One billion. That is the number of health questions people bring to Google every single day, according to Hema Budaraju, who leads AI quality and experiences on Google Search. She delivered that figure on March 17, 2026, during The Check Up 2026, Google's annual health showcase streamed live from its facilities. It is not a projection or an estimate about a future state. It is the present baseline against which Google is now deploying a coordinated set of AI tools across search, wearables, clinical research, and genomic sequencing.

The number is striking on its own. But what makes it structurally significant is what has happened to the character of those queries. Search queries related to health have grown three times longer on average, according to Budaraju. People are no longer typing a keyword - "insomnia" - and waiting for a list of links. They are composing full questions, describing their situation, providing context, and expecting a meaningful response. A user who once searched "back pain" now writes out something closer to a detailed symptom description, specifying duration, location, and what has and has not helped. That behavioral shift is measurable, and it has consequences for how Google has configured its AI products.

The primary vehicle for handling that volume and complexity is AI Mode, Google's conversational search experience. AI Mode is available in more than 90 languages and across more than 200 countries, a rollout that has moved far beyond the English-language, US-only origins of Google's early AI search experiments. Users can now follow up on an AI Overview with a direct question, carrying the exchange into a full back-and-forth conversation without leaving the search interface. The Gemini 3 models power both AI Overviews and AI Mode, bringing reasoning, tool use, and multimodal capabilities to queries that earlier search architectures were not designed to handle.

Multimodal input matters especially in a health context. A user struggling to interpret a lab report - dense with red blood cell counts, hematocrit levels, and hemoglobin measurements - can upload it directly into AI Mode and receive an explanation, along with suggested questions to bring to a doctor at a subsequent appointment. Voice, camera, image uploads, and file attachments are all accepted inputs. The system can respond across the full breadth of what that billion-query-per-day figure contains: symptoms, nutrition, medication interactions, diagnostic results, care navigation, and follow-up after clinical visits.

Google says it aligns its models with clinical guidelines, stress-tests their reasoning and tone, and grounds responses in quality resources. Because health information varies by country, the company also works with national medical institutions: Apollo and Tata 1mg in India, and Medical Note in Japan are named as partners for locally relevant sourcing. The intent is to ensure that the volume of queries is met with locally calibrated, authoritative responses rather than generic answers.

What one billion queries per day means for publishers and marketers

That scale has a direct and contested relationship with the marketing and publishing ecosystem. PPC Land has documented how major health publishers experienced sharp visibility declines following Google's December 2025 core update, with established medical information websites losing positions they had held for years. Zero-click searches increased from 56% to nearly 69% since AI Overviews launched in May 2024. Health YMYL queries - the "Your Money, Your Life" category covering medical content - showed 52% average scroll depth in AI-generated summaries, meaning users consume substantial information before clicking through to an external source, if they click at all.

The tension is not simply about traffic volumes. Google currently blocks ads within AI summaries for health-related queries, classifying healthcare as a sensitive vertical alongside finance and gambling. That restriction means the billion daily health queries are not being monetized through placement within AI-generated answers, even as they generate significant user engagement. Advertisers in the health sector can still appear above or below AI Overviews in traditional positions, but the within-summary placement is categorically excluded. For the marketing community, this creates an asymmetry: AI search is reshaping how users encounter health information at enormous scale, but the advertising access points remain constrained.

Google Search VP Liz Reid, speaking in a March 6, 2026 podcast, argued that the overall volume of questions people ask is growing, not shrinking, and that AI Overviews were driving more than 10% additional queries globally during the third quarter of 2025. Her position - that a larger total query pool benefits the ecosystem despite per-query click-through declines - remains disputed by publishers experiencing ongoing traffic losses. Independent research published in February 2026 estimated Google processes approximately 13.7 billion daily searches in total, across all categories. Health queries represent a substantial subset of that, and the one billion figure gives a concrete measure of just how dominant a single content category has become within the world's largest search engine.

Gemini in caregiving and the Japanese documentation problem

Joelle Barral from Google DeepMind described a Gemini-powered Gem - a saved, specialized instruction set that customizes the Gemini app for a particular task - designed specifically for care workers in Japan. The demographic context is stark. Japan is projecting a shortage of nearly 600,000 health care workers by 2040. According to Barral, caregivers there report that documentation consumes up to 20% of their working hours, a burden that drives high staff turnover. The new Gem allows caregivers to upload voice memos, handwritten notes, or text, which Gemini then converts into standardized professional records. Records can also be translated for rotating caregivers for whom Japanese is not a native language. The feature is protected with enterprise-grade security, and data is never used to train Google's models, according to Barral. The Gem becomes available to Workspace customers in Japan later in March 2026.

