What the obesity TA tells us about GLP-1 launches in AI.

The obesity therapeutic area in AI search is structurally unlike any other category in the May 2026 PharmaGEO public index. It has the most concentrated Share of Voice profile, the most pronounced engine-level divergence between leaders, and the most acute off-label surfacing risk of any TA measured. What follows is a category-level analysis anchored entirely on the public index — not a client case, not an anonymized program, but a pattern-level reading of what the GLP-1 AI search landscape looks like in May 2026 and what it implies for any brand launching into or competing within it.

The GLP-1 concentration problem: top-3 hold 53.9% of SOV

The obesity TA in the PharmaGEO public index is a bipolar market. On OpenAI, the top-3 brands — Wegovy (22.2%), Zepbound (19.5%), and Qsymia (12.2%) — collectively hold 53.9% of total Share of Voice. The remaining 46.1% is distributed across seven or more additional products. This is the most concentrated SOV profile in the four-TA index. Psoriasis by comparison has a top-3 SOV of only 28.2%; lung cancer, 20.9%.

The concentration pattern has a structural driver: AI engines in the obesity TA exhibit a strong category heuristic. When a user asks about weight management medications, engines name GLP-1 receptor agonists first and in depth, before reaching other mechanisms. This is a reflection of the current clinical consensus and the density of GLP-1 clinical content on the web — not a promotional outcome. But the practical effect is that brands outside the GLP-1 class, or GLP-1 brands without the top-tier evidence packages, are competing for less than half the available SOV against two entrenched leaders.

Gemini amplifies the concentration further

The engine-level picture makes the concentration problem more acute. Wegovy holds 22.2% SOV on OpenAI. On Gemini, the same brand holds 34.3% SOV — a 12.1-point premium. Gemini surfaces fewer distinct products per response than OpenAI, amplifying the winner-take-most effect. For a brand entering the obesity TA and trying to displace an entrenched GLP-1 leader, Gemini is structurally the hardest engine to gain share on: the leader's advantage is larger there than anywhere else.

This has direct budget implications. A GEO program that achieves meaningful SOV gains on OpenAI may show little movement on Gemini for the same content investment. Challenger brands entering the obesity TA need Gemini-specific content strategies — focused on current web content and press syndication, which Gemini surfaces more readily than corpus-weighted clinical literature — rather than assuming that OpenAI gains will translate.

The off-label leakage problem: a systemic AI risk

The most significant compliance signal from the obesity TA is not about the obesity-indicated brands. It is about brands that are not obesity-indicated appearing in obesity answers. The May 2026 PharmaGEO public index records measurable Answer Rate for Ozempic and Mounjaro in response to obesity queries. Perplexity's SOV chart for the obesity TA includes Ozempic at approximately 6% and Mounjaro at approximately 3%.

Both brands are approved only for Type 2 diabetes. Neither holds an obesity indication. Their appearance in obesity AI answers is a systemic behavior — AI engines connecting same-molecule branded products across indication boundaries based on the public clinical literature, which extensively discusses the GLP-1 class's effects on body weight. This is not a brand-specific error and not a promotional claim by any manufacturer. It is the engine doing what retrieval engines do: drawing on all available literature about a molecule regardless of indication-specific approval status.

But each such mention is a pharmacovigilance exposure. An HCP or patient receiving an AI answer about obesity medications that includes a T2D-only brand, unprompted, has received an off-label indication reference through a channel outside any MLR review process. This is the "AI as pharmacovigilance surface" problem, and it is most acute in the GLP-1 class because of the molecule-level overlap between obesity-indicated and T2D-indicated agents.

FDA labels lead obesity citations — and what that means for MLR

The citation source stack in the obesity TA is dominated by regulatory documents. In Perplexity's obesity answers, the top two citation sources by use count are the FDA label for Zepbound (accessdata.fda.gov, 36 citation uses, rank 1) and the FDA label for Wegovy (32 citation uses, rank 2). These labels include boxed warnings, full contraindication lists, and REMS information.

The implication for brand teams is uncomfortable but important. The AI answer layer in obesity is already delivering mandatory safety information — boxed warnings, contraindications, REMS instructions — to prescribers and patients, framed by the engine without promotional review. The engine does not distinguish between safety content and product claims. It cites what is most authoritative. In obesity, the most authoritative sources are regulatory labels.

A brand entering the obesity TA that focuses its GEO program exclusively on clinical evidence content and ignores the regulatory citation stack is optimizing for the fourth-most-cited source while leaving the top three — FDA labels and manufacturer PIs — entirely outside its strategy. Ensuring that the brand's own PI and FDA label are structured, current, and cross-linked appropriately is not a regulatory housekeeping task. It is the top GEO content priority in this TA.

