Your next publication will be read by AI before it reaches a doctor.
Is it ready?
A structured methodology for assessing and improving how scientific publications perform in AI systems.
Talk to us about GEOHow HCPs access scientific evidence has shifted
AI tools — ChatGPT, Perplexity, Gemini, Claude — are now a primary entry point for many healthcare professionals looking for answers about clinical data, treatment outcomes, and emerging evidence. That shift is accelerating.
But those AI systems don't read publications the way a human does. They work from extracted text, parsed into chunks, embedded as vectors, and retrieved against a query. The quality of that extraction — and the structure of the original content — has a direct effect on whether your key scientific messages appear in an AI-generated response, and how they're represented when they do.
Most publications are still built, reviewed, and approved without considering the implications of AI discoverability.
What a GEO assessment includes
GEO works at three levels — from the strategic narrative that frames your programme, through the individual assets that carry your evidence, to the competitive landscape your publications operate in.
Scientific platform and narrative
Before looking at individual assets, we review the scientific platform and core narrative — how key messages are structured, what should be leading versus supporting, and how the overall evidence story reads to AI systems. Optimising at the narrative level creates the foundation everything downstream depends on.
- Scientific platform and message hierarchy review
- AI visibility audit of platform-level documents
- Message clarity and differentiation assessment
- Recommendations for narrative structure and framing
Asset-level assessment
Individual manuscripts and publications are evaluated against GEO criteria: how well AI systems can read, parse, and retrieve the content — and whether they represent key claims accurately and completely. This is where specific, actionable recommendations are generated for your publications team.
- Encoding and text extraction quality check
- Structural analysis against GEO optimisation criteria
- Semantic similarity scoring against a pharma-specific prompt library
- AI visibility testing across ChatGPT, Claude, Perplexity, and Gemini
- Prioritised recommendations report
Competitive assessment
AI systems don't evaluate publications in isolation — they rank and choose between competing sources in real time. A competitive assessment maps how your content performs relative to competitors in shared disease areas, and identifies where the structural or narrative gaps are.
- Comparative AI visibility against competitor publications
- Benchmarking across shared indication and disease area queries
- Identification of structural and narrative gaps driving competitor advantage
- Strategic recommendations for closing competitive gaps
A few examples
The corrupted p-value
A key efficacy endpoint — p<0.001, a highly significant result — was extracted by AI systems as p5.001. A single encoding corruption transformed a landmark finding into an unintelligible value. The paper looked correct in the PDF. It had passed every standard QC check. But for every AI system that has since tried to summarise it, that endpoint was broken.
Do you know how many of your communications already have this issue?
Structure, not data
Two publications from the same disease area, published in the same high-impact journal. In AI systems across ChatGPT, Claude, Perplexity, and Gemini, one was consistently cited as the more authoritative answer — including to questions about the other company's drug. The divergence came down to structure. Not data quality. Not journal. Not timing.
Is your content currently the source of authority for your own scientific statements?
What GEO assesses
GEO assessments cover four domains — each addressing a distinct way a publication can fail to perform in AI systems.
Publication structure and language
Whether the manuscript is written in a way that allows AI systems to retrieve individual claims accurately. How a document is structured, how precisely language is used, and how self-sufficient individual sections are all affect retrievability.
Technical encoding quality
Text extraction errors are more common than most teams realise. We've identified systematic encoding issues in published work that passed every standard QC check — because no one was checking for AI readers. These errors invisibly distort how data are read.
Semantic retrievability
We assess, and quantifiably score, how well the key claims in a manuscript align with the kinds of questions HCPs actually ask. A publication that structures its data poorly will score worse than a structurally similar competitor — even in the same journal.
AI visibility testing
We test how publications perform across the major AI systems, and what you can do about it — how often they're cited, how they're represented in generated responses, and how that compares to competitor publications in the same clinical area.
Who this is for
Pharma medical affairs and publications teams
Working on programmes where AI visibility of the evidence base matters. Particularly relevant for:
- Programmes with active HCP-facing evidence dissemination
- Teams preparing for or following up major congress outputs
- Publications with competitive overlap where AI-generated responses matter
- Programmes where a single manuscript underpins a key clinical narrative
MedComms agencies
Responding to client demand for AI visibility analysis, or looking to add a credible GEO offering to their portfolio.
- Bring GEO to clients as a standalone service
- Integrate into existing publication planning engagements
- Differentiate your agency offer as AI questions grow from clients
GEO is not SEO
Traditional SEO is about ranking in search engines — keyword optimisation, metadata, backlinks. GEO is about something different: whether the content of a publication is structured so that AI systems can retrieve individual claims accurately, represent key messages faithfully, and cite the publication in generated responses.
The two are related at the edges but distinct in method and purpose. GEO is a scientific communications problem, not a search marketing problem.
Find out how your publications perform
The simplest starting point is a single manuscript assessment. It's quick and gives you a clear picture of where you stand — and what the highest-value actions are.
Talk to us about GEO