Hospital marketers are often told that "4.5 stars is the line". Two 2026 surveys give that claim some support, and both have limits that a CMO should understand before turning it into a target in someone's KPI sheet. This post is for hospital and multi-clinic marketing leads who own reputation numbers. It reads the data carefully, shows the arithmetic of moving a rating, and gives you a way to test whether a threshold exists in your own enquiries.
The short version is that ratings behave like a filter and not a dial. Below a certain point, a share of patients will not consider you. Above it, additional stars buy less than recent, specific, answered reviews do. Neither survey is Indian, and both ask about stated behaviour, not recorded bookings, so we treat them as hypotheses to test locally. Stated behaviour often overstates how strictly people filter when they are in pain or in a hurry, and that is another reason to test.
What the two surveys say
The rater8 2026 Patient Choice Report surveyed 992 US adults in April 2026. It found that 47% had used an AI tool to research healthcare providers, up from 31% in 2025, and that 55% had cancelled or avoided booking a provider because of online reviews. It also found that 75% would not book a provider rated below 4.0 stars, and 44% would not book below 4.5. rater8 sells reputation software, so this is a vendor's survey, though the methodology is published.
The BrightLocal Local Consumer Review Survey, published in February 2026, asked 1,002 US adults about local businesses in general, not healthcare. It found that 31% will only use a business rated 4.5 stars or higher, and that 74% look for reviews written in the last three months.
Read together, the two surveys do not say that patients choose 4.5. They say that a large minority use a high cut-off, a larger group use a lower one, and most check how recent the reviews are. For a hospital in India, those proportions will differ by city, specialty and the platform patients use. Treat the surveys as an argument to look at your own distribution, not as a benchmark to copy.
What the threshold means in practice
A threshold changes how you should report. If the cut-off sits near 4.0 for many patients and near 4.5 for some, then:
- Below 4.0 is a visibility problem. A large share of searchers exclude you before reading anything.
- Between 4.0 and 4.4 is a competitive problem. You are considered, but you lose comparisons with nearby alternatives that show 4.5 or above.
- At 4.5 and above the return on additional rating points falls. Effort is better spent on recency, response quality and review content.
The rating that matters is the one shown on the search result or map card, which Google displays to one decimal place. A 4.46 shows as 4.5 and a 4.44 shows as 4.4. That rounding is invisible in your dashboard and obvious to a patient.
The arithmetic of moving a rating
Averages move slowly once you have volume. If a profile has n reviews with an average rating of a, and you add k new five-star reviews, the new average is (n × a + 5k) ÷ (n + k). To reach a target t, you need k ≥ n × (t − a) ÷ (5 − t).
A worked example, with invented numbers. A hypothetical 150-bed hospital in Nagpur has 400 reviews averaging 4.3. To reach a displayed 4.45 it would need k ≥ 400 × 0.15 ÷ 0.55, which is about 110 additional five-star reviews with no other new reviews at all. Every new review below five stars raises that number.
Three consequences follow.
- Volume protects and punishes. A branch with 40 reviews at 4.2 needs only about 24 new five-star reviews to reach 4.5. A flagship with 4,000 reviews at the same average needs about 2,400.
- Location-level profiles are the unit. Patients see the profile of the branch near them. A group average of 4.4 hides one branch at 3.9.
- Improving service moves the rating before marketing does. If a department's reviews mention waiting time at billing, no request campaign will fix it. Send the pattern to operations.
Why review text now matters as much as the number
If 47% of patients are using AI tools to research providers, as the rater8 report found, then the written content of reviews is being read and summarised by software as well as by people. We have not seen a published test of how any AI system weighs review text for healthcare queries, so this is an inference and not a finding. It is a sensible one: a review that says "the orthopaedic team explained the physiotherapy plan clearly" contains the department, the topic and the experience in language an answer engine can reuse. A review that says "good" contains nothing.
You can ask for specific reviews without scripting them. A request message that asks the patient to mention which department they visited, and what they found useful, is a prompt. A message that dictates the content is not, and Google prohibits fake and incentivised content under its contributed content policy. Also avoid filtering who is asked: sending requests only to patients who gave a high satisfaction score is the practice usually called review gating, and it conflicts with Google's guidance on not selectively soliciting reviews.
