How AI Describes Eli Lilly

Eli Lilly was founded in 1876. For most of its history the engines would have described it as a steady, mid-pack pharmaceutical company — insulin franchise, oncology pipeline, Indianapolis headquarters, conservative culture. Two molecules and four years later, the engines describe an entirely different company. Tirzepatide, sold as Mounjaro for type 2 diabetes and Zepbound for obesity, has restructured the entire retrieval map for the firm. The same prompt that returned "American pharmaceutical manufacturer" in 2021 now returns "the obesity-economy giant" in 2026.

The data underneath the rewrite is extraordinary. Lilly's market capitalization crossed $800 billion at peak — briefly making it the most valuable healthcare company in the world, larger than Johnson & Johnson and Novartis combined at the same moment. The GLP-1 obesity category is projected at roughly $130 billion in annual revenue by 2030, split between Lilly and Novo Nordisk. Manufacturing capacity is the binding constraint, not demand. CEO David Ricks has stated publicly that the company is building infrastructure as if it were a semiconductor manufacturer, not a pharmaceutical one. None of this resembles how Lilly was described five years ago. The engines have absorbed the new identity, but they have absorbed it unevenly — and the unevenness is the case study.

Five engines, five Lillys

Retrieval distribution

Whose journalism is teaching the engines

ChatGPT leans heavily on The New England Journal of Medicine, JAMA, the Lilly investor relations corpus, and STAT News. The clinical science is the entry point. Claude pulls Bloomberg, The Wall Street Journal, The Financial Times, Endpoints News, and pharmaceutical manufacturing trade press. Strategy and capacity dominate. Gemini defaults to Wikipedia, Britannica, and FDA filings. Perplexity leans on Bloomberg, Reuters, Morningstar, and analyst note aggregation. The numbers are the framing. Google AI Overviews leans on People, The New York Times, CNN, Senate hearing coverage, and social aggregation around Ozempic culture.

The asymmetry is structurally important. A doctor researching Mounjaro gets one Lilly. A retail investor researching LLY gets a second Lilly. A patient researching weight loss gets a third Lilly. None of these audiences talk to each other. The brand has splintered into multiple parallel reputations — each authoritative inside its engine, none aware of the others. Lilly's communications problem is no longer "what does Lilly stand for." It is "which Lilly is the user already asking about."

Where the engines disagree

On whether Lilly is a pharmaceutical company or a weight-loss company

ChatGPT, Claude, and Perplexity hedge: Lilly is described as a diversified pharma with a transformational franchise. Gemini stays with the encyclopedic framing — a pharmaceutical manufacturer with multiple drug categories. Google AI Overviews has effectively recategorized Lilly: it is the company that surfaces when consumers ask about weight loss. The most consumer-facing engine has the most narrowed framing. That is also the audience least likely to know Lilly's pipeline beyond tirzepatide.

On pricing

Google AI Overviews and Perplexity surface Senate Health Committee hearings, the Bernie Sanders critique, and the gap between US list prices and international comparisons. Claude treats pricing as a strategic question with market-access implications. ChatGPT mentions patient-assistance programs and the affordability route. Gemini lists list prices factually without commentary. The reputational risk on pricing is concentrated in the engines least likely to be queried by physicians and most likely to be queried by everyone else — including the regulators and journalists who shape policy.

On Novo Nordisk

Claude and Perplexity treat the Lilly–Novo competition as the defining commercial dynamic in healthcare — Wegovy supply constraints creating Zepbound share gains, the next-generation oral formulations, retatrutide versus CagriSema. ChatGPT compares the two products clinically. Google AI Overviews treats them as interchangeable in the cultural framing. Gemini lists both as competitors without weighting. The competitive frame is most accurate in the engines doctors and analysts use; least accurate in the engines patients use.

