The first public ranking of pharma brands in AI answers.

How the PharmaGEO Index is computed

For each therapeutic area, the PharmaGEO Index runs a standardised set of 80 prompts that real patients, HCPs, payers and competitive analysts ask about brands. Each prompt is executed across the configured LLM (ChatGPT, Gemini, Perplexity or Claude), in the configured language (US English, UK English, French, Spanish), and the resulting AI answer is parsed for three core GEO and AEO metrics: appearance rate, share of voice, and top sources.

The index is the same instrument we ship inside PharmaGEO for paying brand teams — here we expose the result for a public selection of brands, indications and engines, so that the pharma industry can read AI visibility at a glance. Full methodology on how share of voice is measured in LLMs is in the research hub.

The three metrics that define pharma AI visibility

• Appearance Rate (AR) — the share of prompts in which the brand is mentioned at all by the LLM. AR answers the simplest question: does this brand exist in the AI's worldview for this indication? AR is the floor metric for any GEO program.
• Share of Voice (SOV) — the share of total brand mentions a single brand captures across the prompt set. SOV measures dominance, not presence. A brand can have 93% AR but only 14% SOV — it is named everywhere, but never first. Read the deep dive on AR vs SOV.
• Top Sources — the most-cited sources the LLM relies on when constructing the answer. The source mix changes the brand-relevance of the answer, the safety information surfaced and the competitive framing. See how the source stack changes by TA.

Which LLMs the Index tracks

The public Index currently exposes results for OpenAI ChatGPT, Google Gemini, Perplexity and Anthropic Claude. Inside the PharmaGEO paid platform, the same instrument runs across eight LLMs (adding Mistral, Grok, Copilot and DeepSeek), three language markets and twelve therapeutic areas, with month-over-month tracking and engine-level alerts. The same brand can rank #4 in one engine and #8 in another — read the engine-divergence study.

Therapeutic-area coverage

The public Index currently covers five therapeutic areas: atopic dermatitis, obesity, lung cancer (NSCLC), menopause vasomotor symptoms and psoriasis. Each TA carries 80 prompts that span the patient-to-prescriber decision funnel: disease information, treatment options, safety, comparator, payer/access, and quality of life. Source stacks differ markedly by TA — in oncology, NCCN guidelines dominate Perplexity citations; in obesity, FDA labels lead; in psoriasis, specialist disease sites take the largest share. Full TA citation analysis.

How to read the leaderboard

Use Appearance Rate first to see which brands the engine knows about at all. Then read Share of Voice to see who dominates the conversation — a high AR with low SOV is a brand that is named but never first, a competitive vulnerability. Cross-check with Top Sources: a brand that depends on one source archetype (only its own .com, or only one society site) is fragile. The brands with durable AI visibility tend to be cited via consensus across independent, high-authority sources. See the eight findings from the May 2026 index.