Do Patients Actually Want AI Scribes? What the 2025–2026 Research Says

If you are thinking about adding an AI scribe, you have probably wondered how your patients will react. A lot of new data on that question came out in the past year. It splits into two camps that seem to contradict each other.

Ask the general public, and people are wary. Ask patients who just had a visit with an AI scribe, and most of them liked it.

Both findings are real. Most of the gap comes from who gets asked, and whether they've actually seen the technology work. That gap matters a lot for how you introduce a scribe in your own exam room.

The skeptical camp: the general public

Canada, 12,153 adults (JAMIA, 2026). This is the largest study so far. Chandrasekaran and Moustakas analyzed an online survey of 12,153 Canadian adults, fielded February 6 to March 10, 2025 (JAMIA 2026;33(2):263–272, doi 10.1093/jamia/ocaf218). Only 28.3% knew clinicians were using AI scribes at all. 61.8% were reluctant about future use. 39.3% reported some or very high comfort. 57.4% said they would trust AI documentation with human oversight, and 49.5% expected AI scribes to improve their interactions with clinicians. The authors summed it up as "conditional trust and comfort" alongside reluctance to adopt.

England, 4,039 adults (Healthwatch England / YouGov, 2026). This nationally representative poll ran April 16–27, 2026, and was published July 16. It found:

• 48% were somewhat or very comfortable with an AI scribe at a routine health check.
• Comfort dropped sharply for sensitive visits: 29% for sexual health, 28% for mental health, 23% for domestic abuse.
• 81% wanted clinicians to say when an AI scribe is in use and to ask for consent.
• 69% would be more comfortable if the clinician committed to checking the note for accuracy.
• Overall support for rollout was split almost evenly: 38% support, 37% oppose.
• Of the respondents who'd had an appointment in the past year, nearly 90% didn't know whether AI scribing had been used.

United States, general AI attitudes (Ohio State / SSRS, 2026). This survey isn't about scribes, but it shows the trend. In a poll of 1,007 U.S. adults fielded January 16–20, 2026, 42% said they were open to AI being used in their care, down from 52% in 2024. Public sentiment about AI in medicine is getting cooler, not warmer.

The positive camp: patients who have actually had an AI scribe in the room

Stanford Health Care, 2,202 patients (JAMIA Open, 2026). Shah et al. surveyed outpatients after visits where an ambient scribe was used, May to July 2025 (JAMIA Open 2026;9(3):ooag104). 70.1% found it helpful, and 73.6% wanted it used at future visits. Clinicians had to tell everyone in the room before turning the tool on, and patients could decline.

Kaiser Permanente (The Permanente Medical Group, 2025). This was a small patient survey of 118 respondents, reported alongside TPMG's evaluation of 7,260 physicians and roughly 2.5 million encounters. About two-thirds were comfortable with the technology, 26% were neutral, and 8% reported some discomfort. 47% said their doctor spent less time looking at the computer.

Emory, 8,120 patient surveys (Applied Clinical Informatics, 2026). Patel et al. compared post-visit surveys from visits with and without an ambient scribe (February to April 2025; Appl Clin Inform 2026;17(1):82–88). Patients whose clinician used a scribe had slightly higher odds of being satisfied with time spent with the provider (OR 1.13). The other five domains showed no detectable difference, including communication, attention, and after-visit summary quality. In other words, the scribe made things slightly better and didn't make anything worse.

The VA. Since ambient scribes started rolling out in October 2025, more than 986,000 VA appointments have used one. According to vendors and a VA official, veterans declined less than 1% of the time (Nextgov, September 2026). That figure isn't peer-reviewed, but it's the largest real-world refusal rate on record.

Smaller studies point the same way. In a Singapore time-motion study, clinicians' eye-contact time rose from 69.6% to 77.1% of the visit. Of 39 patients surveyed, 69.2% felt the physician focused on them more and none reported discomfort (JMIR Med Inform 2026;14:e85580). In an Australian dermatology clinic survey of patients who'd had little prior exposure, 82.1% were comfortable with AI scribe use, although more than half still had concerns (Australas J Dermatol, July 2026).

Why the numbers disagree

1. Experience. In the Canadian and English polls, most people were judging something they had never seen. The Stanford, Kaiser, Emory, and VA numbers come from people who had just been through a visit with a scribe. This suggests the abstract idea of "an AI recording my doctor's visit" sounds worse than a normal visit where the doctor looks at you instead of a keyboard.

2. Who answers the survey. Post-visit surveys pick up people who answer post-visit surveys. Stanford's authors say plainly that they couldn't calculate a response rate and that 59.8% of respondents were 70 or older. A pre-launch UC Davis survey had a 20% response rate and skewed female, older, and White (JMIR Med Inform 2025;13, Leiserowitz et al.). Read the positive numbers as "acceptable to many patients who respond," not "acceptable to everyone."

