Picture this: it’s early morning, and you wake up with a strange pain in your chest. You don’t call a doctor. You don’t open your web browser and carefully choose among hospital websites, medical journals, or nonprofit health organizations. Instead, you type a question into a search engine. Within seconds, an AI-generated paragraph appears at the top of the page. It tells you what might be wrong, what to watch for, and whether you should see a doctor. You read it, feel a little better, and close the tab. You never click a single link. You may not even know which organization wrote the information that just shaped your health decision. This scenario is becoming the default way millions of people encounter health information online, and it marks a quiet but important shift in how trust is built, how knowledge is shared, and how the relationship between health organizations and the public is changing. Search engines used to be doorways. People would search, scan a list of results, and choose a source that seemed credible. The website’s name, its design, its tone, and its authors all contributed to whether a reader believed the information. Now, AI-generated summaries and health tools are collapsing that journey. The answer and the source are being pulled apart, and in many cases, the source is disappearing entirely.
This isn’t just a feeling or an anecdote; recent data makes the pattern clear. An analysis published last year by Pew Research Center tracked the actual web browsing behavior of a nationally representative panel of 900 U.S. adults over one month in 2025. The researchers looked at how people interacted with Google search results, especially when AI Overviews appeared at the top. What they found was striking. When an AI Overview was present, users rarely clicked on the links cited within the summary itself, doing so in only 1% of cases. Even more surprising, the presence of the AI summary seemed to change how people treated the rest of the results. When AI Overviews were on the page, users clicked through to the ordinary search results below just 8% of the time. When no AI Overview was present, they clicked through 15% of the time. In other words, the AI summary did not just serve as a helpful starting point; it often became the entire experience. For anyone who creates health content, these numbers are reason to pause. You can publish a thorough, accurate article about diabetes, insomnia, or a new treatment, and it may be used in an AI-generated answer. But the reader may never visit your page, never see your logo, never learn your mission, and never encounter the context and care that only the original source can provide. They walk away with the information but without the relationship.
The stakes are especially high in health. A KFF poll from March found that about two-thirds of adults, 68%, said they had sought physical or mental health information or advice from an internet search engine in the past year. Search is one of the first places people turn when they are worried, confused, or in pain. They search for symptoms, treatment options, medication side effects, and mental health resources. They search when they are alone at night, when they are caring for a sick child, or when they are trying to decide whether to call a doctor. In the past, a search engine offered a menu of sources: a government agency like the Centers for Disease Control and Prevention, a university hospital, a professional medical association, or a trusted nonprofit. The user could weigh those sources and choose. Now, with AI summaries becoming more common, people are increasingly receiving an answer that has been assembled from multiple sources, condensed into a few sentences, and delivered without the anchor of an identifiable author or institution. The information may be accurate, and it may even be helpful. But it enters a person’s mind in a different way. It feels like it came from the search engine itself, or from some anonymous, all-knowing intelligence. That can create a false sense of authority, or the opposite, a vague skepticism, because there is no way to tell who actually behind the words.
For health organizations, this shift changes the very meaning of publishing information. For decades, hospitals, public health agencies, and advocacy groups treated online content as both a public service and a way to build recognition. They wanted people to see themselves as reliable guides, to come back when they had more questions, and to understand the values and expertise behind their recommendations. A person who visits a hospital’s website and reads its health library not only gets answers; they also get a sense of the institution’s voice, its areas of specialty, and its commitment to patient education. They might sign up for a newsletter, fill out a contact form, or remember which organization helped them in a moment of worry. But when an AI summary delivers the health information, all of that context is stripped away. The reader receives the “what” without the “who.” This is not just a branding problem. It’s a trust problem. Trust in health information is built on accountability, transparency, and recognition. If people cannot tell where a piece of advice came from, they cannot evaluate why they should believe it. And if they never reach the source, they miss important warnings, nuance, and messages that might require a human touch. An AI answer can say “seek immediate care if you experience severe chest pain,” but it cannot look the reader in the eye and convey the urgency of an emergency physician who has seen too many people wait too long. It cannot adjust its tone to a frightened patient. It cannot invite a follow-up question or offer empathy.
In response, some organizations are beginning to adapt by learning how to be seen in AI systems. They are experimenting with techniques, often after the fact, to make their content more attractive to AI algorithms. This practice is called Generative Engine Optimization, or GEO. It is a close relative of search engine optimization, the old art of making websites rank higher in traditional search results. But instead of targeting lists of links, GEO targets AI-generated answers. Health care systems and pharmaceutical companies are among the early adopters, largely because they cannot afford to disappear. If a patient asks an AI tool about a common condition, the drug that treats it, or the risks of a procedure, the organization wants its expertise to be included in the AI’s final answer. To do that, they are restructuring their content in ways that AI systems are more likely to recognize and quote. That might mean writing clear question-and-answer sections, using consistent and simple language, embedding facts and statistics in explicit sentences, and adding structured data that tells software exactly what a page is about. Some studies suggest these efforts can work. One study found that GEO techniques increased a source’s visibility in AI responses by up to 40%, although the effects varied depending on the topic and type of content. For some publishers, that is a hopeful sign. It suggests that health organizations are not entirely at the mercy of the new systems; they can still shape, at least partly, how often and where their information appears.
But the deeper question is whether visibility in an AI summary is enough. Even if an organization’s content is quoted, readers may never click through, never learn the source, and never feel connected. That leaves health communicators with a difficult balancing act. They need to optimize their content for AI-powered tools today, but they also need to find ways to maintain direct relationships with the people they serve. AI tools and search engines must also take responsibility. They should make sources more visible, not less. Instead of burying citations in small text or clickable links that no one clicks, an AI answer could label its authority in plain language: “This information comes from the World Health Organization” or “According to Mayo Clinic.” People should be able to see the source as they read the answer, not just after digging through a footnote. At the same time, individuals can learn to ask simple questions when they encounter health information: Who wrote this? What is their evidence? How can I check the original? Health literacy has always been important, but it has become even more important in an age where an AI assistant can speak with calm confidence about anything. The challenge is to make good use of AI without letting it become an invisible veil between people and the institutions that work to keep them healthy. We are moving toward a world where information is faster, easier, and more convenient than ever. The goal now is to make sure it is still human and still trustworthy. That means keeping the source in sight, not just as a tiny link at the bottom of an answer, but as a meaningful part of the information itself. After all, people do not need only answers; they need to know whom to turn to when the answers are not enough.