Methodology
Measured like research. Read like a benchmark.
A systematic, repeatable measurement of how AI represents your brand, configured to your market reality and traceable back to every answer.
Theralexis measures and interprets what AI says, then prioritises the actions to improve your AI representation.
72AI Representation Score
Over time
- Baseline: 61
- Re-measure: 66
- Latest: 72
+11 since baseline
- Direct competitorsRank among selected peers
- #2 of 6
- Therapeutic areaGap to the TA benchmark
- +8 vs benchmark
- Launch excellenceQuartile among peer launches
- Top quartile
- Industry benchmarkPercentile across pharma
- 67th percentile
Configured, not generic
Business priorities, competitors, stakeholder personas, markets, languages and key messages define the scope. Never a generic AI benchmark.
Observed, not asserted
Every finding comes from AI responses collected and evaluated first-hand, under consistent conditions, with variation measured rather than removed.
Built to be acted on
Results are benchmarked, explained and turned into prioritised actions for the teams that own them.
01Align Mirror your market reality
Same therapeutic area. Not the same measurement.
What AI tells a cardiologist in one country is not what it tells a GP or a patient in another. Each stakeholder, market, language and AI system can result in a different answer, so each measurement mirrors the market a brand team actually works in.
| Configured on | Scope AA global brand team | Scope BA country team preparing a launch |
|---|---|---|
| Therapeutic area | The same therapeutic area | |
| Stakeholder personas | Cardiologists | Payers and patient profiles |
| Markets and languages | Comparing countries and regions | One market, in its own language |
| Competitors | Established direct competitors | Current and emerging therapies |
| Questions | Competitive positioning and treatment choice | Disease understanding and patient journey |
| Key messages | Global brand positioning | Disease education messages |
Different use cases call for a modular system that can evolve across the product lifecycle.A generic GEO setup cannot optimally serve every business case.
02Benchmark See where you lead, lag or are missing
Benchmark your brand from every angle.
How AI represents your brand, benchmarked against competitors, your strategy, your therapeutic area, peer launches and the wider industry.
AI Visibility measures whether, and where, your brand appears. AI Representation measures how it is positioned, compared and described.
View measure definitions
| Measure | The question it answers | What you see |
|---|---|---|
| AI VisibilityThe baseline: do you appear, and where? | ||
| Visibility | Is your brand mentioned in the answer? | The share of answers that mention it, for the brand and for its molecule. |
| Share of voice | How much of the conversation does it take? | Its share of all mentions, including repeated mentions within one answer. |
| Prominence | When several treatments are named, is yours named first? | The share of answers in which your brand is the first of the tracked treatments named. |
| AI RepresentationThe competitive measurement: how are you represented when you appear? | ||
| Brand Representation | ||
| Treatment positioning | Where in the treatment pathway does AI place you? | First-line, later-line or not positioned, per stakeholder and market. |
| Preference | When treatments are compared, which does AI prefer? | How often AI prefers your brand when it compares options, and the reasons it gives. |
| Attribute ownership | Which attributes does AI associate with you rather than with competitors? | Efficacy, safety and tolerability, adherence, convenience, mechanism: the attributes agreed with you. |
| Strategic Alignment | ||
| Key messages | Are your key messages present in the answer? | Present, partly present or absent, message by message. |
| Therapeutic Accuracy | ||
| Evidence accuracy | Is the scientific evidence represented accurately? | Each answer read against agreed reference standards, response by response. |
| Source Intelligence | ||
| Sources and entities | Which sources and entities are associated with the answer? | The sources cited, the societies, KOLs, organisations and domains named, and the brands and claims they appear alongside. |
Every measure is reported per stakeholder, market and AI system, against your direct competitors and your own trend. Attributes, messages and comparisons are agreed with you at the start, so the scorecard measures what your strategy actually says.
Traceable, end to end
From a number on the scorecard to the answers behind it, and the action it calls for.
No finding is a black box. Every result can be followed back to the responses and conditions it came from.
IllustrativeAn invented example to show the method. Not client data and not a measurement.
03Deep dive Understand what sits behind the numbers
Scorecard finding
Associated with tolerability
29%vs 57% for the leading competitor
Share of answers associating each brand with tolerability. Cardiologists · Germany · four AI systems.
The responses
Read the answers, not a summary
- DatePersonaMarketAI system
- DatePersonaMarketAI system
- DatePersonaMarketAI system
Each answer that associated tolerability with the competitor rather than with you, kept with the date, persona, market and AI system it was observed under.
The sources
Which sources are associated with the answer
- Society guidelineCompetitor
- Medical portalCompetitor
- Medical portalCompetitor
- KOL reviewCompetitor
- Your approved evidenceNot cited
A society guideline, two medical portals and a KOL review are cited alongside the competitor's long-term tolerability claims. Your approved evidence is not cited.
04Act Know what to change, and who owns it
Prioritised action
Who, where, what
- Who
- Medical Affairs + Brand
- Where
- Owned medical content; relevant external channels
- What
- Assess whether approved long-term tolerability evidence can close the identified representation gap
Re-measured in the next cycle, which begins again from Align
Existing approved content first; a genuine gap is identified only where necessary. Your teams act through their own channels and approved content. Theralexis does not create, approve or publish content.
How this becomes a solutionWhy you can trust the findings
AI answers vary. That variation is not noise. It is part of what we measure.
Variability is measured and reported, not averaged out of sight. Five realities of AI systems, and how the measurement is built around each.
The five realities, and our approach to each
Reality
Different AI systems, and the same system asked twice, can answer the same question differently.
Our approach
Responses are collected repeatedly under consistent conditions. Variation is measured, not removed, and a gap between brands is read against it.
Reality
It matters who asks. A cardiologist in Germany can get a different answer than an endocrinologist in Spain, or a patient.
Our approach
Personas built with you to resemble the HCPs, payers and, where relevant, patients you engage are assessed consistently across the AI systems in scope.
Reality
Representation is local. Answers differ by country and language.
Our approach
Each market is measured independently, in its own language, and compared.
Reality
AI systems change, and so can how they represent a brand.
Our approach
Every response keeps its date and conditions, so repeated measurements under the same protocol can be compared.
Reality
A citation shows which source an answer draws on, not how the answer represents your brand.
Our approach
Sources are captured. Representation is scored on what each answer actually says, and stays traceable to it.
Theralexis measures and interprets how AI systems represent your brand; it does not control, guarantee or manipulate what they say.
Data and confidentiality
No client datasets needed. Nothing of yours made public.
This page describes what is measured and why the findings hold. The configuration behind a given assessment, its questions, personas and scoring, stays confidential, for the same reason your results do.
No client datasets required
Beyond agreeing the scope with you, Theralexis needs no customer, HCP or patient-level datasets, no access to your systems and no integration.
Confidential by default
Your scope, findings and recommendations are yours. Theralexis research and benchmarks are market-wide, and never contain organisation-specific assessments or recommendations.
Retained and dated
Every response is kept with its measurement conditions, so a finding can be re-read and a change can be compared.
Start with one brand, one market
See how AI represents your brand.
Tell us the brand and the market. We will show you how AI systems represent it, how it compares, and where your teams could act first. No obligation, no sales presentation.