AR vs SOV in obesity: the 4x compression trap

The obesity TA also provides the most striking example of the AR-vs-SOV compression problem in the May 2026 index. Wegovy holds an Answer Rate of 93.2% on OpenAI — the brand appears in nearly every relevant obesity prompt response. Its SOV is 22.2%. The AR-to-SOV ratio is approximately 4.2:1.

A brand team reporting only Answer Rate could read Wegovy's 93.2% figure and conclude that the brand has near-total dominance of AI obesity answers. The SOV figure tells a different story: even the most frequently cited brand in the most concentrated TA in the index holds only 22.2% of the conversation surface. Three to eight other brands are named alongside it in every answer where it appears.

For challenger brands entering obesity, this has a specific strategic implication. Reaching a high Answer Rate is achievable with sustained content investment, but Answer Rate alone does not translate to competitive differentiation. SOV — the share of brand-token surface area in answers that mention multiple brands — requires content that gives the engine a specific, differentiating reason to expand its description of your brand relative to others named in the same answer. Clinical subgroup data, mechanism differentiation, and cardiovascular outcomes content are the levers that move SOV; they give the engine material to write more about a specific brand rather than merely naming it alongside the category standard.

A pattern-level playbook for a metabolic launch in AI search

Reading the obesity TA's structural patterns in the May 2026 PharmaGEO public index produces a set of strategic priorities that any GLP-1 or metabolic brand team can apply. These are category-level patterns, not prescriptions for a specific program.

Priority 1: own the regulatory citation stack before the web content stack

FDA labels are the #1 and #2 citation source in obesity AI answers. A brand's full prescribing information and patient medication guide — hosted on both accessdata.fda.gov and a manufacturer-controlled PI page — are the first GEO assets, not the last. These should be current, structured, and accessible. A label that has not been updated to reflect current indication language or that lacks a clearly indexed manufacturer-hosted PI page is structurally disadvantaged before any content program begins.

Priority 2: build Gemini-specific content separately from OpenAI

Gemini concentrates SOV at the top of the obesity TA (Wegovy 34.3% vs 22.2% on OpenAI). For a challenger brand, closing the Gemini gap requires different content than closing the OpenAI gap. Gemini surfaces recent web content and press syndication more readily than corpus-indexed clinical literature. A GEO program in obesity should have separate content tracks for each engine, not a single track assumed to lift all platforms uniformly.

Priority 3: monitor off-label AI surfaces as a pharmacovigilance function

The measurable appearance of T2D-only GLP-1 brands in obesity AI answers is not a problem a brand team can eliminate — it is a systemic AI behavior driven by the molecule-level clinical literature. But it is a surface that pharmacovigilance and MLR functions need to monitor continuously, because it constitutes an off-label indication channel operating outside regulatory oversight. The monitoring protocol should be structured: defined prompts, defined engines, defined frequency, and a documented escalation path when off-label brand mentions exceed a threshold.

Priority 4: invest in cardiovascular outcomes content as a SOV differentiator

In the obesity TA, cardiovascular outcomes data is the primary dimension along which brands are differentiated in AI answers beyond weight loss efficacy. The SELECT trial for semaglutide and SURMOUNT-MMO for tirzepatide have created a citation-rich outcomes environment that engines draw on in clinical queries. A brand without outcomes data in cardiovascular endpoints will be described primarily on weight reduction efficacy — a metric on which multiple brands can make similar claims. Outcomes data creates the content differentiation that allows an engine to say something specific and distinct about a brand, not just name it alongside the category leader.

Priority 5: size your expectations against structural SOV ceilings

The top-3 brands in obesity hold 53.9% of SOV, leaving 46.1% for seven or more other brands. The structural ceiling for a new entrant's SOV — assuming it displaces no one and simply gains share from the tail — is in the single digits initially. Modeling GEO success against realistic SOV ranges based on TA concentration — not against the leader's figure — is the only way to set achievable targets. An obesity brand projecting 30% SOV on a twelve-month timeline based on the category leader's current position is measuring against a ceiling that required years of clinical content investment to build.

The obesity TA in AI search is the clearest illustration in the May 2026 PharmaGEO public index of how GEO, pharmacovigilance, regulatory affairs, and brand strategy intersect. It is not a marketing problem with a content solution. It is a cross-functional measurement and monitoring challenge that requires all four functions operating from the same data set.

Want a real audit on your brand? Request a sample report or get the full PharmaGEO Playbook.