Talk to us

In pharma, context is everything.

AI answers depend on who is asking. ceel.ai measures them in context: the persona, the step in the treatment journey, and your label and evidence.

A conversation, not a prompt

Every persona’s journey, mapped as one AI conversation.

Physicians and patients don’t ask AI a single question. They have a conversation, and each answer shapes the next question. ceel.ai follows it step by step and checks every answer.

PatientUnited States · English
Treatment journey
PulmonologistGermany · German

  1. PatientI’ve had a dry cough for months and get out of breath on the stairs. What could it be?

    AI answerNames IPF as a possible cause
    01Symptoms to disease

    PulmonologistWhich HRCT patterns point to IPF rather than other interstitial lung diseases?

    AI answerCites the current guideline

  2. PatientI was just diagnosed with IPF. What treatments are there?

    AI answerTalvirex mentioned, 2nd
    02Treatment options

    PulmonologistWhat is first-line antifibrotic therapy for IPF today?

    AI answerTalvirex mentioned, 1st

  3. PatientAre there other options besides Talvirex?

    AI answerBalanced, on label
    03Alternatives

    PulmonologistWhen would you choose an alternative to Talvirex?

    AI answerBalanced, on label

  4. PatientIs Talvirex better than Norbisan?

    AI answerNorbisan favoured
    04Head-to-head

    PulmonologistHow does Talvirex compare with Norbisan on FVC decline?

    AI answerTie, both cited

  5. PatientI get diarrhoea on Talvirex. Should I stop taking it?

    AI answerAdvises stopping. SmPC says manage it
    05Side effects

    PulmonologistHow are GI side effects of Talvirex managed in practice?

    AI answerMatches the SmPC

  6. PatientIs there proof that Talvirex works long-term?

    AI answerCites a patient forum
    06Clinical evidence

    PulmonologistWhat do the 52-week phase 3 data show?

    AI answerCites the phase 3 paper
On trackWorth a lookOff labelIllustrative example · Talvirex and Norbisan are fictional

Personas

It starts with who is asking.

Every measurement runs on behalf of a persona. You set it up in two steps: the stakeholder type, then the details that make it specific.

Step 1Choose the stakeholder
  • Healthcare ProfessionalA physician involved in diagnosis, treatment and care.
  • PatientA person receiving medical treatment or care.
  • PharmacistA healthcare professional specialising in the safe and effective use of medicines.
  • Nurse / CaregiverA nurse or caregiver providing hands-on patient care.
Step 2Define the persona
Healthcare Professional
Language
German
Country
Germany
AttributesOptional. Narrow down who this persona is, or leave blank to keep it general.
Specialty
Pulmonology
Practice setting
Academic centre
Career stage
Senior
Brand adoption
Evaluating
Patient caseload
Not set

  • Language and country, always

    AI answers change with both, and so do labels and guidelines. Every persona has them.

  • Attributes, when you need them

    Leave them blank for a broad view of a stakeholder group. Set them to answer a specific question.

  • The same persona, every time

    A persona asks the same way in every run, so changes in the answers come from the models, not from the questions.

From persona to prompt

Attributes steer every prompt.

The persona sets the detail, length, complexity and vocabulary of each generated prompt. The same journey step, two personas:

Healthcare Professional
  • Pulmonology
  • Academic centre
  • Senior
  • Germany
Detail
Length
Complexity
Clinical vocabulary
FocusEndpoints, comparators, discontinuation rates
Generated prompt · Side effects

“In the phase 3 data for Talvirex in pulmonary fibrosis, how do discontinuation rates due to diarrhoea compare with Norbisan, and which management strategies were used?”

Patient
  • 45–54
  • Family caregiver
  • Digital-first
  • United States
Detail
Length
Complexity
Clinical vocabulary
FocusDaily life, what is normal, when to call the doctor
Generated prompt · Side effects

“I just started Talvirex. Is it normal to have diarrhoea, and should I stop taking it?”

