AI was supposed to replace Googling your symptoms. Instead, apparently, we have invented another place to worry about them.
That is the more useful finding inside a new Luth Research study on how people use artificial intelligence while looking for health information. The company paired passive digital tracking of 1,538 consumers with a survey of 537 adult patients who had used AI for health information during the previous year. Rather than finding a clean migration from search engines and health websites to AI, Luth found people moving among all of them. Roughly 80% of AI health-information users also use other sources, creating what the company calls a group of “hybrid researchers.”
The phrase is corporate research language, but the behavior underneath it is wonderfully human. We are repeatedly told that new technologies will replace old ones because replacement makes a much cleaner story. Streaming replaces television. Search replaces directories. Messaging replaces phone calls. AI replaces search. Actual behavior tends to be less elegant. We keep the old thing, add the new thing, discover that each is good for something slightly different and build ourselves a longer route.
Health information may be an especially clear example because the question a patient is asking rarely stays the same for very long. A person may begin with, “What does this term mean?” Then: “Should I be worried?” Then: “What are the usual treatments?” Then: “Is this source credible?” Then: “What should I ask my doctor?” Each question can send the person toward a different tool, because speed, simplicity, familiarity and trust are not the same requirement.
Luth’s findings describe exactly that division of labor. Forty-seven percent of surveyed AI users said they use AI to dig more deeply into a doctor’s explanation, while 44% said asking AI is easier than asking their physician. AI is useful at the point where somebody wants an immediate explanation in ordinary language and does not want to wait for another appointment, search through a long article or admit that they did not understand something the first time.
Then, according to the company’s research, many of those people leave the AI tool and go somewhere else. Health-information sites were the most common companion source, used by six in 10 AI users. Luth’s accompanying analysis says 73% verify AI-provided health information against a source they already trust. As people move closer to an appointment or treatment decision, the company says use of established health-information sites rises while reliance on AI falls.
That is a very different picture from replacement. It is interpretation layered on top of interpretation.
The older health-information journey was already messy. You noticed a symptom. You searched for it. You found a terrifying result. You refined the search. You found a calmer result. You visited a hospital page. You checked a patient forum. You texted somebody who once had something vaguely similar. Eventually, maybe, you called a clinician. The joke about Googling symptoms ending in a diagnosis of imminent death worked because the process mixed genuine information-seeking with anxiety, uncertainty and the irresistible human desire to know what is happening before anyone qualified is available to tell us.
AI does not remove any of those motives. It changes the first interaction. Instead of receiving ten blue links and deciding which one might contain an answer, the user can ask a conversational system to synthesize the answer immediately. That can reduce friction. It can also introduce a new question almost instantly: “Can I trust what it just told me?”
And there is the extra stop.
If the answer is important enough, the user may now do exactly what Luth is observing: ask the AI, search the terminology, visit an established health-information source, compare the results and possibly return to the AI with a more informed question. The technology that promised a shorter route can become another layer in the route because the answer has to be interpreted and then the interpretation has to be checked.
That does not make the tool useless. In fact, the opposite may be true. Luth found that people using both AI and health sites spent three times longer researching chronic conditions than people using AI alone, and patients who had lived with a diagnosis for 11 years or more were 27% more likely to behave as hybrid researchers. Experience does not necessarily produce loyalty to one information source. It may teach people how to combine several sources according to what each one does well.
That is a useful correction to the recurring fantasy that technology simplifies human behavior by removing steps. Sometimes it does. More often, removing one friction point reveals the next one. Search made information easier to find, which created the problem of deciding which information deserved trust. Social platforms made people easier to reach, which created the problem of deciding whose advice deserved attention. AI makes explanations easier to obtain, which creates the problem of deciding when an explanation is reliable enough to act on.
Health raises the stakes because the difference between understanding and deciding matters. Asking an AI to translate a clinical phrase into plain English is not the same as asking it whether to start, stop or change treatment. Luth’s research suggests users themselves may already be making some version of that distinction. Its company blog says the closer patients get to preparing for an appointment, the more their use of health-information sites increases and their reliance on AI declines. AI appears strongest in the exploratory part of the journey; established sources become more important when the decision feels real.
That does not mean every patient follows the pattern, and it does not establish that every verification step is adequate. Luth’s sample consists specifically of adults who had used AI for health information during the prior year, so the study is describing behavior among AI health users rather than all patients. But within that population, the result is useful because it replaces a dramatic technology story with a much more believable human one.
We accumulate methods.
Most of us did not stop asking family members questions when search engines arrived. We did not stop using search engines when social media made communities searchable. We did not stop visiting websites when apps appeared. The new method usually settles into the existing collection, and over time we develop an informal sense of which doorway feels right for which question.
AI may be becoming the doorway for questions that feel embarrassing, confusing, too small for an appointment or too urgent to leave unanswered until morning. Search may still be where we look for the exact phrase, the drug name or the official site. A trusted health publisher may be where we go when the answer needs institutional weight. The clinician remains the person who can actually examine us, understand the broader medical context and make decisions with us rather than merely describe possibilities.
Seen that way, the patient journey has not been replaced at all. It has acquired another interpreter. And because we are humans, we immediately gave ourselves another place to check the interpreter.
SOURCE NOTES
• Luth Research / PR Newswire, Sept. 16, 2026: AI Patient Journey Study announcement
• Luth Research: “AI Didn’t Replace the Patient Journey. It Added a Step.”
• Reporting note: Luth Research says it passively tracked 1,538 consumers and surveyed 537 patients, all adults 18+ who had used AI for health information in the prior year. Fieldwork was completed in 2026. Survey and behavioral findings are treated here as company-reported.
Study findings attributed to Luth Research materials cited in SOURCE NOTES. Cultural framing is RMN's.