Research Finds Nearly 30% Of Google Ai-Cited Domains Are Absent From Traditional First-Page Results

WEST PALM BEACH, FL – A major new study of Google’s AI Overviews has uncovered something businesses and website owners cannot afford to ignore: appearing on the first page of traditional search results and being selected as a source for Google’s AI-generated answers are no longer necessarily the same achievement.

Researchers analyzed 55,393 Google searches across 19 subject categories and found that nearly 30% of the domains cited in AI Overviews did not appear anywhere on the conventional first page displayed for the same search.

The finding does not mean that traditional search engine optimization is obsolete. Google says its generative AI search features remain rooted in its established search ranking and quality systems. However, the research provides strong evidence that Google’s AI systems can draw from a broader collection of sources than the pages presented to users in conventional first-page results.

For businesses, this represents an expansion of the semantic-search principles Google has used for years. A website must still be crawlable, authoritative and competitive in conventional search, but Google’s AI can now investigate multiple related questions and assemble an answer from a broader collection of sources. Websites therefore need useful information addressing the complete range of questions and considerations surrounding a customer’s decision.

Nearly 30% of AI-Cited Domains Were Missing From Page One

The study, conducted by researchers at Washington University in St. Louis, examined searches performed over a 40-day period from March 13 through April 21, 2026.

Researchers collected trending search queries from the U.S. version of Google Trends and conducted the searches through fresh browser sessions originating from Northern Virginia. The searches covered 19 categories, including business and finance, health, law and government, technology, shopping, travel, automotive, science and entertainment.

Of the 55,393 searches studied, 7,583 generated an AI Overview. Researchers collected 61,212 reference URLs cited within those AI-generated responses and compared them with the traditional first-page results shown for the same queries.

Averaged across the 7,583 AI Overviews, only 25% of the AI-cited domains overlapped with the top five traditional results. The overlap increased to 41.4% when the top 10 results were considered and 70.2% across the complete first page.

That left 29.8% of AI-cited domains absent from the corresponding first page. At the individual reference level, 28.5% of the URLs came from hosts that were not represented anywhere on the first page for that search.

The researchers also found that these off-page sources did not score lower under the credibility measure used in the study. However, that measurement should not be interpreted as a universal quality evaluation of every cited commercial or local-business website.

The complete study, titled“Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact,” is currently an academic preprint. Its results should therefore be considered significant new evidence rather than a final, universal measurement of every type of Google search.

The study was also based on trending searches, not a dataset created specifically around local services or commercial buying decisions. Nevertheless, its central finding is important: an AI Overview is not simply a summary assembled from the traditional first-page links displayed for the same query.

Google’s Query Fan-Out Helps Explain the Difference

Google’s own documentation provides a likely explanation.

Google says AI Overviews and AI Mode may use a process called query fan-out. Instead of conducting only the exact search entered by the user, Google’s AI can generate and run several related searches covering different aspects of the original question.

Consider someone searching:

Who is the best property management company for an out-of-state landlord in Las Vegas?

Google does not disclose the exact related searches it would generate for a particular question. However, possible searches might address:

• Property management fees in Las Vegas
• Services for out-of-state property owners
• Tenant screening procedures
• Rent collection and owner disbursements
• Maintenance coordination
• Property manager reviews and reputation
• Experience managing single-family rental homes
• Frequency and quality of owner reporting

Google can retrieve information associated with those related searches and use it to support its response.

A company could therefore be cited because its website contains particularly useful information about tenant screening, remote-owner communication or maintenance procedures, even if that company does not occupy one of the highest conventional positions for the original broad search.

Google describes query fan-out as a set of concurrent, related searches used to gather additional information needed to address a user’s question. The company separately explains that this process can help its systems locate a wider and more diverse set of supporting web pages than a classic web search would ordinarily display.

Keyword rankings continue to matter, but they no longer tell the entire story.

Questions Are Far More Likely to Produce AI Overviews

The structure of a search also appears to make a substantial difference.

Across all 55,393 searches in the study, AI Overviews appeared 13.7% of the time. Among searches written as questions, however, the activation rate rose to 64.7%. Non-question searches generated AI Overviews only 9.5% of the time.

Longer searches were also more likely to trigger an AI response. AI Overviews appeared for 58.1% of searches containing six or more words, compared with 10% of one-word searches.

These results are especially relevant as people become more comfortable entering complete questions instead of short keyword phrases.

