the answer lands before your website

it is late. somebody has a new diagnosis, or a dosing question they are embarrassed to ask again. they type it into chatgpt, or perplexity, or claude, or gemini, before they open anything with a brand logo on it. sometimes before the appointment. that first answer is already shaping what they believe, what they fear, and what they will ask the clinician tomorrow.

pharma and life sciences has a ux problem. patients, caregivers, and clinicians already get answers about conditions and drugs somewhere else first. the brand site is often not even the second stop. that late-night search is not a one-off. the data says the same thing.

pharma spends years making the on-label story accurate. medical, legal, and regulatory review, mlr for short, is the reason a claim can be trusted. marketing builds the sites, portals, and patient pages that carry it. and still the question often gets answered by an ai chatbot.

that isn't because a marketer didn't write enough pages. the verified words exist. they sit behind brand sites and review cycles, while people already ask chatbots that answer in the conversation itself. the hard work is getting those verified words into that first answer, in language a person can use.

language is half the gap. half of insured adults say at least 1 part of how their insurance works is hard to understand¹. what it covers, what they will owe, or basic terms like deductible. health information on the web has the same problem. people skim under time pressure. lower-literacy readers often cannot skim at all. nielsen norman group research, including a pfizer-sponsored rewrite of a major pharma product site, showed that simpler writing and clearer layout helped². people found the right information faster and liked the experience more. hhs's health literacy online guide makes the same demand³. design for people who do not have strong reading skills, and for people who do not have a lot of time. the demand for better, clearer ux is not new. the channel people use when brand ux fails them is.

patients already left the homepage

rock health's late-2025 consumer survey of 8,000 us adults found that 1 in 3 had used an ai chatbot for health information. that share doubled in a year, from 16% to 32%. among those users, 64% do it at least weekly⁴. they ask about treatment options, symptoms, and prescription drugs. kff's separate tracking poll lands nearby: about 1 in 3 adults turn to ai for health information or advice⁵. different method, so treat it as direction, not the same study. among the reasons people give kff for using ai, speed is first. difficulty accessing or affording care shows up too, especially for younger and lower-income users⁵.

none of that is surprising if you have sat with what patients say they lack. across 95 studies and 32,493 adults with cancer, unmet information needs still ranked first, with psychological needs second⁶. the gap is old. the channel people use to fill it is new.

clinicians are in the same chat

doctors do a version of this too. an iqvia-led physician survey in partnership with medscape put generative-ai use highest for understanding guidelines, clinical decisions, and drug information: dosing, interactions, safety⁷. that is the work medical affairs already owns.

they are not outsourcing judgment. the same survey shows how physicians check what generative ai gives them. most cross-check trusted guidelines, databases, or peer-reviewed publications. many check the original source. many rely on their own clinical knowledge. almost nobody says they usually skip the check⁷. access got faster. verification is still the job.

that is the trust gap in 1 line. some physicians already say they want pharma's help reading what the model said, not another brochure⁷.

cost still keeps a lot of specialty care behind a delay or a denial. more of the diagnosis and treatment conversation lands with the primary care doctor patients can actually get in to see. in 2024, about 1 in 6 us adults delayed or went without care because of cost⁸. when money and access are not limiting, primary care is still the first place many people say they would take most health concerns⁹. tools built for that setting are catching up. a 2025 primary-care case series tested openevidence, an evidence-linked clinical ai, as support for common chronic-care decisions¹⁰. it mostly reinforced physician plans rather than replacing them. the point for pharma is the same either way. more first answers, patient and clinician, are forming outside the specialty brochure.

most brands never make the answer

if the recommendation happens in the chat, the next question is whether your evidence is even in that answer. a published cross-platform study found only about 7% of cited domains showed up on 3 or more major ai platforms¹¹. most brands are invisible on most of the models their patients already use. (the publisher sells visibility tools; the conflict is real. the low overlap still matches what teams see when they check their own names by hand.)

showing up early still beats polishing a page that never enters the first answer.

mlr is not the villain. destinations are.

so why does translating information to patients and hcps still feel so hard?

mlr is real. it exists so a claim stays on-label and a patient is not sold a fantasy. when every delay gets blamed on review, the real miss is usually earlier: nobody scoped the claim so it could survive review.

and the words that survive review still have to sound like people. verbatim medical jargon is not how patients, caregivers, or most clinicians want to be spoken to. in a mixed-methods survey of 205 adults, 91% preferred a jargon-free explanation of the same clinical information, and described the jargon version as confusing and cold¹². plain language is not dumbing down the science. it is the readability and vernacular work that makes an approved claim usable once a chatbot or a person has to say it out loud.

marketing is not off the hook either. too many programs still ship destinations for a search era people partially left: a brand site, a portal, a pdf library, a drip that assumes the reader starts on your url. the reader starts in a chat. if your verified copy is not findable there, you lose the first answer by default. write it so a model can name the source and a person can use the words.

neither function alone builds the bridge. medical affairs has the evidence clinicians check after the chat. marketing knows how to make language land. mlr decides what is allowed to be viewed by patients, caregivers, and hcps. the work is getting those 3 to treat the chatbot answer as a surface, not as a threat that arrives after the campaign is locked.

john duffield writes as the future patient after a complex care journey. he also worked in pharma marketing; the 3 lines he puts on the job come from that patient story: be in the answer; help people use ai better; show up for the emotional gap between appointments¹³. that framing comes from the inside of recovery rather than from a brand deck¹³.

one question, one cite-able answer

pick a question you would ask a chatbot about a friend, a loved one, or yourself dealing with an illness. ask it. the reply will likely sound nothing like the medical jargon on a brand site. that gap is the illustration. chatbots train on how people speak to each other, so the first answer already sounds more human than most on-label pages. the work is getting verified words into that same reply, in language a person can use.

the real work is meeting people where the answer arrives. that includes the marketer who is also a patient, and the hcp buried in visits. they are sorting through instant answers, half-understood coverage rules, and short attention, trying to make sense of something that used to wait for an appointment. that is a ux problem. build for efficiency and for where people already are, not for the old guard. the brand site can still matter. it is just rarely where understanding starts anymore.