Illustrative example · Talvirex and Norbisan are fictional

Same side effect, different needs. ceel.ai checks every answer in the context of the persona who asked.

Backed by research

How it works

Where does AI get our medicine’s story wrong, and what do we do about it?

No prompts to write, no manual tracking. ceel.ai does the heavy lifting, so your teams can focus on the actions. Fully transparent, fully traceable.

  1. 01
    Personas
    Physicians, patients and caregivers, by country and language.
  2. 02
    The treatment journey as a conversation
    Questions generated per persona, step by step.
    1. Symptoms to diseaseHow symptoms are linked to the disease, and when to see a doctor.
    2. Treatment optionsWhich treatments AI recommends, and where your medicine fits.
    3. AlternativesWhat else physicians and patients could choose instead.
    4. Head-to-headHow your medicine compares with named competitors.
    5. Side effectsWhat AI says about safety, tolerability and managing side effects.
    6. Clinical evidenceWhich trials and data AI cites, and how it describes them.
  3. 03
    Monitored over time
    Share of answer over time, with alerts.
  4. 04
    Checked against your label
    Every answer, in every country.
Physician
DE · German
Patient
US · English
Caregiver
FR · French
Story breaks · Patient · United States · Side effects
“I get bad diarrhoea on my IPF medicine. Should I stop taking it?”
Answers advise stopping rather than managing the side effect.
Tracked over timeOngoing
Share of answer (SoA)33%
Alert
Side effects: SmPC deviation in the US
Checked against
  • Label, per country
  • Clinical evidence
  • Cited sources
Method: for each persona (a physician in Germany, a patient in the United States, a caregiver in France), ceel.ai generates the questions they would ask AI along the treatment journey, as one conversation from symptoms to disease through treatment options, alternatives, head-to-head comparisons and side effects to clinical evidence. The graphic marks where the story of a fictional medicine breaks: patients asking about side effects are advised to stop treatment. Share of answer is tracked over time (33%), and an alert flags the SmPC deviation in side-effect answers in the US. Every answer is checked against the label in each country, clinical evidence and cited sources. Demo data.

  • Conversations, not prompt lists

    The treatment journey is modelled as the conversation a physician, patient or caregiver has with AI, from first symptoms to clinical evidence, in their country and language. You choose where it starts, for example at treatment options for physicians. ceel.ai generates the questions, so nobody has to write or maintain prompts.

  • Monitored over time, with alerts

    Leading models answer these conversations on an ongoing basis. Follow share of answer and narratives over time, and set alerts for what matters, such as a drop in one country. For each new publication or claim, see when it reaches AI answers and how long it takes.

  • Where the story breaks, and what to do

    Every answer is checked against your label in each country and the evidence. You see the step where the story of your medicine breaks, and get concrete actions for Medical Affairs, Brand and Market Access, each traced to its evidence.

ceel.ai and GEO

Conversations, not isolated prompts.

ceel.ai is not a GEO tool. Generative engine optimisation (GEO) tools track whether a brand is visible in answers to isolated prompts. We analyse whole conversations, because that is how people use AI: one question leads to the next, and each answer builds on the last. Mapping these conversations, persona by persona along the treatment journey, shows what Medical Affairs, Brand and Market Access need to know beyond visibility.

Core capabilitiesceel.aiTypical GEO
How visible are my brands and medicines in AI answers?
What is the AI reach¹ of each medicine across relevant HCP and patient questions?
Which narrative does AI tell about each medicine?
Do safety statements in AI answers match my SmPC?
Which scientific publications does AI cite for each narrative, resolved to the actual paper?
Does my new congress data reach AI answers, and how fast?

¹ AI reach: how often a medicine appears across relevant AI questions and answer touchpoints.

A cited source shows what a model referenced, not what shaped its answer. Personas are simulated profiles: they show what AI tells a defined audience, not what any individual was told. Findings go through your medical, legal and regulatory review.

See where your narrative stands, and what to do next.

A 30-minute conversation about your indication, your competitors and your audiences.

Talk to usAbout ceel.ai