A search for “roofing company West Palm Beach” may continue to resemble a familiar local search. A question such as “Which roofing company in West Palm Beach has experience replacing tile roofs after storm damage?” gives Google considerably more information to interpret, investigate and answer.

Businesses should not respond by creating hundreds of thin pages targeting every imaginable question. Google specifically warns that creating separate content for every possible search variation primarily to influence rankings or AI responses may violate its scaled content abuse policy.

The better approach is to make a website genuinely informative about the questions customers consider before choosing a provider.

Google Says AI Search Still Depends on SEO

The emergence of AI-generated answers has led to a growing industry of services described as generative engine optimization, answer engine optimization, GEO or AEO.

Google’s position is more restrained. In its official guidance updated July 10, 2026, Google states that optimizing for its generative AI search features remains fundamentally search engine optimization because AI Overviews and AI Mode are rooted in Google’s search index, ranking systems and quality systems.

To be eligible for consideration as an AI supporting source, a page must be indexed and eligible to appear in Google Search with a snippet. Crawlability, internal linking, textual content, page experience, relevant images, accurate structured data and sound technical construction remain important.

Structured data can still help Google understand pages and make them eligible for certain rich results. However, Google says there is no special structured data required specifically for AI Overviews or AI Mode.

For local businesses, Google also recommends maintaining accurate information through Google Business Profile. Ecommerce companies should continue supplying complete and current product information through Google Merchant Center.

AI visibility is therefore not a replacement for SEO. It is an expanded result of doing SEO thoroughly.

Google Has Rejected Several Popular AI SEO “Hacks”

Google’s July 2026 guidance is unusually direct about several tactics being promoted as shortcuts to AI-search visibility.

According to Google:

• An llms.txt file does not improve visibility or rankings in Google Search. Google Search currently ignores it.
• There is no special schema markup required for AI Overviews or AI Mode.
• Content does not have to be divided into artificially small sections so AI can understand it.
• Companies do not need to rewrite their websites in a special style intended only for AI systems.
• Websites do not need to capture every long-tail keyword or create a separate page for every variation of a question.
• Seeking inauthentic brand mentions across the internet is not a dependable shortcut to AI visibility.

These points are addressed directly in Google’s section on generative AI search practices businesses can ignore.

The final point requires careful interpretation. Google is not saying that legitimate press coverage, independent reviews, industry citations, authentic customer discussions and authoritative references have no value.

Google’s guidance is narrower: its generative AI features can consider what is being said about products and services across blogs, videos and forums, but seeking inauthentic mentions is not as useful as it may appear because Google’s AI features rely on both its quality systems and spam protections.

Google does not provide an official checklist defining every placement it would consider inauthentic. Therefore, it would be an overstatement to claim that every paid article, press release or guest contribution is automatically disregarded. The defensible conclusion is that artificially increasing the number of brand mentions does not, by itself, establish the authority needed to appear in AI-generated search results.

Original, ‘Non-Commodity’ Information Is Becoming More Valuable

Perhaps the most useful part of Google’s guidance is its recommendation to create non-commodity content.

Commodity content is information that could have been written by almost anyone. Examples include generic articles such as “Five Benefits of Hiring a Property Manager” or “Seven Tips for Choosing a Personal Injury Lawyer.” Thousands of websites and AI tools can produce substantially similar versions of those subjects.

Non-commodity content provides information that comes from actual knowledge, experience, research or a recognizable point of view.

For a service business, this might include:

• Original data collected from completed projects
• Detailed explanations of how the company handles difficult situations
• Actual costs, timelines or performance ranges
• Case studies containing measurable outcomes
• Firsthand observations from experienced professionals
• Comparisons based on real customer circumstances
• Answers to questions that competing websites address only superficially
• Photographs, videos, charts or documents that substantiate the information

The objective is not to make content sound as though it was written for a machine. It is to provide the facts, evidence and expertise that people find useful and that an AI system may need when constructing a reliable answer.

This is where an experienced business may have an advantage over a larger competitor with a generic website. A smaller company that clearly documents what it knows may provide more useful supporting material for a specific question than a nationally recognized company offering only broad promotional language.

As AI systems become increasingly capable of reproducing information that is already widely available, original, non-commodity information is becoming more valuable.

Enjoyed This Content?

Like our coverage of search, SEO, AI, and digital publishing? Add SEARCHEN NETWORKS® as a Google Preferred Source to discover more of our articles in Google Search.

Add SEARCHEN NETWORKS®