Many Americans Turn To Ai For Health Answers Despite Accuracy Warnings

Traditionally, individuals asked wellness questions of their superior wellness attraction provider. Confidence successful that supplier arsenic a root of trustworthy wellness accusation has been consistently precocious successful caller years and was astatine 90% successful April, according to study information from nan Annenberg Public Policy Center (APPC) of nan University of Pennsylvania.

The emergence of nan net created fresh entree to a 2nd root of master wellness knowledge – authorities wellness authorities specified arsenic nan Centers for Disease Control and Prevention (CDC), Food and Drug Administration (FDA), and National Institutes of Health (NIH). Although assurance successful these sources eroded during nan Covid-19 pandemic, comparative to astir different institutions assurance successful these institutions has remained high, ranging from 66%-68% successful April for nan 3 wellness agencies.

The advent of generative artificial intelligence (AI) has added a 3rd root of wellness knowledge and recommendations. Whether this root is trusted – and, if so, to what grade – is an unfastened question, but its beingness is each but unavoidable. A Google hunt for nan reply to a wellness question, for example, will elicit a consequence led by an "AI Overview" of nan hunt results containing a disclaimer successful mini people astatine nan end: "This is for informational purposes only. For aesculapian proposal aliases diagnosis, consult a professional. AI responses whitethorn see mistakes."

Although Google notes that its AI summaries whitethorn not beryllium accurate, immense numbers of Americans are exposed to these answers and astir see them reliable, according to nan latest APPC wellness survey. Conducted successful April 2025 pinch complete 1,600 U.S. adults, nan study finds that astir 8 successful 10 adults opportunity they're apt to spell online to reply a circumstantial mobility astir wellness symptoms aliases a wellness condition. Nearly two-thirds of this group opportunity they person seen AI-generated responses. 

Highlights

The Annenberg Public Policy Center's April 2025 wellness study finds that:

• Most (79%) U.S. adults opportunity they're apt to look online for nan reply to a mobility astir a wellness denotation aliases condition.
• Three-quarters (75%) of group who hunt online opportunity that AI-generated responses supply them "sometimes" (45%) aliases "often aliases more" (31%) pinch nan reply they need.
• Most Americans (63%) deliberation AI-generated wellness accusation is somewhat (55%) aliases very (8%) reliable.
• Nearly half (49%) are not comfortable pinch wellness attraction providers utilizing AI devices alternatively than their acquisition unsocial erstwhile making decisions astir their care.

Almost 1 successful 3 group (31%) who usage an online hunt motor to find wellness accusation opportunity nan AI-generated summary consequence often aliases ever gives them nan reply they request while different 45% opportunity it sometimes gives them nan reply they need. Nearly two-thirds (63%) find AI-generated results to beryllium somewhat aliases very reliable. Yet astatine nan aforesaid time, almost half of U.S. adults (49%) opportunity they are not comfortable pinch wellness attraction providers utilizing AI devices alternatively than their acquisition unsocial erstwhile making decisions astir their care.

"Despite nan disclaimers that travel immoderate AI-generated summaries, location is imaginable for disorder and moreover harm among susceptible individuals if they are not alert that these responses are not a substitute for nan personalized master wellness guidance that their wellness attraction supplier tin offer," says Kathleen Hall Jamieson, head of nan Annenberg Public Policy Center and co-editor of nan book "Realizing nan Promise and Minimizing nan Perils of AI for Science and nan Scientific Community" (Penn Press, 2024), which is disposable astatine nary costs online.

APPC investigation expert Laura A. Gibson added, "AI platforms are not needfully updated successful existent clip and whitethorn incorporate outdated information. Skepticism is warranted."

Where do group activity wellness accusation online?

While astir group opportunity they activity online wellness information, location isn't a azygous root aliases type of root to which they turn. The study finds that nan astir celebrated resources are hunt engines specified arsenic Google aliases Bing, pinch 71% of respondents saying they often aliases occasionally get wellness accusation from these types of sources. Following hunt engines, nan astir communal sources group study utilizing are online web sources (48%) for illustration WebMD aliases healthline.com; websites of ample aesculapian investigation and objective entities (44%) specified arsenic nan Mayo Clinic and Cleveland Clinic; nonprofit organizations (37%) focused connected circumstantial topics for illustration nan American Cancer Society (cancer.org), American Heart Association (heart.org), and American Diabetes Association (diabetes.org); and national authorities resources (35%) specified arsenic nan CDC, nan NIH, and nan FDA.

Fewer group study getting accusation often aliases occasionally from authorities aliases section authorities wellness sites (14%), podcasts (18%), nan World Health Organization (19%), aliases societal media sources for illustration Facebook, X (formerly Twitter), TikTok, YouTube, aliases Reddit (24%).

Most opportunity AI-generated wellness accusation provides what they request and is reliable

Among users who opportunity they hunt for wellness accusation online, astir two-thirds (65%) study that they person seen AI-generated responses astatine nan apical of nan results connected hunt engines. On Google, for example, this is branded an "AI Overview," and connected Bing, arsenic a "Copilot Answer."

Seniors property 65 aliases older who opportunity they hunt for wellness accusation online are importantly little apt than younger group to opportunity they've seen AI-generated wellness responses atop their hunt results. Under half (49%) of those property 65+ opportunity they person seen AI wellness responses, compared pinch 69% of 18- to 34-year-olds, 75% of 35- to 49-year-olds, and 63% of 50- to 64-year-olds.

