Why direct-to-consumer ad spend stopped predicting where the AI engines land — and what pharmaceutical companies should do about it.

The U.S. pharmaceutical industry spent an estimated $8 billion on direct-to-consumer advertising in 2024. That is the largest DTC ad category in America. Bigger than auto. Bigger than financial services. The ads run during the network news, during sports, during prestige drama — built to put a drug name in front of a patient so the patient asks the doctor about it.

The campaign worked for thirty years. It is working less now.

New 5W research measured how AI answer engines — ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews — describe and rank the 25 largest pharmaceutical companies. We ran more than sixty patient and consumer prompts, five times per engine, in clean sessions. The finding is simple. Ad spend does not predict citation share. What patients self-research does.

Lilly and Novo own the answer

Eli Lilly leads the citation surface at an estimated 12.5% share. Novo Nordisk sits at 11.5%. Together, two companies hold roughly a quarter of all corporate citations across the pharmaceutical category. Their drugs — Ozempic, Wegovy, Mounjaro, and Zepbound — generated roughly $70 billion in combined 2025 sales, and more importantly, became cultural events. Ozempic shows up in late-night monologues, New York Times features, and Wikipedia entries the AI engines retrieve all day long.

Now look at Merck. Merck makes Keytruda, the best-selling prescription drug in the world. Merck ranks fifth in citations. Oncology drugs are prescribed by oncologists. Patients do not type Keytruda into a chatbot the way they type Ozempic. Revenue is one scoreboard. The answer engine is a different one.

AbbVie is the cleanest test case

AbbVie has been among the largest direct-to-consumer advertisers in U.S. healthcare for a decade — Humira, Skyrizi, Rinvoq — running heavy national TV and digital schedules. AbbVie ranks sixth in citation share at an estimated 5.0%. Below Pfizer, whose lift is the residual recognition dividend from the COVID-19 vaccine. Below Johnson & Johnson, which carries diversified-conglomerate weight. AbbVie is paying premium rates on prime-time inventory. The engines are not paying it back at the same multiple.

That is not an indictment of DTC. It is a measurement of where the channel stops compounding. A spot makes a drug name familiar. Familiarity by itself does not produce a retrievable record. What produces the record is the editorial coverage, the trial data, the Wikipedia entry, the analyst note, the pharmacist explainer on Drugs.com, the STAT investigation, the Reuters earnings story. The spot can seed all of that. It does not substitute for it.

What is actually happening in the chatbox

More than a third of U.S. consumers now begin product research with an AI engine, not Google. In healthcare, the share is rising fast. Patients open ChatGPT and ask what is the difference between Ozempic and Mounjaro, which works better for sleep apnea, what the side-effect profile looks like, whether their insurance is likely to cover it. The engine returns a paragraph. That paragraph names companies. Lilly and Novo show up by name in roughly a quarter of all pharma answers we tested.

By the time the patient gets to the doctor, the brand conversation has already happened. It happened in the chatbox. The doctor is not the first stop anymore. The doctor is the third stop, after the chatbot and the friend who tried it.

The playbook for pharma communications teams

  • Audit corporate citation share quarterly. Trial readouts, pricing news, and patent events move the surface faster than the budget cycle.
  • Connect the drug brand to the company name in every retrievable surface. If the public record says “the maker of Ozempic” without naming Novo Nordisk, you are leaving citation share on the table.
  • Earn the business and health editorial record. Reuters, Bloomberg, STAT, and Fierce Pharma feed pharma citations disproportionately.
  • Keep the reference layer accurate. Company filings, Wikipedia, and Drugs.com are heavily retrieved. Errors there propagate.
  • Treat trial readouts and approvals as citation events. A New England Journal of Medicine paper is a citation engine for years.
  • Engage pricing and access coverage directly. AI answers surface pricing context inside corporate queries. Concede that framing and the engines write it for you.
  • Stop reading DTC TV spend as a citation strategy. It is a channel, not the infrastructure.

The bigger picture

Pharma built thirty years of brand awareness on television. The industry is still buying that channel at scale. The buyer moved. The buyer is in the chatbox now, asking the AI engines a question that used to be the first thing said in the exam room.

Citation share is the new market share. Eli Lilly and Novo Nordisk understood it without naming it. They built the drugs people talk about, fed the editorial and reference record around those drugs, and let the engines do the rest. Merck makes the best-selling drug on earth and ranks fifth in the answer. That gap is not closing on its own. Closing it is a communications discipline. We call it AI Communications. AbbVie, Pfizer, Bristol Myers Squibb, and every other heavy DTC spender have a choice to make about whether they fund the channel or the infrastructure.

AI Communications is a mix of journalism, psychology, and engineering.

The full Pharma / Rx AI Visibility Index 2026 is live at 5wpr.com — no paywall, no PDF, no email gate.

Important framing: this Index measures AI citation share for corporate communications and marketing strategy purposes only. It is not medical advice. Medication decisions must be made with a qualified physician or pharmacist.

Ronn Torossian is the founder and chairman of 5W AI Communications, the AI Communications Firm. He is the publisher of Everything-PR and the author of two best-selling editions of For Immediate Release.