Updated October 26, 2026

Illustration: Quartz
AI already reads X $TWTR-rays and takes notes while patients talk with their doctors. Many patients cannot say whether the technology has played any part in their own care. That gap between what the tools do and what patients know is at the center of a 2026 survey from Pew Research Center.
Pew published "Americans want transparency when AI is used in their healthcare" on August 25, 2026. Eileen Yam, Giancarlo Pasquini and Emma Kikuchi wrote the short read. It draws on a survey of 3,488 U.S. adults conducted June 22 to 28, 2026, through the Center's American Trends Panel. Pew says the results represent the full U.S. adult population.
The survey poses a practical question. When a health care provider uses an AI tool, should the patient be told? That question sits at the core of AI in healthcare disclosure. Pew notes that debate over the legal and ethical duty of medical professionals to get patient consent before using AI tools is moving quickly. The survey does not take a side in that debate. It records what patients say they expect.
The responses are not uniform across every use. Pew asked about tools that can shape medical decisions, such as reading scans or explaining lab results. It also asked about background work, such as scheduling appointments and ordering prescription refills. Patients draw distinctions between those categories, and the data show how wide those distinctions are.
Pew also asked how much control patients feel they have over AI in their care, and whether they want more. A separate set of questions asked whether a provider has ever used AI in their treatment. For those who said yes, Pew asked how well they understand what the tool did.
This slideshow walks through five findings drawn from that report. They cover whether disclosure matters at all, which clinical tasks patients most want flagged and which administrative tasks draw less concern. They also cover how much say patients believe they have and how often they do not know whether AI was involved. Each slide reports Pew's figures, including differences by age, gender, education, race and ethnicity where the survey provides them. All figures come from a single survey. They describe a snapshot of opinion in June 2026 rather than a change over time.
Most U.S. adults say doctors should disclose AI use

Illustration: Quartz
Majorities in every age, gender, education and racial or ethnic group say it is extremely or very important for health care providers to disclose AI use. In the Pew Research Center survey, no group broken out falls below 62% on that measure.
A small minority sees little need for it. Another 16% call disclosure somewhat important. Only 7% say it is not too or not at all important.
Women place more weight on disclosure than men do. Some 74% of women call it extremely or very important. Among men, the figure is 68%. That works out to a six-point gap.
Age produces a steady pattern. Each older bracket in the survey gives disclosure a higher rating than the one before it. Adults ages 18 to 29 come in at 68%, and those ages 30 to 49 at 69%. The share rises to 73% among adults ages 50 to 64. It reaches 76% among those 65 and older. That is an eight-point gap between the youngest and oldest groups.
The figures also differ by race and ethnicity. Some 75% of White adults call disclosure extremely or very important. The share is 67% among Black adults and 66% among Hispanic adults. That leaves a nine-point gap between White and Hispanic adults. Among Asian adults, the figure is 62%. Pew's estimate for Asian adults represents English speakers only.
Education does not follow a straight line. College graduates and adults with some college both stand at 75%. Postgraduates come in lower, at 70%. Adults with a high school education or less are at 67%, the lowest of the four education groups. The highest shares sit in the middle of the education range rather than at either end. The distance from top to bottom is eight points.
Pew places these views alongside a fast-moving debate. It concerns the legal and ethical duty of medical professionals to get patient consent before using AI tools. The survey measures what patients expect from their providers. It does not describe what any rule currently requires.
Patients want to know when AI reads scans or diagnoses

Illustration: Quartz
At least eight in 10 U.S. adults want to be told when a health care provider uses AI for work that can shape their care or medical decisions. Analyzing medical scans, making a diagnosis and explaining lab results all reach that level of support.
Pew Research Center asked about each task separately. For AI that analyzes medical scans, 81% of adults say patients should be told. Another 10% say they should not, and 9% are not sure.
Making a diagnosis produced the same split. Some 81% want to be informed, 10% do not and 9% are unsure.
Explaining lab results lands just behind. Some 80% of adults want to know when AI plays that role. Another 11% say disclosure is unnecessary, and 9% are not sure.
These three tasks have one thing in common. Each can influence a patient's care or a medical decision. Opposition to disclosure stays small for all of them and never rises above 11%. The undecided share is 9% for each of the three.
Support differs by race and ethnicity, though majorities of every group want to be told. White adults are more likely than Black or Hispanic adults to want disclosure for each task.
On diagnosis, 86% of White adults want to be informed. That compares with 72% of Hispanic adults and 64% of Black adults. For scans, the figures are 85%, 78% and 65%, respectively.
Lab results follow a similar pattern. Some 84% of White adults want to be told when AI explains their results. The shares are 75% for Hispanic adults and 65% for Black adults.
Among Asian adults, 83% want disclosure when AI makes a diagnosis. The figure is 84% for scans and 77% for lab results. Pew reports these Asian estimates for English speakers only.
The survey does not explain why these groups answer differently. It records the gaps without giving a cause. Even the lowest figure in the breakdown, 64% of Black adults on diagnosis, still represents a majority.
Fewer patients want disclosure for note-taking and scheduling

