If your model of how patients and families get health information online treats search engine results pages as a list of curated links (where users click through to authoritative sources such as NHS, Mayo Clinic, peer-reviewed journals, and professional medical societies to read content), the 2025 data on Google AI Overviews shows the model is outdated. AI-generated summary responses now appear at the top of 84–92% of health-related Google searches. Symptom queries trigger an AI Overview 92% of the time. Common health questions trigger one 92% of the time. Treatment queries trigger one 90% of the time. Condition queries trigger one 84–88% of the time. AI Overviews are now a routine feature of health information searches, shaping the initial interpretation of what patients understand about their symptoms, conditions, and treatments, before they click through to any source link.

The structural shift this represents is hard to overstate. For a generation of patients and families, "looking it up online" meant typing a query into Google, scanning the top links, and clicking through to authoritative-looking sources. The information layer was edited content from organisations that had editorial processes, fact-checking standards, and accountability for accuracy. The AI Overview layer is generated content produced by Google's models based on indexed web content, presented as a synthesised answer at the top of the search results page.

The data observation: for most health information queries, the patient now reads a Google-generated summary before (and often instead of) reading authoritative source content. The summary may cite sources; the patient may or may not click through; the summary itself becomes the de facto initial information.

The query type breakdown is informative for which kinds of health information are most affected:

  • Common health questions ("does pus mean it is infected?"): 92% trigger AI Overview
  • Symptom queries ("dementia symptoms"): 92%
  • Treatment queries ("dementia treatment"): 90%
  • Condition question queries ("what is dementia?"): 88%
  • Condition queries ("dementia"): 84%

The pattern: more specific, action-oriented queries (symptoms, treatments, common questions) trigger AI Overviews more reliably than generic terms (condition names alone). The reason is straightforward: Google's AI Overview system is designed to be useful for specific question-shaped queries, where a synthesised answer adds clear value. The result is that the queries most likely to influence patient action (symptom interpretation, treatment understanding, common-question framings) are also the queries most likely to be shaped by AI Overviews.

Three structural data observations follow.

The first observation: the AI Overview layer has effectively become the default health information layer for online queries. Patients searching for symptom information, treatment information, or common health questions encounter AI-generated synthesis as their first information layer in 84–92% of cases. The traditional list-of-links interface, with click-through to authoritative sources, is no longer the primary user experience for health information search.

The second observation: the AI Overview content quality determines patient information quality at scale. Hundreds of millions of US adults search Google for health information annually. If AI Overviews provide accurate, well-sourced, balanced summaries, patient information quality improves at scale. If AI Overviews provide subtly incorrect, biased, or low-quality summaries, patient information quality degrades at scale. The content quality of one product (Google AI Overviews) now substantially shapes US patient information access.

The third observation: the source citation behaviour of AI Overviews affects the downstream visibility of authoritative health information sources. When AI Overviews answer the query at the top of the page, click-through rates to source content decline. Authoritative health information producers (Mayo Clinic, Cleveland Clinic, NIH, medical specialty societies) increasingly produce content that informs AI Overview summaries but that fewer users read directly. The economic and editorial incentives for authoritative content production shift accordingly.

The data observation: the 84–92% AI Overview prevalence in health searches is a structural change in the patient information landscape. It happened quickly (Google AI Overviews expanded broadly in 2024-2025) and the implications are substantial.

Three structural implications follow for organisations whose operations intersect with patient health information.

The first implication: healthcare communication strategy needs to engage with AI Overview content as a primary distribution channel. Health systems, professional medical societies, and health information producers need to produce content that AI Overviews can effectively synthesise: clear, well-sourced, structured for AI ingestion. The traditional model of producing content for human readers who navigate to the source remains important but is no longer sufficient. The new model includes producing content that informs AI Overview synthesis.

The second implication: patient education programmes need to engage with what patients are likely seeing in AI Overviews. Pre-clinical visit patient education traditionally assumed patients arrived with whatever they remembered from authoritative sources they had read. The 2025 reality: patients arrive with whatever AI Overviews told them. Clinical education programmes that engage with this reality (discussion of common AI Overview content, common AI Overview gaps, and how patients can verify AI-generated health information) will be more effective.

The third implication: the patient consent and disclosure landscape will need to engage with AI-generated health information. If patients are forming health beliefs based on AI Overviews, the question of who is accountable for accuracy, what disclosure is appropriate, and what regulatory framework applies becomes operationally relevant. The current legal and regulatory infrastructure treats Google as a search engine rather than as a health information producer; the 84–92% AI Overview prevalence suggests the framing may need to evolve.

The data observation about who is shaping this: Google is the primary AI Overview provider for US search. The corporate decisions Google makes about AI Overview content quality, source citation, accuracy review, and topic-specific behaviour shape patient information at population scale. The accountability for this shaping is concentrated in one organisation. The implications for public health information governance are substantial and largely undeveloped.

For health systems, professional medical societies, and patient advocacy organisations setting 2026 communication strategy, the data observation calibrates expectations: the patient information landscape has changed structurally. Strategies that engage with the AI Overview layer as a primary distribution channel (both producing content that informs AI Overviews and helping patients navigate AI-generated health information critically) will be aligned with the actual patient experience. Strategies that continue to operate as if patients read authoritative content directly will be misaligned with where patients are actually getting health information.

The data is direct: 84–92% of health searches now return AI-generated summaries. The implications for patient information, healthcare communication, regulatory frameworks, and authoritative source visibility are operationally substantial. The 2026 strategic landscape for health information producers is shaped by this reality.