cognitive one. When a billion people ask the same question and receive the same synthesized answer, the diversity of interpretation that drives understanding disappears. Learning does not happen when everyone agrees. It happens when someone encounters a framing they had not considered, a source that contradicts their assumption, an argument that forces revision. A single AI-generated answer, delivered with the confidence of a textbook and the accountability of no one, forecloses that encounter.

Unlike a human expert, AI does not justify its reasoning. It does not refine its position through experience. It does not take responsibility for errors. And its errors are significant: Google's own internal analysis of Gemini 3 found it produced incorrect information twenty-eight percent of the time.^[39] At five trillion annual searches, that translates to tens of millions of wrong answers every hour.^[40] A Mount Sinai study found AI chatbots "highly vulnerable" to spreading harmful health misinformation.^[41] Google's AI has claimed that Barack Obama was the first Muslim president, declared Hulk Hogan dead, and cited an April Fool's satire about microscopic bees powering computers as fact.^[42]

When the answer has no author, you cannot evaluate the author's credibility. When you never visit the source, you never encounter the context the source provided. When one voice speaks for everyone, disagreement, the engine of thought, disappears.

Every Voice Wants to Be the Only Voice

The pattern extends far beyond one company. OpenAI's ChatGPT now processes 2.5 billion prompts per day across 900 million weekly active users, capturing roughly seventeen percent of all global digital queries, the first double-digit crack in Google's near-monopoly in two decades.^[43] In 2026, OpenAI began running cost-per-click advertisements inside the chat interface itself.^[44] The answer is the ad unit. The user never leaves. Microsoft made Copilot the default search handler for the Windows 11 taskbar, Start menu, and File Explorer in a January 2026 update.^[45] Millions of searches that previously went through a browser are now handled by Copilot directly, pulling from Bing's index, presenting AI-generated summaries, often without the user realizing they have conducted a web search at all. The AI is not in the browser. It is the operating system.

Apple will unveil an AI answer engine called "World Knowledge Answers" at WWDC in June, transforming Siri from an assistant into an oracle that generates summaries blending text, images, video, and local results.^[46] iOS 27 replaces "Hey Siri" with a "Search or Ask" prompt.^[47] Apple partnered with Google to power it with a customized version of Gemini.^[48] Two competing walled gardens. The same underlying voice.

Perplexity, valued at twenty-one billion dollars, has been sued by the New York Times, the BBC, and Reddit for large-scale content reproduction.^[49] Cloudflare's research found Perplexity using undeclared "stealth" web crawlers to bypass web application firewalls.^[50] The company's entire value proposition is that you never need to follow the links it cites. It calls this aggregation. Publishers call it extraction. The better Perplexity works, the less traffic flows to the people who created what it summarizes.

Meta has embedded AI directly into Instagram and Facebook, answering questions inside the app in under five seconds.^[51] Instagram search now operates as a conversational AI engine: users discover content through AI-designed feeds, suggested creators, and natural-language queries rather than by browsing the open web.^[52] The user never leaves. The source never gets visited. The pattern is the same across every company. Ingest the web. Dissolve authorship. Speak in a single voice. Keep the user inside the walls. They are not competing to give you better search. They are competing to be your source of truth, the one voice you trust so completely that you stop asking where the answer came from.

Sixteen Years and Six Months

On Google's own publisher showcase page, there is still a case study for The Planet D. "Thanks to ads, we've grown exponentially," Bouskill is quoted as saying. "Before using our ad network, we made a solid six-figure income, but thanks to a mixture of ads, affiliates and partnerships, we were able to push that into seven figures."^[53]

The page does not mention that The Planet D no longer publishes. It does not mention the ninety-percent traffic collapse. It does not mention the layoffs, or the six months that undid sixteen years of work. Google's own case study for its own ad network now promotes a site that Google's own AI made economically unviable.

The couple is on YouTube now, creating video content for a platform owned by the same company that consumed their written work. They have rebuilt on Google's terms, inside Google's walls, subject to Google's next algorithmic decision. They are, in the language the industry uses without irony, pivoting.

Somewhere today, a person will type "best time to visit Kyoto" into Google's new intelligent search box. AI Mode will return a fluent, confident, three-paragraph answer. It will synthesize the accumulated knowledge of travel writers who spent years visiting Kyoto in every season, comparing temples in spring rain and autumn light, noting which ryokans are booked a year in advance and which shrines are empty at dawn. The answer will be adequate. It may even be correct. It will have no author. It will generate no visit. It will pay no one.

The person who knew the answer first is nowhere to be found.

Footnotes

Disclosure: Sage.is uses AI tools in its editorial and product workflows. This article was researched and drafted with AI assistance.