The Future of The Brand.com

Here’s a question the industry does not yet have a settled answer to.

A patient asks an AI about a therapy and gets a clear, well-organized response. Does that end it? Or does the answer prompt them to check — to look for the source, confirm what they heard, find the thing the model couldn’t give them?

Both are happening. What nobody can say with confidence yet is the ratio, or how it shifts by therapeutic area, by question type, by how much is at stake for the person asking. That uncertainty is the whole strategic problem, and it’s what six executives from pharma, agency, procurement, and technology spent the second half of a recent DHC Group roundtable working through.

What is established is narrower, and it’s enough to act on: the path in has changed.

The entry point moved upstream

Chris Lee of Pfizer put it plainly. The way into brand.com increasingly runs through whichever LLM a patient happens to be using — and those starting points are far more dispersed than brands are used to managing. His conclusion isn’t that the destination stopped mattering. It’s that the source of truth has to exist wherever the journey begins, so that accurate information is present before someone is several steps in.

Stephanie Schulman of Publicis Health framed what follows as a split in the site’s job. Brand.com now serves a machine audience and a human audience at the same time, and she expects the balance between them to keep shifting. The machine-facing role — being the record of truth about a therapy — grows regardless of what the click data eventually shows.

The human-facing role is the one still in motion. Her read: the people who do arrive increasingly show up having already engaged with AI, which means higher intent and higher trust needs. They’re there for what the model couldn’t finish — financial assistance, enrollment, ongoing support.

There’s some evidence on verification

Christine Maguire of Fullspan Health brought the closest thing the panel had to data on the question. Citation drives trust, and trust drives action: among users who come to validate and verify after encountering a cited brand, Fullspan is seeing a 2.2x conversion lift.

That’s worth sitting with, because it describes the behavior in question. Those people did continue past the answer. And the ones who did so from a cited source converted at more than double the rate.

Which reframes the strategic question away from will they still visit and toward something more useful: what determines whether they do, and what happens when they land.

Build for the visit you’re likely to get

If the arriving visitor is fewer in number but further along, the site has to be worth their trip.

Dan Haller of Heartbeat drew the line around value. Financial resources, and access to educators and nurses that fill real gaps in traditional care, give someone a reason to come. Broad disease education already covered by a hundred other sources doesn’t.

Lee described that logic applied on Eliquis.com — a savings and support experience streamlined around the high-value actions increasingly driven from AI search, optimized for AI natively rather than keyword search alone, with media pointed at those actions rather than general awareness. Delivering people where they need to go, rather than asking them to browse and discover.

Measure the answer and what follows it

Schulman proposed four measures for the machine-facing job: presence, source visibility, answer fidelity, and actionability — tracked across models, and specifically within the questions patients most often ask about a therapy.

She also reframed share of voice itself. Who wants to own an incorrect answer? The more meaningful measure is share of accurate, balanced, actionable answers.

Estefania García Zapata of Takeda added the near-term stake: without visibility, someone else is defining the story a patient receives. But visibility alone isn’t the goal — it needs accuracy, the right sentiment, enough context, and a clear next step.

Lee put the ultimate test furthest downstream. Upstream metrics still matter, but the question is whether behavior actually changes. He was candid that it’s hard to measure, and clear that it’s the point.

Barbara Salami of Decyd offered a way in that doesn’t require solving attribution first: map the patient journey end to end, count who enters each stage, see how long they take to move, and find where they fall away. That’s also where you’d see the verification behavior — or its absence — in your own data rather than in anyone’s projection.

Which may be the most useful takeaway available right now. The industry doesn’t have a shared answer to what happens after the AI answers. Individual brands can start finding out for themselves.