Beyond caregiving documentation, Barral described the deployment of AMIE - Google's conversational medical AI built on Gemini models - into a real-world feasibility study with Beth Israel Deaconess Medical Center. That study, announced the week before the event, found AI can support clinical reasoning and history-taking with empathy, matching physicians in diagnostic quality. Patient trust in AI for clinical settings also improved after interactions with AMIE, according to results from the study. Google and Included Health are now preparing a nationwide, randomized study to evaluate conversational AI across diverse virtual care workflows - described as the first of its kind in scale and methodology.

Fitbit: medical records, sleep, and glucose

Rishi Chandra, who leads the Fitbit team, announced three upgrades to Fitbit's personal health coach, which entered public preview last autumn and is built with Gemini.

Sleep tracking is the first area. Google says it has improved sleep stage accuracy by 15%, trained on a diverse set of sleep profiles to ensure inclusivity. The improvement is rolling out to public preview users over the coming weeks. The system now captures sleep interruptions and restlessness during the night, as well as improved nap detection. A white paper on the methodology has been published.

The second development is metabolic science. A Fitbit study on predicting insulin resistance using wearable data was published this week in Nature. Starting next month, public preview users in the United States will be able to connect a continuous glucose monitor through Health Connect, enabling the coach to show how specific workouts or meals affect glucose levels.

The third, and structurally most significant, is medical records integration. Starting early next month, public preview users in the United States will be able to store and view their medical history in Fitbit and share it with the coach for personalized guidance. The initial flow requires users to find their doctor or health system, log into a provider portal, and verify identity - at which point historical and future records sync automatically. A more advanced flow, being developed in coordination with the Centers for Medicare and Medicaid Services and companies including b.well, CLEAR, and Athenahealth, uses CLEAR identity verification - requiring a selfie and a valid photo ID - to locate records automatically, without requiring manual portal logins. Chandra stated medical records are not used for advertising, a commitment documented in Fitbit's public privacy policies.

CMS and the interoperability push

Amy Gleason, appearing from the Centers for Medicare and Medicaid Services, gave a personal account of the fragmentation problem Google is attempting to address at scale. Her daughter Morgan has a rare disease and was misdiagnosed for over a year because her medical history was scattered across different clinics and systems - a gap the family bridged by carrying records and faxing documents themselves. Uploading those records to an AI tool later surfaced a misrecorded diagnosis from 16 years ago: what had been documented as ulcerative colitis was in fact microscopic lymphocytic colitis. That correction made Morgan eligible for a clinical trial.

CMS launched its Health Tech Ecosystem last July. When the initiative began, around 60 companies pledged to work toward interoperability. According to Gleason, that number has since grown to over 700 organizations. Google committed to two specific priorities: eliminating repetitive patient intake paperwork - what the initiative calls "kill the clipboard" - and developing conversational AI to help patients manage their health data between clinical visits. The scale of the interoperability effort reflects how far health data remains from the real-time exchange that now defines most other digital services.

CVS Health and Health100

Sonny Sgagliardich, executive director of strategic products with CVS Health's Health100 platform, described the structural challenge concisely. A patient's health information can sit across an insurer, a pharmacy benefit manager, multiple pharmacies, a primary care physician, a specialist, and emergency and urgent care visits - potentially ten separate entities, none of which share data efficiently. Health100 is designed to consolidate that landscape into a single experience. Built on Google Cloud infrastructure, it is described as an open platform - available to users beyond CVS Health - that will provide a longitudinal health record pulling together insurer data, pharmacy history, provider records, and wearables. An AI health partner, described as the primary interface of Health100, will monitor and anticipate rather than wait to be asked. In 2026, the plan is to ship the longitudinal record, care-finding and scheduling tools, and the AI partner simultaneously.

Rural health, clinical education, and the $10 million commitment

Dr. Michael Howell, Google's Chief Health Officer, joined two Arkansas-based health leaders to discuss collaboration on rural care and clinical education. Dr. Claude Pirtle, president of the Heartland Whole Health Institute, and Dr. Sharmila Makhija, founding Dean and CEO of the Alice Walton School of Medicine, described a context where chronic disease accounts for roughly 90% of Medicare expenditures and health care spending has reached 18% of US GDP. Arkansas ranks 46th in obesity and 44th in hypertension among US states, according to Pirtle. Self-insured employers working with the institute report 6% to 11% annual health care premium increases.