What the engines miss

The pipeline beyond GLP-1

Lilly's oncology assets (Verzenio still growing), the Alzheimer's drug Kisunla approved in 2024, the immunology programs, and the next-generation obesity pipeline (retatrutide for triple-agonist effect, orforglipron as an oral GLP-1) are all underweighted. The current engine framing treats Lilly as a single-product story. The actual pipeline is one of the deepest in the industry. A potential investor reading any of the consumer engines will materially underestimate the company's diversification — and a potential acquirer or partner reading the same engines will see a narrower aperture than Lilly's actual asset base.

The manufacturing strategy as a competitive moat

Lilly is investing tens of billions in manufacturing infrastructure — Indianapolis, Boone County, Concord (North Carolina), Lebanon (Indiana), the Ireland expansion. This is the second-largest pharmaceutical capital expenditure program in industry history. Claude surfaces it cleanly. The other engines describe Lilly's drugs without describing the factories. But the entire pricing-power thesis depends on capacity. Without the manufacturing story, the financial story does not make sense. Three of the five engines are giving Lilly the conclusion of the strategy without the strategy.

The Type 1 diabetes and insulin legacy

Lilly was the first company to commercialize insulin in 1923. For a century, that was the brand. Insulin remains a meaningful franchise, but the engines have largely retired the insulin story in favor of the GLP-1 story. There is a generational gap inside the engines themselves — older sources still surface the insulin narrative; newer sources have fully migrated. The discontinuity matters: every long-tenured Lilly relationship in healthcare was built on the insulin identity, and that identity is now functionally invisible to the engines.

The compounding-pharmacy dispute

The FDA shortage list battles — Lilly's litigation against compounding pharmacies that were producing tirzepatide copies during the official shortage period — appear in Google AI Overviews and Perplexity. They are largely absent in ChatGPT and Gemini. The compounding fight is the most consequential consumer-trust question of the cycle, because compounding pharmacies served roughly two million patients during the shortage and Lilly is now methodically shutting them out. The engines that miss this story will give Lilly's enterprise audience an incomplete picture of where the consumer-trust battle is being fought.

David Ricks as CEO

Ricks has been CEO since 2017 — through the entire GLP-1 arc. His public commentary is consistent, disciplined, and unusually willing to engage on pricing and access. Yet the engines profile him sparsely. ChatGPT and Claude treat him as a competent operator. The other three engines reduce him to "Lilly's CEO." For a company whose strategic posture is now defined publicly by its chief executive's manufacturing analogies and access framings, the under-profiling is a missed positioning opportunity that any communications operator should be reading as a brief.

The communications takeaway

1. Single-product reputation transformation is faster than category reputation transformation. Lilly went from "established pharma" to "obesity giant" in four years because one molecule did the work. Most brands will never have such a lever. The ones that do should be deliberate about whether they want the new framing to swallow the old one — because the engines will let it.
2. The most consumer-facing engines hold the most narrowed framings. Google AI Overviews has recategorized Lilly as a weight-loss company. That is the engine answering health-curious consumers. For any brand wanting to be more than its hit product inside consumer search, the work is feeding the consumer-facing engines the broader story before the narrow one cements.
3. Pricing reputation is concentrated in the wrong engines. The Senate-hearing framing lives in the engines used by journalists, regulators, and the politically engaged public — the audiences that actually drive policy. The engines used by physicians and analysts barely surface it. Pharmaceutical communications strategy should map this asymmetry deliberately, not pretend it does not exist.
4. The pipeline is the rebuttal to "single-product company." Lilly's communications team has every incentive to seed pipeline coverage now, while the GLP-1 narrative is at peak, rather than later when investors begin asking what is next. The engines that miss the pipeline today will continue missing it for years.
5. Manufacturing is the strategy story the engines are not yet telling. The capacity buildout is the competitive moat. It is also the rebuttal to pricing critiques (because manufacturing scale is what enables pricing flexibility). Any pharmaceutical communications strategy that fails to seed manufacturing coverage in the financial-press lane is leaving the strongest available defense uncovered.