How patients pick a doctor

Eighty-four percent of U.S. patients say they check online reviews before choosing a new healthcare provider, according to a December 2024 survey of 1,008 adults by patient-experience firm rater8. That sits at the high end of a band recent surveys put between 71% and 90%, and the spread owes more to question wording and fielding date than to a real swing in behavior. What has moved further is trust in what patients read once they get there: the share of Americans who say they trust an online review as much as a personal recommendation from a friend peaked at 84% in 2016 and 2017, fell to 42% by early 2025, and recovered to only 49% in 2026, per BrightLocal’s Local Consumer Review Survey. Patients are reading more reviews than ever. They are trusting them less than they did a decade ago.

The share who check reviews keeps climbing

Three surveys fielded between 2020 and 2025, run by three different firms with three different panels, land within 11 points of each other on the core behavior. Software Advice’s 2020 patient-experience survey (498 qualified respondents, screened from 854) found 90% of patients use online reviews to evaluate a physician, and 71% use online reviews as the very first step in finding a new doctor. Tebra’s sixth annual Patient Perspectives Report, surveying 3,964 U.S. adults and fielded June 29, 2025, put the read-before-you-book figure at 79%. rater8’s December 2024 panel of 1,008 patients put it highest, at 84%. None of the three asks an identical question — “use,” “start with” and “check before choosing” are not interchangeable verbs — which is the honest reason the numbers do not line up to the decimal.

Star ratings now function as a cutoff, not a nudge

Where the surveys agree more precisely is on what happens once a rating drops below a threshold. Tebra found 53% of patients will not consider a provider with fewer than 4 stars. rater8’s 2026 Patient Choice Report, published June 3, 2026 on a fresh sample of roughly 1,000 adults, found 75% will not book with a provider rated below 4.0 — and that 55% have walked away from a doctor entirely because of what they read online, up 15 percentage points from the prior year’s edition of the same report. Provider responses to reviews carry weight too: 66% of rater8’s 2026 respondents said a provider’s reply to feedback directly influences their trust, up 24 points on 2025, and Tebra separately found 70% of patients said a reply to a negative review improved their opinion of the provider. Reviews have also started to override cost: in the 2020 Software Advice survey, 43% of patients said they would go outside their insurance network for a provider with strong reviews.

Trust in reviews is a separate, older story

The behavior above measures whether patients use reviews. A longer-running BrightLocal series measures something narrower: whether consumers trust a review as much as they would trust a recommendation from someone they know. That number is not healthcare-specific, but as the closest tracked proxy for general review credibility it is the backdrop against which the read-before-you-book figures sit. It peaked at 84% in 2016 and 2017, had fallen to 79% by 2020, and cratered to 42% in BrightLocal’s January 2025 edition (1,026 U.S. consumers) before ticking back up to 49% in the 2026 edition (1,002 U.S. consumers). Patients keep opening the review tab. Fewer of them now treat what they find there as equivalent to a friend’s word.

AI summaries are becoming the next reference check

rater8’s 2026 report also caught the next layer forming on top of the review stack. Thirty-six percent of patients now cite AI tools as influential when searching for a doctor, up from 17% a year earlier — enough to pass both a plain Google search (34%) and a doctor’s own recommendation (32%) as a stated influence. Thirty-seven percent said they trust Google’s AI Overviews most for healthcare searches specifically. The same report found 66% of patients had encountered incorrect AI-generated provider information, and that 60% of that group said they trusted the AI summary anyway, without checking it further. If the last five years were about patients reading reviews instead of asking a friend, the next data point to track is whether they read an AI summary of the reviews instead of the reviews themselves.

Methodology

Sources

Software Advice, “How Patients Use Online Reviews: What You Need to Know,” 498 qualified U.S. patient respondents screened from 854, fielded 2020. rater8, “How Patients Choose Their Doctors” report, 1,008 U.S. adult patients, fielded December 2024, SurveyMonkey panel. rater8, “2026 Patient Choice Report,” published June 3, 2026, approximately 1,000 U.S. adult patients. Tebra, sixth annual “Patient Perspectives Report,” 3,964 U.S. adults, fielded June 29, 2025. BrightLocal, “Local Consumer Review Survey 2025,” 1,026 U.S. adult consumers, published January 29, 2025. BrightLocal, “Local Consumer Review Survey” (2026 edition), 1,002 U.S. adult consumers. BrightLocal, “Historical Trends in Consumer Review Behavior,” cited for the 2016–2017 trust peak.

Scope limits

All seven surveys cited here are U.S. consumer or patient panels; no comparably rigorous, comparably sized UK or EU-wide survey of patient review behavior turned up during research for this brief, which is a genuine gap for cross-market comparison, not an editorial choice to exclude those markets. The three “read/use/check reviews” figures (71–90%) come from three different questions on three different panels and are not a single tracked series — they are presented side by side, not averaged. The BrightLocal trust figure measures general consumer trust in reviews, not a healthcare-specific question, and is used here as the closest available proxy for review credibility over time. Two of the seven sources (rater8, Tebra) are vendors that sell reputation-management or patient-engagement software to healthcare providers; their survey questions and framing may reflect that commercial interest, though the topline figures are consistent with the non-vendor Software Advice and BrightLocal data.

Updates

This brief will be revised as new data is published. Corrections are handled under our corrections policy.