3. What the question asks. "Was it helpful?" asked about a visit that already happened is a different question from "Would you want this in the future?" asked of strangers. The answers shouldn't be expected to match.

Who is more hesitant

Several subgroup patterns show up in more than one study, and you should plan around them:

• Women. In Canada, men had higher adjusted odds of comfort (1.13), trust (1.21), and willingness for future use (1.38). At Stanford, 84.0% of men vs. 76.1% of women wanted future use.
• People in poorer health, and with lower education or income. In the Canadian data, all of these were tied to less favorable attitudes. Privacy concern was the strongest negative predictor, cutting the odds of future use to 0.65.
• Younger adults, sometimes. At UC Davis, patients aged 18–30 were more skeptical than those over 51. In the dermatology survey, older patients worried more about privacy and younger patients about accuracy. "Older patients will object" is not a safe assumption.
• Anyone at a sensitive visit. The England data is the clearest here: comfort roughly halves for sexual health, mental health, and abuse disclosures.

UC Davis also asked what worried patients. Of the comments raising concerns, 39% were about documentation accuracy and 15% about privacy. At the same time, 23% of respondents were already frustrated that their physician focused too much on typing. Patients don't love the status quo either.

What this means for how you introduce a scribe

The research points to a consistent playbook. None of it is complicated:

1. Tell every patient, every time, and give them a real option to say no. The 81% "tell me and ask me" figure from England matches what UC Davis patients said ("previsit patient education and obtaining permission were viewed as very important"). In many states it's also the law. See our guide to AI scribe consent and recording laws and the 2026 state AI disclosure rules.
2. Lead with the benefit patients care about: your attention. The most consistent positive finding across studies is less screen time and more eye contact. "This lets me look at you instead of my computer" works better than "this saves me time."
3. Say out loud that you review every note. In England, 69% said this would make them more comfortable. In Canada, 57.4% said they'd trust AI documentation with human oversight. It's also true: you sign the note. Our post on AI scribe accuracy and hallucinations covers why that review matters.
4. Have a ready answer to "what happens to the recording?" Privacy is the top concern in several studies. Know whether your vendor keeps audio, for how long, and whether it trains on your data. Also know that patients may request the recording itself. Our HIPAA guide for AI tools lists the questions to ask vendors.
5. Pause the scribe for sensitive content. That includes sexual health, mental health, abuse, substance use, and confidential adolescent time. The England data shows comfort halving in exactly these areas. Our adolescent confidentiality workflow covers a pediatric-specific version.
6. Make "no" easy and painless. A patient who declines should get exactly the same visit. If declines run well above the low single digits, your script is probably the problem, not the patients.
7. Track declines and complaints from day one. Count them as part of your rollout metrics (see our rollout guide). This is your own local version of the research above, and it's more relevant to your patients than any national poll.

For patients who need an interpreter or don't speak English, the experience is different, and so are the accuracy risks. See AI scribes in non-English and interpreter visits.

Where the evidence is still thin

• No study has tracked the same patients before and after experiencing a scribe. The "experience closes the gap" explanation is plausible, but it's an inference from two different populations, not a measured effect.
• Almost all positive data comes from large academic systems and integrated networks (Stanford, Kaiser, Emory, the VA). There's very little from solo practices or small clinics, where patients may know the clinician better and the consent conversation looks different.
• Few studies separate out tools. Most don't name the vendor, so you can't infer that patients prefer one product over another.
• Behavioral health is underrepresented. Given the England sensitivity numbers, therapists should be especially careful about assuming acceptance. Our therapy AI privacy guide covers the extra considerations.

The bottom line

The headline that "most patients don't want AI scribes" and the headline that "patients love AI scribes" both come from real data. Patients who have never experienced an AI scribe are cautious, and that caution is growing. Patients who have had one used in their visit mostly report a better or unchanged experience, and very few say no when they're asked properly.

The deciding factor is the first conversation. Tell patients, ask them, explain that you review every note, and pause the scribe when a visit turns sensitive. Once that conversation goes well, most patients care about whether you're looking at them, not which software is listening.

If you're still choosing a tool, browse the AI medical scribes in our directory. You can also compare options for family medicine or small clinics. Popular choices include Freed, Abridge, Nabla, Heidi Health, and Suki. Check each vendor's consent tools and audio-retention policy before you sign.

This article is for informational purposes only and is not legal or medical advice. Survey findings reflect the populations and dates studied and may not generalize to your patients. Consent and recording requirements vary by state; confirm current rules with counsel and verify product features directly with each vendor.