The Heartland institute, which moved into its current building in May 2025, is preparing to launch a chronic disease reversal project in the coming months, focusing on behavior change upstream of clinical intervention. The Alice Walton School of Medicine opened its doors last July, welcoming a first class of 48 students. A 110-acre health care campus in Bentonville, Arkansas, is in early design stages with the first building targeted for 2029.

Google.org committed $10 million today to fund organizations reimagining clinical education in the AI era, aimed at improving high-quality, person-centered care. The Council of Medical Specialty Societies and the American Academy of Nursing are the first two recipient organizations.

YouTube: 1 trillion views and an AI Ask button

Dr. Garth Graham, global head of YouTube Health and a cardiologist, anchored his presentation in scale. As of December 2024, YouTube crossed 1 trillion cumulative views on health videos globally. With nearly half of the world's health workers already using YouTube for clinical content - reviewing surgical techniques, studying for exams, learning new protocols - Graham argued the platform is positioned to address part of the global health workforce shortfall. The World Economic Forum projects a gap of around 10 million health care workers by 2030.

The new feature is an Ask button appearing on eligible health videos, giving users direct conversational access to an AI tool grounded in the specific video's content. A student can ask for a time-stamped explanation of a particular concept, request that terminology be translated into simpler language, or prompt the tool to generate multiple-choice quiz questions based on the video. The system is expanding through a partnership with the College of Nursing at the All India Institute of Medical Sciences in India for mandatory nursing courses, and with the International Federation of Red Cross and Red Crescent Societies for first-responder training starting in Latin America.

Dr. Latha, who leads the nursing program at AIIMS, described the impact in a direct quote included in the presentation: "What excites me most is the tool's power to convert passive viewing into active learning. In resource-variable settings, the ability to clear doubts in real time really builds the critical thinking and self-directed learning essential for competent nursing practice."

Google Research: from cells to the planet

Yossi Matias, VP and head of Google Research, outlined the foundational research behind the product announcements. MedGemma, Google's open medical model for text and image interpretation, was released in May last year and has since reached more than 3 million downloads, with thousands of applications built on it. AIIMS New Delhi is using it for outpatient dermatology screening trials.

On breast cancer, results from studies conducted with the UK's NHS and Imperial College London - published in Nature Cancer - found that AI can detect 25% of breast cancers missed by human specialists and could safely reduce 40% of the workload for screening specialists, with potential for faster reporting times as a downstream effect. A separate open-source tool, DeepSomatic, has been released to identify genetic variants in cancer cells. Cell2Sentence Scale is a 27 billion parameter open model, built with Yale University to understand single-cell behavior, which recently generated a new hypothesis about cancer cell behavior. Earth AI, a collection of geospatial models with an agentic layer, is being applied to shift public health from reactive to proactive - predicting outbreaks and identifying local vulnerabilities. Researchers from Harvard, Mount Sinai, and Boston Children's Hospital have published work using it, examining measles vaccination at zip-code level across the United States.

Roche and the $150 genome

Chris Bakan, head of computational science R&D at Roche Diagnostics, described a sequencing advance with direct implications for cancer research. Roche's SBX sequencing technology recently completed a whole-genome sequence in less than four hours - a record, according to Bakan. The first complete sequencing of the human genome took approximately 13 years and cost around $3 billion. With SBX, combined with Google TPU-based training stacks and AI methods including DeepVariant, the same process now costs under $150 and takes a few hours. Roche and Google Cloud have collaborated on genomics infrastructure for over a decade. Bakan described the partnership as building a world-class genomics research technology stack, aimed at eventually bringing advances from research labs into clinical use.

The competitive and regulatory backdrop

Google's event landed one week after Amazon expanded its Health AI agent from the One Medical app to Amazon.com and the Amazon Shopping app on March 10, 2026, as covered by PPC Land. Amazon's system is built on Amazon Bedrock as a multi-agent architecture and can book appointments, explain lab results in the context of an individual's history, and connect patients to licensed clinicians - directly competing for the same conversational health queries that now number in the billions daily on Google. Amazon's January 2026 launch of Health AI within One Medical was the first move in what is becoming a direct platform competition over health query volume.

The stakes extend beyond query share. A Stanford study published earlier this year found that AI medical models produce severe clinical errors in 22% of cases across evaluated systems, with errors of omission - failing to recommend critical tests or treatments - accounting for 76.6% of the most harmful mistakes. That finding applies to the category broadly, not specifically to Google, but it frames the scrutiny these deployments will continue to attract as they scale. Handling one billion health queries per day through AI systems is an achievement of engineering. Whether those responses consistently meet the standard that health queries demand is a question regulators, clinicians, and publishers are actively examining.

