Label Ledger
For regulated markets
What AI claims about your medicine.
Checked against your own label, kept on the record.
ChatGPT, Claude, Gemini and Perplexity answer questions about your products every day, unreviewed, unrecorded, and increasingly trusted by the patients and prescribers who ask them. AlphaCitation sits in on those conversations, checks every claim against your own approved label, and turns each discrepancy into a governed case with an audit trail you can verify yourself.
A regulated communication, held to a regulated standard, whether or not you wrote it.
Capture
Every answer captured verbatim, the moment it's given, and sealed.
Check
Each claim checked against your own approved label.
Prove
An audit trail you can verify yourself.
Why this exists
Every promotional and medical communication your company issues goes through review: medical, legal, regulatory. AI answers don't.
When someone asks an assistant whether your drug is right for a pregnant patient, how it compares to a competitor, or what it's approved to treat, the answer is generated on the spot, at scale, with no reviewer, no record, and no obligation to match your label. That answer behaves like a promotional claim and reaches the same audiences your promotional claims reach.
But until now it has been invisible to the teams responsible for exactly this class of risk: off-label exposure, unsubstantiated benefit claims, competitive misrepresentation, safety-population boundaries.
Label Ledger makes that channel observable, checkable, and defensible. Not “how visible is your brand”: what is AI actually claiming, is it true against your own regulatory record, and can you prove what was said and when.
What it detects
Five classes of divergence
Claims your label doesn't support
AI credits your medicine with a therapeutic benefit its current approved label doesn't carry, often a benefit that's real for the drug class or a sibling product, wrongly attached to yours.
Off-label population exposure
Highest severityAI presents your medicine as appropriate for a population your own label explicitly excludes: for example a paediatric use or use in pregnancy. This is the finding that routes straight to patient safety, at the highest severity.
Your fact, credited to a competitor
A real, on-label advantage of yours that AI attributes to a competitor instead: the other side of losing the claim.
Real benefits AI under-tells
Verified, on-label benefits that AI systematically omits or buries, so the picture a patient or prescriber receives is quietly narrower than the approved one.
Your own prohibited claims, repeated
Claims you've specifically designated as never-to-be-made, appearing in AI answers about your brand, checked against a list you provide, kept deliberately separate from the independent regulatory checks above.
How it works
01–05
- 01
Capture
Every answer the major AI engines give about your brand is captured verbatim the moment it's observed, with the exact prompt, the model that produced it, and a precise timestamp. Sealed so it can't be quietly edited later.
- 02
Check against your own regulatory record
Each claim is checked against your medicine's own approved label: an independent, primary regulatory source, not your marketing copy and not the AI's opinion of itself. Findings resting instead on claims you've registered yourself are labelled as exactly that, never blended in.
- 03
Open a governed case
Every real discrepancy becomes a case, not just an alert: routed to the right risk domain (patient safety, promotional/regulatory, comparative, market access, or narrative drift), given a severity, and assigned an owner.
- 04
Close only on proof
A case is marked fixed only when a fresh, real re-scan confirms AI has actually stopped saying it, never on someone's unverified say-so. If it resurfaces later, the case reopens itself.
- 05
Prove it independently
The full life of every case, first detected, every recurrence, the verified close, and every underlying captured answer, is anchored so it can be verified by an outside party, with no need to trust our database.
Built for the teams who own this risk
Pharmacovigilance & Patient Safety
Off-label and excluded-population exposure surfaced the moment AI starts implying a use your label rules out.
Regulatory Affairs
The AI channel treated as what it is: an unreviewed surface making promotional-grade claims, now monitored against the approved label with a record for every one.
Medical Affairs
Where AI misstates the science, under-tells a real benefit, or drifts from the evidence, with the verbatim answer and the label reference side by side.
Legal
A tamper-evident, independently verifiable evidence chain of exactly what was said and when, and of everything done about it.
Corporate Affairs & Market Access
Competitive misattribution and narrative drift, including the access and comparative claims that move commercial ground.
The evidence chain
Independently verifiable: not “trust our database.”
Most compliance tooling asks you to trust that its records are accurate. This doesn't.
Every captured answer is sealed into a tamper-evident chain, each record cryptographically linked to the one before it, so nothing can be altered, inserted, or removed after the fact without visibly breaking the chain. Case resolutions are sealed the same way, so a case's entire history is as fixed as the observations that opened it.
Periodically, the current state of that chain is anchored to an independent, third-party timestamping authority using an open, published standard: the same class of trusted timestamp used for legally binding e-signatures and software code-signing. That anchor is proof, from a party with no stake in the outcome, that the record existed in exactly its current form at that time and hasn't been back-dated or rewritten since.
Illustrative: an actual sealed case record
sha256: 7f2a8c91e3d4f206b5a17c8902ef4d61b3c8a90f…
chain_hash: 3b91fa20c76e8419d0527bc1f93a6de40128c7e…
anchored: RFC 3161 trusted timestamp, independently verifiable
Every finding carries its verbatim capture, its source reference, and its place in the chain, exportable as an evidence packet.
What that means in practice:you, your auditors, or a regulator can verify the record independently, using standard tools, without taking AlphaCitation's word for anything. If AI told a patient your medicine was safe in pregnancy on a given date, and later stopped, both facts, and the timeline between them, are provable by someone who doesn't trust us at all.
How we keep it honest
This is a monitoring and evidence system, not regulatory advice and not a replacement for your own review process. We surface and prove what AI is saying and where it diverges from your approved label; the regulatory judgment on what to do about it stays with your team.
We check against the primary regulatory record and against claims you register yourself, and we keep those two things structurally separate. The independent regulatory check is the strong one. Anything resting on your own self-asserted claims is labelled as exactly that, never presented as independent verification. We don't invent findings: where the evidence isn't there, we show nothing rather than guess. Coverage today is strongest for medicines, where an independent, current label exists to check against.
Request a confidential briefing
See it run on one of your own molecules.
Label Ledger is delivered as a scoped, configured engagement, not a self-serve plan. Tell us one product and the markets that matter, and we'll come back with a confidential briefing: including a live read on what AI is currently saying about that product, checked against its own label, with the real captured answers and the evidence chain in front of you.
Configured to your brands, competitors, markets and personas. Covered by NDA on request.