What 2026’s Biggest Studies About Trust in Public Health, Science, and Technology

At The Light Collective a lot of time goes into arguing about trust and representation: who has it, who lost it, who is trying to buy it back with a chatbot. This spring handed us something better than argument. Four large, credible polls landed between March and April 2026, each measuring a different corner of one question. How much do people trust the institutions, experts, and algorithms now shaping their health?

People reach for new health tools at a pace nobody forecast, while trusting those tools less than almost anything else in their lives. Below is what each poll found, and what it means for patients living inside that gap. For patients with lived experience, this is hopefully a useful roundup of all the latests stats and studies for your keynotes, talks, conference circuits.

Poll 1: Crisis of our own confidence

Confidence in personal health judgment is falling fast, faster than any single news cycle can explain on a global scale. The 2026 Edelman Trust Barometer Special Report: Trust and Health, now in a fifth year, surveyed more than 16,000 people across 16 countries. People’s confidence in their own ability to find reliable answers and make good decisions fell 10 percentage points in a single year, down to 51 percent, with statistically significant drops in 14 of 16 countries measured.

This number deserves that we pay attention to a growing gap, since the poll measures trust in health information and people’s confidence in their own decisions rather than trust in scientists as an institution, and those two things are not interchangeable. What Edelman really captured is the feeling of having concerning symptoms try to find trusted knowledge on the internet, and half the planet now reports some version of that experience.

Edelman also found that 70 percent of people believe at least one of six widely debunked health claims about food, vaccines, or medicine, and the striking part is that these beliefs cut evenly across education levels and do not stem from disliking doctors. Even credentialed people who trust their own physician still carry a debunked belief or two, which means confusion of what’s real and what’s not has long since stopped being something about “the uninformed public” and turned into the ordinary water that nearly everyone is swimming in. We can’t keep clutching our pearls about misinformation, because we’ve reached a crisis point in not knowing what to trust.

Poll 2: AI is everywhere in healthcare.

AI moved into health decisions on both sides of the exam table, and it happened in about two years. A West Health-Gallup survey fielded in late 2025 found that one in four U.S. adults, more than 66 million people, have used AI tools or chatbots for physical or mental health information. A KFF Tracking Poll on Health Information and Trust released in March 2026 put that figure at roughly a third of adults in the past year. Either way, asking a chatbot about your body stopped being an early-adopter habit. Tens of millions now get a second opinion this way…or a first one.

Patients are not making this shift alone. An American Medical Association 2026 physician survey found that 81 percent of doctors now use AI in their practices, up from 38 percent in 2023, so a minority curiosity became something four in five physicians touch. Both sides of that table changed, and changed in a hurry. This is well described in a recent piece called ‘Health systems govern the Tip of the Iceberg.” In a 2025 survey, Offcall reported that 67% of physicians reported using AI daily, 81% were dissatisfied with their employer’s AI adoption speed, and 71% reported having little to no influence on which tools their organization deploys. Figure Below:

Poll 3: We’re taking advice from AI as a source we openly distrust, and asking for more

High adoption met low trust, and that collision is the whole story these polls tell together. Among patients, KFF found that just 6 percent of adults who used AI for health advice say they trust it “a lot,” though about seven in ten express at least a fair amount of trust. West Health-Gallup data points the same direction from another angle: among recent AI health users, roughly a third trust the accuracy of what they get, a third distrust it, and a third land in the middle. Millions follow guidance they say, out loud, they doubt.

Doctors hedge right alongside their patients, because even at 81 percent adoption, AMA found 88 percent of physicians say AI tools still need formal validation for safety and efficacy before they will fully trust them, with 86 percent flagging data privacy as a top concern. People best equipped to judge these tools use them daily and keep asking for proof the tools actually work. Patients and physicians rarely agree on much, and here they agree on skepticism.

Poll 4: We’d never share our medical history, except for the part where we do

Privacy trust issues are running high and gets overridden constantly, which says more about the system than about the people inside it. A KFF poll found that about 77 percent of adults are concerned about the privacy of personal medical information shared with AI tools. Yet 41 percent of people who used AI for health reported uploading their own medical information to do it, which works out to roughly 13 percent of all adults.

A whole tension sits inside that one comparison. We want privacy, and people override it because a chatbot is fast, free, and awake at 2 a.m. when every clinic is closed. Make a safe channel slow and a convenient channel instant, and convenience wins even against a person’s own stated judgment, which means patients are not behaving foolishly so much as responding rationally to a system that built the careful choice into the hard path and left the risky one a single tap away.

This crisis of trust was a decade in the making.

We’ve been tackling trust issues for years in healthcare through our work with The Light Collective, and now have substantial independent data to support the need to lead rethink how we design and build tech. Trust in this arena did not collapse in a single data scandal or acquisition of internet companies. It thinned slowly, year after year like the cracked foundation of a building that was never up to code in the first place. While institutions kept assuming patients’ patience would last forever. Pew’s long series shows the building is ready to fall apart: confidence that scientists act in the public’s best interest stood at 87 percent in April 2020 and fell to 73 percent by late 2023, recovering only slightly since. Over a longer arc, the share of Americans saying science has a mostly positive effect on society dropped from 73 percent in 2019 to 57 percent. Each yearly dip looked survivable on its own. Stacked across a decade, they became a crisis.

This slow at the foundation of trust in our institutions and the internet has a root cause, and it is not simply that people grew irrational or too stupid to make informed decisions about their health. The public kept asking for honesty about data, plain answers, and a seat at the table, and too often got polished reassurance instead. Health systems sold patient records to data brokers. Consumer genetic testing companies changed terms after people had already spit in the tube, and they kept asking for more trust even after they got hacked and bankrupted. App makers buried consent in language nobody could parse. Misalignment with the public compounded quietly, one broken promise at a time, until 66 million people decided a chatbot was as good a bet as the system that kept letting them down.

There is a faint “let them eat cake” quality to how the trusted institutions met this moment. Many of the people charged with serving the public interest did not do enough to understand what patients were actually experiencing because they never partnered with patients in the first place, then expressed surprise when those patients walked toward whatever felt responsive at 2 a.m., as though surprise were any substitute for having paid attention. That walk toward AI reads as a verdict on a decade of not listening, delivered by people who finally stopped waiting.

Trust breaks faster than it heals, which is the hard math facing anyone who wants to fix this. Rebuilding will take longer than the breaking did, and it cannot be done with a friendlier interface or a fresh campaign. It calls for substantive change that visibly serves the public: real transparency about where data goes, real consequences when companies sell people out, and real representation for patients in how these tools get evaluated and bought. One finding offers a place to plant the first stake. Across every decline Edelman measured, “my doctor” stayed the most trusted source for honest health information. A human in a room, bound by a duty to look out for you, still holds trust these apps have not earned, and that points straight at what rebuilding requires.

Patients reading this can treat AI health tools the way you would treat a brilliant acquaintance who is usually right and occasionally, confidently wrong: fine for orientation and questions, no substitute for a person legally obligated to look out for you. Before pasting a medical history into a chatbot, ask where that information lands, because most tools will not volunteer an answer. For anyone working in health care or policy, the consensus in this data is the assignment, and the clock on it started years ago.