AI-generated content is already appearing prominently in some of Google’s highest-stakes search results, according to new research from UK SEO and content agency Jodana. The study found that 12.1% of 670 top-ranking pages that could be scored were classified as AI-likely, with Legal and Insurance searches recording substantially higher rates than Health and Finance.
AI-Generated Content Is Reaching Google’s YMYL Results, New Study Finds
Google’s search results are becoming an increasingly important battleground for generative AI—but a new study suggests the issue is no longer limited to experimental chatbots or low-quality content farms.
AI-generated or AI-assisted material is already appearing among the top results for searches involving health, finance, legal rights and insurance, according to research published by UK SEO and content agency Jodana.
Those categories are particularly sensitive because Google classifies many of them as Your Money or Your Life (YMYL) topics, where inaccurate information can cause significant financial, health or personal harm.
Jodana examined the top-ranking page for 1,000 real Google search queries, split evenly across Health, Finance, Legal and Insurance. Researchers retrieved the actual content of the ranking pages and analyzed it using AI-detection software.
Of the 1,000 pages reviewed, 670 received a usable AI-detection score. Among those, 12.1% were classified as AI-likely.
The distribution was uneven.
Legal pages produced the highest rate, at 15.9%, followed closely by Insurance at 15.5%. Health came in at 8.2%, while Finance registered 8.9%.
The difference is notable because Legal and Insurance content was roughly twice as likely to be classified as AI-generated as Health and Finance content.
Why Legal and Insurance stand out
The research does not establish that AI-generated content is inherently inaccurate, nor does an AI detector prove that a page was entirely written by an AI system.
That distinction is important.
AI detection tools can produce false positives and may identify content as AI-generated even when humans have substantially written or edited it. The Jodana study therefore provides a signal about the prevalence of content classified as AI-likely, rather than definitive proof of how every page was produced.
Still, the numbers raise a more consequential question about editorial accountability.
Someone searching for an explanation of an insurance claim, legal right or contractual obligation may act directly on what they find in Google. If the underlying information is incomplete, outdated or wrong, the cost of that mistake can be considerably higher than getting a bad restaurant recommendation or an incorrect product description.
That makes the production process behind ranking content increasingly relevant.
Google’s quality systems face a new test
Google has spent years emphasizing experience, expertise, authoritativeness and trustworthiness—often summarized within the SEO industry as E-E-A-T—for content where accuracy matters.
The search giant has also repeatedly said that its systems are designed to reward helpful, reliable content rather than content simply produced to manipulate rankings.
Generative AI complicates that model.
A well-prompted large language model can produce fluent explanations of medical conditions, financial concepts, insurance policies or legal terminology in seconds. The resulting article may look authoritative even when it contains subtle factual errors or lacks the contextual nuance a qualified professional would provide.
That creates an unusual problem for search engines.
The challenge is not necessarily identifying whether a paragraph was generated by an AI model. It is determining whether the information is accurate, appropriately sourced, current and responsibly reviewed.
AI detection may not be the right answer
Jodana’s research was prompted partly by Anthropic’s announcement that Claude would watermark AI-generated text.
That development raises an obvious industry question: if AI-generated material becomes easier to identify, can search engines use that information to distinguish trustworthy content from synthetic content?
The answer is not straightforward.
A piece written entirely by a language model and carefully fact-checked by a qualified professional could be more useful than poorly researched human-written content.
Conversely, an article written by a human can still be inaccurate.
That means an AI-origin label alone cannot function as a quality score.
For Google, the more difficult problem is likely provenance and verification: who produced the information, what sources were used, whether an expert reviewed it, when it was last updated and whether claims can be independently verified.
Search behavior is already changing
The issue becomes even more complicated as Google’s own AI-powered search features increasingly answer questions directly.
Google’s AI Overviews can synthesize information from multiple sources and present an answer above traditional search results. That changes the role of publishers: instead of simply competing for a blue-link ranking, websites increasingly compete to become sources within an AI-generated answer.
Jodana’s broader report also examines Google’s AI Overviews, alongside the roles of Reddit and YouTube in YMYL search results.
Those platforms matter because users increasingly look beyond conventional publisher websites when researching sensitive questions.
Reddit can provide first-person experiences. YouTube can provide demonstrations and explanations. AI systems can synthesize information across sources.
Each format has advantages—and different forms of risk.
Enterprise publishers need stronger AI governance
For publishers, banks, insurers, healthcare companies and legal-information providers, the findings point toward a broader change in SEO strategy.
Simply publishing large volumes of AI-generated articles is becoming a weaker proposition, particularly in categories where Google’s quality systems and users demand evidence of expertise.
Enterprise content teams should instead be able to document who reviewed AI-assisted content, which sources support important claims, when information was last checked and what happens when regulations or medical guidance change.
That creates a new layer of content operations.
SEO teams, editors, subject-matter experts, compliance departments and AI governance teams increasingly need to work together.
The objective is not necessarily to eliminate generative AI from publishing. AI can help with research, summarization, drafting and content operations.
The objective is to make the human accountability layer visible and meaningful.
The next SEO battle may be transparency
Jodana’s research offers a snapshot rather than a definitive measurement of AI’s presence in Google’s YMYL results. Its use of AI-detection software also means the 12.1% figure should not be interpreted as proof that exactly 12.1% of the pages were actually generated by AI.
But the underlying question is harder to dismiss.
As generative AI makes high-quality-looking text cheap and abundant, search engines have to distinguish between content that sounds credible and information that deserves to be trusted.
For users researching symptoms, insurance coverage, legal rights or financial decisions, that distinction matters.
The next phase of search quality may therefore depend less on whether publishers disclose that AI was involved in writing an article and more on whether they can demonstrate that the information was checked, sourced and owned by someone accountable for getting it right.
Market Landscape
Generative AI is reshaping the economics of online publishing.
Tools from OpenAI, Anthropic and Google can dramatically reduce the cost and time required to produce written content. That creates obvious productivity opportunities for enterprise marketing and publishing teams.
The risk is greatest in YMYL categories, where content quality has consequences beyond search traffic.
Google’s AI Overviews add another layer because publishers now have to optimize not only for conventional rankings but also for visibility within AI-generated answers. Meanwhile, Reddit, YouTube and specialist communities give users alternative sources of information and first-hand experience.
The competitive advantage is consequently shifting from content volume to information credibility.
For enterprise SEO teams, that means structured editorial review, expert authorship, citations, transparent sourcing and regularly updated information are likely to become increasingly valuable alongside conventional technical SEO.
Top Insights
- Jodana found AI-likely content across YMYL search results, with 12.1% of scorable top-ranking pages classified as AI-likely across health, finance, legal and insurance queries.
- Legal and insurance results showed the highest rates, reaching 15.9% and 15.5%, raising questions about verification of high-stakes information.
- AI detection does not establish authorship or accuracy, meaning publishers and search engines need stronger approaches to provenance, expert review and content accountability.
- Google’s AI Overviews change the search landscape, requiring publishers to compete for inclusion in machine-generated answers as well as conventional organic rankings.
- Enterprise SEO teams face a governance challenge, balancing generative-AI productivity with expert review, reliable sourcing and transparent editorial accountability.
Get in touch with our fintech expert