Among those who person seen AI-generated responses astatine nan apical of results erstwhile searching for wellness information, much than 3 successful 10 (31%) opportunity AI ever (4%) aliases often (26%) provides them pinch "the reply you need." Nearly half (45%) opportunity AI sometimes gives them nan consequence they need. About 1 successful 5 (18%) opportunity they seldom (13%) aliases ne'er (6%) get nan reply they request and 6% were unsure.

Most group who hunt for health-related accusation don't extremity pinch an AI-generated response. About two-thirds (65%) opportunity they ever (31%) aliases often (34%) proceed searching by pursuing links to circumstantial websites aliases different resources. Another 4th (24%) opportunity they sometimes travel nan links, while 8% seldom aliases ne'er spell beyond nan AI-generated response.

Most group who hunt for wellness accusation online opportunity they find AI-generated wellness accusation to beryllium reliable. Nearly two-thirds (63%) find nan accusation to beryllium reliable, pinch 55% saying nan accusation is somewhat reliable and 8% saying it is very reliable. A 4th of online searchers (25%) opportunity AI-generated wellness accusation is mostly unreliable.

Sharing online wellness accusation pinch one's wellness attraction provider

When visiting their expert aliases wellness attraction provider, fewer group who hunt for wellness accusation online (12%) opportunity they ever aliases often talk pinch their expert aliases wellness attraction supplier astir wellness accusation they saw online. Two-fifths (40%) sometimes do and astir half opportunity they do truthful seldom (29%) aliases ne'er (17%).

Among group who regularly sojourn their wellness attraction supplier and hunt for online wellness information, astir work together (80%) that their wellness attraction providers are mostly bully astatine explaining wellness issues and treatments to them. Nevertheless, respondents are mixed astir whether their wellness attraction providers are dismissive aliases do not person nan clip to talk online wellness accusation they bring to their visits. A 3rd (32%) opportunity there's usually not capable clip astatine aesculapian appointments to talk pinch their providers astir wellness accusation they saw online and 40% consciousness location is capable time, while 28% neither work together nor disagree. As for nan providers' receptivity to nan online accusation brought successful by a patient, 22% work together wellness attraction providers will disregard what they bring in, but 37% disagree pinch that statement. Four successful 10 (41%) neither work together nor disagree.

Are you comfortable pinch wellness attraction providers utilizing AI tools?

Nearly half of those surveyed (49%) opportunity they are not comfortable pinch wellness attraction providers utilizing wellness attraction AI devices alternatively than their acquisition unsocial erstwhile making decisions astir nan patient's care. More than a 3rd (36%), however, definitive astatine slightest immoderate comfortableness pinch their wellness attraction providers utilizing AI devices beyond their ain expertise/experience.

Despite nan wide unease pinch wellness attraction providers utilizing AI, much than 4 successful 10 group surveyed (41%) deliberation their wellness attraction supplier regularly uses wellness attraction AI devices sometimes (31%), often (9%), aliases ever (1%), alternatively than their acquisition unsocial erstwhile making decisions astir diligent care. Over a 4th (26%) opportunity their wellness attraction providers seldom aliases ne'er usage AI, while a 3rd (33%) are not judge whether they usage AI aliases not.

Many usage smart devices to way their wellness data

Overall, 93% of respondents opportunity they ain a smartphone aliases smartwatch. Adults 65 and older are importantly little apt (83%) to ain 1 than younger people, who scope by property group from 98% to 95%.

About 6 successful 10 (59%) smart instrumentality users – telephone aliases watch – study that they usage an app connected their instrumentality that summarizes wellness information, specified arsenic steps per day, slumber clip and slumber quality, aliases bosom health. Forty percent of instrumentality users opportunity they do not usage specified an app connected their device. By age, those pinch a smart instrumentality who are betwixt nan ages of 18 and 34 (62%) and 35 and 49 (61%) are importantly much apt to opportunity they usage specified an app than those property 65 aliases older (51%).

A mostly of smart instrumentality users (52%) opportunity they'd beryllium apt to stock immoderate of their summary wellness accusation pinch their expert aliases different wellness attraction provider. This includes nan 3% who opportunity they do truthful already. More than a 3rd (36%) opportunity they are not apt to stock this accusation pinch their wellness attraction provider.

While a sizable number of smart instrumentality users do not wish to stock their summary wellness attraction information pinch their doctors aliases wellness attraction providers, nan immense mostly of study respondents are comfortable having their pharmacies nonstop them alerts astir their prescriptions. Eight successful 10 of respondents opportunity they either already get alerts (31%) from their pharmacies aliases would beryllium apt to motion up for specified alerts (49%).

The Annenberg Science and Public Health survey

The study information travel from nan 24th activity of a nationally typical sheet of 1,653 U.S. adults conducted for nan Annenberg Public Policy Center by SSRS, an independent marketplace investigation company. Most person been empaneled since April 2021. To relationship for attrition, replenishment samples person been added complete clip utilizing a random probability sampling design. The astir caller replenishment, successful September 2024, added 360 respondents to nan sample. This activity of nan Annenberg Science and Public Health (ASAPH) study was fielded April 15-28, 2025. The separator of sampling correction (MOE) is ± 3.4 percent points astatine nan 95% assurance level. All figures are rounded to nan nearest full number and whitethorn not adhd to 100%. Combined subcategories whitethorn not adhd to totals successful nan topline and matter owed to rounding.