Google has been running a search experiment that most people missed — and its implications are staggering. The company appears to be testing how AI-generated content performs when injected directly into search results at scale, and early data suggests the results are cannibalizing traffic from human-created pages at an alarming rate.
The experiment, first reported by Search Engine Land, involves Google surfacing AI-generated pages — some with little to no original reporting, sourcing, or editorial oversight — in prominent search positions. These aren’t the AI Overviews that have drawn so much scrutiny over the past year. They’re something different. Full web pages, generated by AI systems, ranking alongside and often above content produced by newsrooms, independent publishers, and subject-matter experts who’ve spent years building authority.
This matters enormously.
For an industry already reeling from the traffic declines caused by Google’s AI Overviews and the Search Generative Experience rollout, the prospect of AI-generated pages claiming prime real estate in organic results represents a new and more direct threat. If Google’s algorithms can’t reliably distinguish — or worse, don’t prioritize distinguishing — between machine-generated filler and carefully reported content, the economic model underpinning digital publishing faces another body blow.
Google has long maintained that it evaluates content based on quality, not provenance. The company’s official guidance states that AI-generated content isn’t inherently penalized, provided it meets the standards of E-E-A-T: experience, expertise, authoritativeness, and trustworthiness. That framework sounds reasonable in theory. In practice, the experiment suggests the guardrails may not be working as advertised.
According to Search Engine Land, the AI-generated pages surfacing in this test often lack clear authorship, original data, or the kind of firsthand expertise that Google’s own quality rater guidelines say should matter. Some pages appear to be thin content dressed up with competent grammar and structured formatting — the exact type of output that large language models excel at producing. They read well enough to fool a casual reader. And apparently, well enough to fool Google’s ranking systems too.
The timing is pointed. Google has spent the last two years aggressively promoting its helpful content system, which was designed to reward pages created for people rather than search engines. The company rolled out multiple updates under this banner, each one supposedly refining its ability to identify and demote low-quality, mass-produced content. Several independent publishers reported devastating traffic losses after these updates, with some seeing 50% to 90% drops in organic visibility. Google told them the algorithm was working as intended — elevating quality, suppressing spam.
But now AI-generated pages are ranking in their place. The irony is sharp enough to cut.
Industry reaction has been swift and pointed. SEO professionals and publishers have taken to X (formerly Twitter) to flag examples of AI-generated pages outranking established sites. Some have documented cases where pages with obvious hallucinations — factual errors characteristic of AI generation — are sitting in top-three positions for competitive queries. Others have noted that the sites hosting these AI pages often have thin link profiles and minimal publishing history, attributes that would typically work against a site in Google’s ranking systems.
Glenn Gabe, a well-known SEO consultant who has tracked Google algorithm updates for years, has been among the most vocal analysts raising concerns about AI content proliferation in search results. His analyses of recent core updates have consistently highlighted patterns where AI-generated or AI-assisted content appears to be gaining ground against original reporting. The data points he and others have compiled paint a picture of a ranking system that is, at minimum, struggling to keep pace with the volume and sophistication of machine-generated pages flooding the web.
And the volume is staggering. Estimates from multiple research firms suggest that AI-generated content now accounts for a significant and rapidly growing share of new pages indexed by Google. Originality.ai, a company that builds AI detection tools, has published research indicating that a substantial percentage of top-ranking content in certain niches already shows signs of AI generation. The web is being reshaped faster than Google’s quality systems can adapt.
Google, for its part, hasn’t said much publicly about this specific experiment. The company’s communications tend to emphasize that its systems are designed to surface the most helpful results regardless of how content is created. A Google spokesperson told Search Engine Land that the company continuously tests improvements to search quality. Standard boilerplate. It reveals nothing about whether Google views the ranking of thin AI content as a bug to be fixed or a feature of a system that treats all content equally by design.
That ambiguity is the problem.
Publishers need clarity, and they need it now. The advertising revenue models that sustain most digital media depend on organic search traffic. When that traffic gets redirected — first to AI Overviews that answer queries without a click, and now potentially to AI-generated pages that replicate the substance of original reporting without the cost — the financial math breaks down completely. You can’t fund a newsroom on content that Google gives away for free or ranks below a machine’s approximation of your work.
The broader competitive dynamics deserve scrutiny too. If AI-generated content can rank effectively in Google search, the barrier to entry for creating a high-ranking website drops to nearly zero. Anyone with access to an API and a domain name can generate thousands of pages overnight. This isn’t hypothetical — it’s already happening. Entire networks of AI-generated sites have been documented, some targeting lucrative affiliate niches, others mimicking the format and tone of legitimate news outlets. Google’s spam team has taken action against some of these networks, but the game of whack-a-mole is one that favors the spammers when content generation is essentially free and instantaneous.
So where does this leave the search industry?
Some analysts argue that Google’s experiment is simply a reflection of reality. AI-generated content is here, it’s proliferating, and Google’s job is to rank the best result for any given query regardless of origin. If an AI-generated page genuinely provides a better answer than a human-written one, perhaps it deserves to rank higher. This is the techno-optimist view, and it has adherents in Silicon Valley.
But the counterargument is powerful. Quality in information isn’t just about surface-level readability or keyword coverage. It’s about accountability. It’s about having a reporter who picked up the phone, an expert who ran the experiment, an editor who checked the facts. AI-generated content can simulate these qualities convincingly without actually possessing them. And when ranking systems can’t tell the difference, the incentive to invest in real expertise evaporates.
There’s a historical parallel worth considering. In the early 2010s, Google faced a similar crisis with content farms — sites like Demand Media’s eHow that mass-produced cheap, formulaic articles optimized for search. Google responded with the Panda algorithm update in 2011, which specifically targeted thin, low-quality content and restored visibility to authoritative sources. It was one of the most consequential changes in Google’s history, and it worked — for a while.
AI-generated content is the content farm problem scaled by orders of magnitude. The pages are better written, more varied, and far cheaper to produce. If the Panda update was Google’s response to a pistol, it now faces a machine gun. Whether Google’s current systems are up to the task remains an open and urgent question.
Recent reporting from multiple outlets has highlighted growing tension between Google and the publishing industry over these issues. The News/Media Alliance, which represents thousands of publishers, has been increasingly vocal about the need for Google to protect original content in search results. European regulators have also signaled interest in how AI-generated content affects competition in search, adding a potential regulatory dimension to what has so far been treated as a purely algorithmic question.
For SEO professionals, the practical implications are immediate. If AI-generated content can rank effectively, then the traditional playbook of building authority through expertise, original research, and quality backlinks may be losing its edge. Some SEOs are already advising clients to focus more heavily on brand-building and direct audience relationships rather than depending on organic search — a tacit acknowledgment that Google’s results may become less reliable as a traffic source.
Others are doubling down on the signals that AI content can’t easily replicate: firsthand experience, proprietary data, original multimedia, and genuine community engagement. These are the qualities that Google’s E-E-A-T framework claims to reward. Whether the algorithm actually does so consistently is the trillion-dollar question.
The experiment Google is running isn’t just a technical test. It’s a stress test of the entire information economy. If the world’s dominant search engine can’t — or won’t — meaningfully distinguish between content created through genuine expertise and content generated by a statistical model predicting the next likely word, then the incentive structure that supports quality information production collapses. Not gradually. Rapidly.
Publishers, advertisers, and regulators should be paying very close attention. Because what Google decides to do with the results of this experiment will shape the economics of online information for years to come. And right now, the signals aren’t encouraging.
WebProNews is an iEntry Publication