Illustration: Quartz
Patients care less about being told when AI handles background and administrative work than when it shapes medical decisions. In the Pew Research Center survey, 72% of U.S. adults want notice when AI takes notes during an appointment. Smaller shares want it for prescription refills, at 64%, and for scheduling, at 56%.
Note-taking draws the strongest call for disclosure of the three. Among adults, 18% say they do not need to be told when AI takes notes on their appointment. Another 9% are not sure.
Ordering a prescription refill from a pharmacy sits in the middle. A 64% majority wants to know when AI handles that step. A quarter of adults, 25%, say no notice is needed. Another 10% are unsure.
Scheduling an appointment produces the weakest demand for disclosure. Just over half of adults, 56%, want to be told when AI books a visit. Another 33% say they do not need to know. That is the largest no-notice group for any task in the survey. The remaining 10% are not sure.
The three tasks form a steady pattern. As support for disclosure falls from note-taking to refills to scheduling, the share saying no notice is needed rises from 18% to 25% to 33%. Even at the bottom of that range, a majority still wants to be informed.
Uncertainty barely moves. Whatever task Pew asked about, between 9% and 10% of adults said they were unsure whether they should be told. The differences between tasks show up almost entirely in the yes and no answers.
The same ordering appears within racial and ethnic groups, though the levels differ. For note-taking, 75% of White adults want disclosure, compared with 69% of Hispanic adults and 62% of Black adults. Among Asian adults, a group Pew reports for English speakers only, the figure is 67%.
For refills, the shares are 65% of White adults, 64% of Hispanic adults and 58% of Black adults. The figure for English-speaking Asian adults is 63%. For scheduling, the shares are 57%, 60%, 46% and 53%, in the same order.
Black adults post a lower share than White or Hispanic adults for each of the three administrative tasks. Their 46% for scheduling means fewer than half of Black adults want notice when AI books an appointment. The survey does not explain why the groups differ.
Most U.S. adults want more say over AI in their care

Illustration: Quartz
Most U.S. adults want more control over how AI is used in their medical care. In the Pew Research Center survey, 63% say they would like more say, while 21% are comfortable with the amount of say they have now.
Pew describes that split as Americans being three times as likely to want more input as to be comfortable with their current level. The two answers sit 42 percentage points apart. Another 16% of adults say they are not sure whether they want more say.
Smaller shares say they already have influence over how AI is used in their care. Some 17% say they have a great deal or quite a bit of say over how AI is used in their health care. Another 16% say they have some say.
Pew combines those two answers into a single figure. By its count, 33% of U.S. adults feel they have at least some say over AI in their care.
Some adults could not place themselves on that scale. A further 14% of adults say they are unsure how much say they have over the use of AI in their health care. That figure comes from a different question than the 16% who are unsure whether they want more say.
The share who feel they have a great deal or quite a bit of say varies by race and ethnicity. Among Hispanic adults, 27% give that answer. Among Black adults, the figure is 23%. White adults are less likely than either group to say so, at 12%.
That puts Hispanic adults 15 percentage points above White adults on this measure. The Hispanic figure is more than double the White figure. Black adults sit 11 points higher than White adults. Among Asian adults, 20% say they have a great deal or quite a bit of say. Pew's Asian estimates represent English speakers only. The survey reports these differences without identifying a cause.
Many patients don't know whether AI was used in their care

Illustration: Quartz
Only 16% of U.S. adults say a health care provider has ever used AI in their care. Another 37% say no provider has, and 46% say they are not sure, according to Pew Research Center.
That makes "not sure" the most common answer to the question. More Americans say they cannot tell than give either a yes or a no. The figures reflect what patients believe about their own care. They do not measure how often doctors, hospitals or clinics actually use the technology.
The share who say yes rises with education. Among adults with a postgraduate degree, 25% say a provider has used AI in their care, 27% say none has and 48% are unsure. College graduates split 22% yes, 33% no and 44% unsure. Adults with some college report 15% yes, 37% no and 48% unsure.
Those with a high school education or less are the least likely to say yes, at 11%. They are also the most likely to say no, at 43%, while 46% are unsure.
The no share moves the other way. It falls from 43% among adults with a high school education or less to 27% among postgraduates. Postgraduates are the only group whose yes and no answers sit within two points of each other. Across all four education groups, the unsure share stays between 44% and 48%. The movement happens between yes and no, not in the size of the uncertain group.
A second question went only to adults who say a provider has used AI in their care. Within that group, 22% say they understand extremely or very well how the technology was used. Some 44% understand it somewhat well, and 32% understand it not too well or not at all.
Pew notes that the share with little or no understanding is larger than the share with a strong grasp. That result comes from a small slice of the public. It describes people who already know AI was involved, not U.S. adults as a whole.
Read together, the two questions point to a gap at two stages. Many patients cannot say whether AI touched their care. Many of those who know it did report only a partial or weak sense of how it was used.
