A Third of America Now Asks the Machine First

What Pew actually measured#

On 25 August 2026 the Pew Research Center published a survey of 3,488 United States adults, fielded between 22 and 28 June. Its headline finding is the number every healthcare seo conversation for the next year will start from: "In all, 34% of Americans ever use AI chatbots for at least one of the eight health and medical reasons we asked about."

Read the framing carefully, because it is the part that gets mangled in summaries. This is a union across eight prompted uses, not a general measure of asking a machine about your health. It is still the largest, cleanest number available on the behaviour, and it comes with a published methodology and a named panel, which is more than can be said for most of the statistics circulating in this category.

Convenience first, diagnosis second#

The reasons matter more than the total. The most common motivation is speed: 28 per cent say getting health information quickly is why they do it. Then 25 per cent for figuring out what is causing symptoms, and 22 per cent each for low or no cost, for information about treatments, and for understanding a diagnosis they already received from a doctor.

Underneath that sit the quieter ones. Twenty per cent use a chatbot to help understand lab results. Eighteen per cent use it to ask about something they are uncomfortable discussing with a person. Fifteen per cent use it to decide whether to see a doctor at all. That last figure is the one a healthcare marketer should sit with, because it is a triage decision being made in a chat window.

Numbers like this get quoted badly for a year, so it is worth planting a marker now. Pew asked about eight specific reasons and reported the union. Anyone who repeats it as "a third of Americans use AI instead of a doctor" has changed the claim into something the survey did not measure and cannot support. Healthcare seo work built on a misquoted statistic tends to collapse the first time a clinician reads the source.

Treat the survey as a map rather than a verdict. It tells you where people are walking, not where they will end up, and the tracks are only two months old. What it does establish beyond argument is that the path exists and is well worn, which is more than most claims about AI search can manage.

The first click is not the first answer#

Anyone who has run medical seo for a clinic knows the traditional funnel by heart. Symptom query, results page, a page that ranks, a form. Two of those four steps are now optional.

Three of the eight reasons are about understanding something a clinician already said. Learning more about a diagnosis, understanding lab results, understanding treatments. That is not a search problem at all, it is a comprehension problem, and it is being solved by a machine rather than by the provider who created the confusion.

For a healthcare brand that is both a rebuke and an opening. The rebuke is that your post-appointment content is not doing its job. The opening is that the content which does do it will be the content the chatbot reaches for, because there is very little good writing in this space and an enormous amount of demand.

Who is asking, and why the age curve misleads#

Usage is not evenly spread. Among adults aged 18 to 29 it is 44 per cent, dropping to 40 per cent for 30 to 49, 32 per cent for 50 to 64 and 17 per cent for those 65 and over.

The obvious reading is that this is a young person's habit and the older cohort is safe. The obvious reading is wrong twice over. Seventeen per cent of the over-65s is a very large number of people in the group that consumes the most healthcare, and the 50 to 64 band at 32 per cent is already a third. Anyone planning healthcare seo services around a demographic exemption is planning around a cliff edge.

Attitude predicts use far more sharply than age does. Among adults who feel more excited than concerned about AI in daily life, 60 per cent are chatbot health users. Among those more concerned than excited, it is 21 per cent. That is a three-fold gap on sentiment against a two-and-a-half-fold gap on age.

The commercial trail here runs straight through the middle of the funnel rather than the top of it. A person reading their own lab results at eleven at night is not a lead in any conventional sense, and they are also exactly the person who will remember which clinic explained the thing clearly. Healthcare seo that only chases the symptom query is hunting the loudest quarry and ignoring the one standing still.

Sentiment moving faster than demographics is the useful finding for anyone planning a healthcare seo budget by audience. You cannot segment on enthusiasm for AI in a media plan, but you can assume the habit spreads along attitude lines rather than age ones, and prowls into older cohorts faster than a simple age curve predicts.

Helpful is not the same as right#

Now the half of the story that healthcare seo decks tend to leave out, because it complicates the pitch.

A physician-led red team study, Draelos and colleagues, assessed 888 chatbot responses to 222 patient-posed, advice-seeking medical questions across internal medicine, women's health and paediatrics, testing four widely used models. Problematic responses ranged from 21.6 per cent for the best-performing model to 43.2 per cent for the worst. Responses judged outright unsafe ranged from 5 per cent to 13 per cent.

The authors' conclusion is blunt: "millions of patients could be receiving unsafe medical advice from publicly available chatbots". Set that beside Pew's finding that 47 per cent of chatbot health users rate the information extremely or very helpful and another 48 per cent rate it somewhat helpful, with only 5 per cent saying it is not helpful, and the shape of the problem is clear. Confidence is close to universal. Accuracy is not.

Only 29 per cent of these users are extremely or very comfortable sharing personal health information with a chatbot, while 26 per cent are not comfortable at all and 42 per cent sit somewhere in the middle. People are using a tool they do not fully trust, for questions they are not comfortable asking a person, and finding it helpful. That is not irrational. It is what happens when the alternative is a three-week wait.

Pew published a companion report the same day on chatbots and loneliness, depression and stress, which is worth reading alongside this one if your clients work anywhere near mental health. The register there is different and the stakes are higher, and the same gap between perceived helpfulness and measured safety runs underneath both.

