Ordestra Explainer

Compliance in Scientific Communication

When communicating scientific evidence, the words you use matter as much as the data you cite. Pharmaceutical and medical communications are regulated because imprecise language can influence clinical decisions. Ordestra applies the same discipline to all scientific communication — not just pharma.

Do-Not-Say Rules

These terms are blocked in every Ordestra output. They are common in press releases and marketing materials but incompatible with evidence-based communication.

“Breakthrough”Implies a transformative advance that may not be supported by the evidence. A single RCT is rarely a breakthrough — it is a data point in a larger evidence base.
“Cure”Overstates the permanence and completeness of a treatment effect. Most interventions show statistically significant improvement, not elimination of disease.
“Game-changing”Promotional language that inflates the significance of findings beyond what the data supports. Common in press releases, unacceptable in evidence communication.
“Proven”Science does not prove — it provides evidence of varying strength. Even well-replicated findings are subject to revision. "Demonstrated" or "supported by evidence" is accurate.
“Safe”Without qualification, "safe" implies zero risk. All interventions have risk profiles. Accurate language is "well-tolerated" or "with an acceptable safety profile in this population".
“Miracle”Implies outcomes beyond what is scientifically possible. Not compatible with evidence-based communication under any circumstance.

The 10 Compliance Checks

Every output runs through this pipeline. 9 checks are deterministic (rule-based). 1 is AI-verified against the source paper.

01
Claim Level — Study design matches the claim language used in the output.
02
Claim Preservation — All finding sentences from the source paper are preserved verbatim in the summary.
03
Attribution Accuracy — Author names, journal, and year are cited correctly and match the source.
04
Do-Not-Say — Banned promotional terms are blocked before delivery.
05
Implied Causation — Association language is not inflated to causal language.
06
Factual Grounding — No fabricated claims, statistics, or author names appear in the output.
07
Length Compliance — Output duration matches the requested summary length within tolerance.
08
Audience Gate — Content is appropriate for the tier and does not include gated features.
09
LLM Factual — AI-verified cross-check against the source paper for factual accuracy.
10
Grounding — Every assertion in the output traces back to a specific section of the source.

These rules exist because pharma-adjacent communications are regulated. Ordestra applies the same framework to any scientific communication — because evidence discipline is not industry-specific.

How Ordestra Applies This

Every Ordestra output runs through the full 10-check compliance pipeline before delivery. Hard failures (claim level violations, fabricated claims, multiple do-not-say violations) block the output and trigger an automatic retry. Soft warnings (grounding, length) are surfaced for your review.

The compliance badge on every output shows exactly which checks passed, which triggered warnings, and what needs your attention. This is the audit trail that generic AI tools cannot provide.

See for Medical AffairsTry the demo

Related Topics

What Are Claim Levels?

How study design determines what language is allowed.

For Medical & Scientific Affairs

Compliance-constrained briefings for MSLs and Medical Directors.

Manual vs Ordestra

Side-by-side comparison of compliance capabilities.

3 free credits. No card required.

Generate your first evidence-constrained summary in under two minutes.

Get Started FreeHear a Sample