GEO for medicines and supplements: measuring AI recommendations and factual accuracy

For medicines and supplements, effective GEO requires appropriate recommendations and verifiable facts about products.

Pharmaceutical AI answers can recognise a brand and still make a material mistake about its ingredient, indications or limitations. As a GEO researcher at GAEO.ru, I review these questions; the commercial scope appears under GEO for pharmaceutical brands.

This is a framework for AI-answer analysis using an existing anonymized August 2026 audit and official sources. It contains no patient treatment recommendations. Medical claims require qualified review before client publication.

A separate dietary supplement mini-audit from October 27, 2026, added 24 complete answers across 4 questions and 6 AI interfaces. The prior medicine study included 120 answers. These are independent baselines covering different products, dates and question sets. Company names, competing brands, store URLs and unique product specifications are omitted.

1. Questions for medicines, supplements and medical devices

The following are illustrative research categories, not verbatim private client prompts. They allow researchers to compare like-for-like answers and verify the right source for each claim.

Medicines, devices and dietary supplements should be treated as separate regulated product categories.

Ingredient comparisons and branded reputation inquiries must be evaluated separately. A supplement product needs evidence specific to the item, including its serving units.

2. Regulatory records, supplement labels and medical sources

For medicines in Russia, consult the State Register of Medicines and the currently approved labeling. The clinical recommendations portal provides a distinct type of evidence with its own medical scope and revision history. The Roszdravnadzor register supports device-identity and intended-use checks.

Other jurisdictions require their own primary registration sources. EMA's medicine finder covers centrally evaluated EU medicines; nationally authorised products require national-register checks.

Dietary supplements require food-product registration and labeling evidence. Rospotrebnadzor explains the product category. NIH distinguishes elemental magnesium from the mass of a magnesium-containing compound. For omega-3, check EPA and DHA on a consistent serving basis: NIH omega-3 guidance. These resources do not endorse individual brands.

3. Trade names, INN, citations and active recommendations

Code literal trade-name presence, INN mentions and active contextual recommendations as distinct outcomes. A description of an ingredient does not establish recognition of its brand. Log citations and any material factual concerns separately.

Unbranded BMR equals answers with the brand's literal trade name divided by all obtained answers to the unchanged unbranded prompt set. Branded questions are excluded from that denominator and used for factual QA. See the BMR methodology.

The supplement brand was correctly linked to its official store in 6 of 6 answers to a branded reputation question. Only 2 of 18 answers to unbranded purchase questions named it. Literal mentions, store citations, explicit ranking positions and factual accuracy require different fields.

4. Study A: a medicine, 120 AI responses

The previously published GAEO pharmaceutical audit ran on August 6–9, 2026: 20 fixed questions across 6 interfaces and 120 complete responses. Fifteen unbranded questions produced 90 answers; 5 branded questions produced 30 additional responses used to check accuracy.

The diagnostic record logged 16 material risks or inaccuracies across 14 unique answers, including 2 related-information concerns. The medicine and manufacturer remain anonymous. There is no documented comparable follow-up or demonstrated GEO growth in the available materials.

5. Study B: dietary supplements, 24 AI responses

On October 27, 2026, all 24 expected answers were obtained: 4 exact questions across 6 AI interfaces. Three unbranded questions concerned magnesium with B6, omega-3 and domestic supplement manufacturers. The fourth branded question concerned reputation and the official store.

Unbranded visibility was 2/18 = 11.1%. Recognizing a named brand does not guarantee inclusion in an unbranded shopping recommendation. The magnesium error was verified against product information. Exact ingredient amounts are intentionally omitted to preserve anonymity.

This was a one-time sample of 4 prompts, with one response per question and interface. No repeated measurement or verified sales attribution is available. The 11.1% supplement figure cannot be directly compared with the 17.8% medicine result as a measure of brand performance. Products, questions, dates and interface panels differed.

6. Medical safety and supplement ingredient errors

Every suspected error needs a preserved answer, source, date, issue priority, qualified reviewer and remediation status. Check indications, contraindications, patient groups and dispensing restrictions first. The client's medical professionals approve all factual corrections.

Potential adverse events must follow the organisation's pharmacovigilance procedures. WHO defines pharmacovigilance through detecting, assessing, understanding and preventing adverse reactions and related problems.

A supplement answer also contained an important factual error: the mass of a magnesium compound was presented as the amount of elemental magnesium. Correct analysis needs the actual label, the unit of measurement and the quantity per capsule or serving. This is a product-accuracy issue, not personalized dosing advice.

7. Separate advertising rules for medicines and supplements

Article 24 of Russian Advertising Law No. 38-FZ sets specific restrictions for medicine and medical-device promotion. Advertising must not guarantee the absence of adverse effects, safety or effectiveness. Prescription products have channel restrictions, and relevant advertising must comply with applicable warning requirements.

Legal status depends on content and distribution. Educational, professional and promotional communications require appropriate legal and medical review. GEO does not override these requirements.

Dietary supplement advertising is covered by Article 25 of Russian Advertising Law No. 38-FZ. Supplement advertising must not present the product as a medicine, describe cases of cure or use research as a direct recommendation for taking a supplement. A notice that the product is not a medicine is required in covered advertising formats.

8. Product facts, quality evidence and external sources

Connect the provider's verified identity, trade name, INN, authorisation and approved product documentation. Separate corporate news, trials and clinical recommendations by date and provenance. Corrections need assigned responsibility.

Schema.org should represent the content actually visible. Google Search Central places particular emphasis on accuracy and trust for health information. GAEO prepares technical research; qualified client professionals verify medical claims.

For supplements, document food-product registration, the current label, actual manufacturer and ingredient quantities per capsule and serving. Independent reviews and test results have different evidentiary roles. A cited review does not prove it caused an AI recommendation. An unmentioned quality protocol also cannot be assumed absent. Full AI answers may contain source links omitted from structured citation fields.

9. Website updates and repeatable GEO measurements

1. Fixed branded and unbranded question inventory with intended audience.
2. Complete responses, observation dates and original source links.
3. Separate mention, recommendation, factual-error and missing-response indicators.
4. Risk register with severity and qualified medical reviewer.
5. Action plan for appropriately approved website and external content changes.
6. Like-for-like follow-up measurements after implemented corrections.

New products, languages and markets need their own baselines. An increase in BMR alone does not establish a better or safer pharmaceutical GEO outcome.

Maintain separate historical prompt panels for medicines and supplements. Follow-up monitoring needs the same questions, interfaces and mention-counting rules. Medical accuracy and the appropriateness of product recommendations are separate outcomes; conversions require real business data.

Frequently asked questions about pharmaceutical GEO

Author: Alexey Yakovlev, independent GEO/AEO expert. This is a methodology for researching AI answers, not treatment advice. See pharmaceutical GEO services and other strategies.