For a provider the practical consequence is uncomfortable but useful. Some proportion of your patients are arriving having already been told something wrong, confidently, by a machine, and they will not volunteer that. Content that gently corrects the common errors is both good medicine and good healthcare seo, because it is the content nobody else is writing and the exact material a model needs in order to stop repeating the error.

What a healthcare seo programme does differently#

So what does a healthcare seo programme actually do differently? Less than the panic suggests, and more than the complacency does.

The mechanics that make a page quotable are the mechanics that already made it good. Self-contained sentences that survive being lifted out of context. Specific numbers with named sources. A direct answer near the top rather than four hundred words of preamble. Google's own guidance on creating helpful content has said versions of this for years, and it happens to describe exactly what a language model needs in order to cite you rather than paraphrase you into anonymity.

Three things a healthcare seo agency should change this quarter#

First, write the post-appointment content nobody writes. Lab result explainers, treatment comparison pages, plain-language versions of a diagnosis. Three of Pew's eight reasons point here and almost no provider serves them well, which makes it the cheapest defensible ground in medical seo right now.

Second, mark the content up properly so a machine can tell what it is looking at. MedicalWebPage and its associated types exist precisely to declare that a page is medical information, who authored it and what it covers. Structured data is not a ranking trick here, it is the difference between being a source and being text.

Third, put the human back in visibly. Named clinician authors with real credentials, review dates, and a stated review process. When a model has to choose between two equally clear explanations, provenance is the tiebreak, and it is the one thing a content farm cannot fake convincingly.

What none of this justifies is loosening the compliance layer to move faster. The FTC and two states acted against Hims and Hers over privacy practices this summer, and analysis of that action makes the point that regulators judge the net impression across every channel, not the privacy policy in isolation. A healthcare digital marketing agency that ships faster by skipping review is building a liability, not an advantage.

None of this is a reason to abandon the ordinary discipline. Site speed, crawlability, internal linking and a sane information architecture still decide whether any of your careful writing is ever found, and the technical layer is where most healthcare sites still lose. Medical seo has always been unglamorous underneath. The new part sits on top of the old part, it does not replace it.

Budget honestly, too. A healthcare seo agency promising to "optimise for AI" as a separate line item is usually selling the same content work with a newer label on the invoice. Ask what will actually be produced, who will review it clinically, and how it will be measured. If the answer is a schema audit and a monthly report, that is a content problem being sold as a technology one.

Do the unglamorous half first. A named clinician, a review date and a clean answer at the top will outperform any amount of clever schema on a page whose opening paragraph wanders through the undergrowth before saying anything. Healthcare seo rewards the same plainness good medical writing always did.

Five moves before your next patient page ships#

The strategic read is simpler than the tactical one. Search did not stop mattering. It stopped being the only place the answer gets given.

The habit is forming faster than the tooling to see it, and that is the awkward middle every channel passes through. Search was invisible to marketers for years before analytics caught up, and paid social was unmeasurable long after it worked. Healthcare seo is in that gap now: the behaviour is documented, the attribution is not, and the teams that wait for clean numbers will start three years behind the ones that read the survey and moved.

Measurement will lag the behaviour, and you should say so out loud to clients before they discover it in a report. A patient who asks a chatbot, reads its answer, then searches your clinic by name and books, appears in your analytics as brand search. The chatbot did the work and the brand channel took the credit. Any healthcare seo services engagement that promises to attribute this cleanly today is overselling.

The platform layer is still moving underneath all of it. ChatGPT added site tools in the same week that Google split its site reputation enforcement along the European Economic Area border, and volatility in ordinary rankings is high enough that one measured window saw 16.71 per cent of top-ten URLs fall out of the top hundred. Building on any single channel's mechanics right now is building on a moving floor.

For providers outside the United States the numbers will differ but the direction will not, and the compliance ground is stonier. A United Kingdom or European clinic handling patient enquiries carries duties that a chatbot answering the same question simply does not, including breach notification obligations that apply the moment things go wrong. The asymmetry is real and it is not going to be legislated away this year.

There is a version of this story where healthcare marketers panic, and a version where they do the obvious thing. The obvious thing is to write clearly, cite properly, name your clinicians and serve the questions people are actually asking. That was good healthcare seo in 2019 and it is good SEO and GEO now. The difference is that the penalty for doing it badly used to be ranking eleventh. Now the penalty is that a machine answers instead of you, sometimes wrongly, in front of somebody deciding whether to see a doctor.

If you want this mapped against your own site rather than in the abstract, that is the work we do. A healthcare digital marketing agency worth paying will start by reading your existing pages the way a model reads them, which usually reveals that the best clinical thinking on the site is buried four scrolls down under a hero image and a form. Come and ask us, or read our healthcare work first and decide afterwards. Either way, do not let the den go quiet while the neighbours are out foraging.

The brush is thick and the ground is shifting, but the direction of travel is legible. Write the clearest true thing, say who wrote it, show your sources, and make it easy to lift. Do that consistently and a healthcare digital marketing agency becomes unnecessary for the hard part, which is the honest test of whether the advice was ever any good.

Frequently asked questions#