Timeline

  • 2016 - Google publishes foundational research on AI detection of diabetic retinopathy, the starting point of a decade of health AI work
  • May 2024 - Google launches AI Overviews in the United States; zero-click search rates begin rising; health YMYL queries face early impact
  • May 2025 - MedGemma released publicly; surpasses 3 million downloads within months
  • May 2025 - Heartland Whole Health Institute moves into its Bentonville building
  • July 2025 - Alice Walton School of Medicine opens; first class of 48 students enrolled; CMS launches Health Tech Ecosystem with approximately 60 founding organizations
  • July 2025 - Google reports AI Overviews driving more than 10% additional queries globally during Q2 earnings
  • December 2024 - YouTube crosses 1 trillion cumulative views on health videos globally
  • December 2025 - Google's December 2025 core update causes sharp visibility declines for health publishers
  • January 21, 2026 - Amazon launches Health AI in the One Medical app
  • January 27, 2026 - Google makes Gemini 3 the default model for AI Overviews globally; AI Mode transitions made seamless
  • This week (March 2026) - Fitbit's study on predicting insulin resistance using wearable data published in Nature
  • Last week (before March 17, 2026) - AMIE feasibility study results with Beth Israel Deaconess Medical Center announced; Google shares UK breast cancer AI research
  • March 10, 2026 - Amazon expands Health AI agent to Amazon.com and the Amazon Shopping app
  • March 17, 2026 - Google hosts The Check Up 2026: Fitbit medical records integration, 15% sleep accuracy improvement, AI Mode health upgrades, Google.org $10 million clinical education fund, AMIE nationwide study with Included Health, YouTube Ask button for health videos, DeepSomatic release, Cell2Sentence Scale 27B model, Earth AI public health deployment, Roche sub-four-hour whole-genome sequencing at under $150
  • Later March 2026 - Gemini caregiver documentation Gem becomes available for Workspace customers in Japan
  • Early April 2026 - Fitbit medical records integration launches for public preview users in the US; continuous glucose monitor connection via Health Connect becomes available
  • 2029 - First building on the 110-acre Bentonville Health Care Campus targeted to open

Summary

Who: Google, through Chief Health Officer Dr. Michael Howell, teams across Google Search, Google DeepMind, Google Research, Fitbit, YouTube, and Google Cloud, alongside partners including the US Centers for Medicare and Medicaid Services, CVS Health, Roche Diagnostics, Beth Israel Deaconess Medical Center, Included Health, the Alice Walton School of Medicine, and the Heartland Whole Health Institute.

What: A comprehensive set of AI health announcements anchored by the disclosure that Google receives over one billion health queries per day - queries that have grown three times longer on average as users shift from keywords to conversational questions. Specific announcements include AI Mode health search upgrades using Gemini 3, a Gemini Gem for caregiver documentation in Japan, Fitbit medical records integration, a 15% sleep stage accuracy improvement, glucose monitor connectivity via Health Connect, a completed AMIE feasibility study at Beth Israel, a nationwide conversational AI study launching with Included Health, a YouTube Ask button for health videos, the DeepSomatic open-source genomics tool, the 27B Cell2Sentence Scale model, Earth AI deployment for public health, Roche's SBX sequencing at under $150 per whole genome, and a $10 million Google.org commitment to clinical education.

When: The Check Up 2026 was streamed live on March 17, 2026. Individual product rollouts are staged across the following weeks - the Japanese caregiver Gem and Fitbit medical records both targeting before end of April 2026.

Where: The event was held as a live stream, available at health.google/the-check-up/. Deployments span primarily the United States for consumer-facing products, with AI Mode health features available globally across more than 200 countries and 90 languages, and specific international initiatives in Japan, India, and Latin America.

Why: One billion daily health queries represents both a product opportunity and a public health responsibility. Google positioned the announcements as the result of ten years of foundational AI health research reaching practical deployment simultaneously across consumer, clinical, and research contexts. The broader backdrop includes a projected global shortfall of 10 million health care workers by 2030, a Japan-specific projected shortage of 600,000 care workers by 2040, persistent fragmentation of patient health records, and chronic disease costs consuming 90% of US Medicare expenditures. Each announced product addresses a distinct point in the chain between a person asking a health question and receiving care.

Share this article
The link has been copied!