CSR Trust Layer · Citation Tracing · ICH E3 Validation · Knowledge Sources

Medical writing done
in hours — not days.

13 AI modules for every regulatory and scientific writing task — with a trust layer that traces every statement to its source, detects staleness when docs change, and validates ICH E3 compliance per section. Your LLM key, your data, your control.

No data sent to third parties · Your own LLM key · Human review on every output

Need expert writers?
13
AI modules
writing + QC in one platform
70%
average time saved per task
vs manual writing
trust layers per CSR section
citation · validation · staleness
100%
human-in-the-loop
you review every output
New

Latest capabilities

ICH E3 CSR Writer

New

Generate full Clinical Study Reports section by section — all 10 ICH E3 sections, 50+ pages, with source context injected from your uploaded protocol, SAP, and TLF tables. Export to Word in one click.

Citation Tracing

New

Every AI-generated statement is anchored to a source document with [N] superscript markers. Hover any marker to see the exact source, section, and supporting quote — full provenance without leaving the page.

Source Staleness Detection

New

Upload a revised protocol or SAP and the platform instantly flags every CSR section that relied on the old document — an orange 'Source Updated' banner appears so nothing slips through.

ICH E3 Section Validation

New

After generation, each CSR section gets an automated ICH E3 compliance check — a per-section score, a pass/warn/fail checklist, and a list of specific issues to address before submission.

Trust Layer — new on CSR Writer

Every statement traced.
Every section scored.

After the CSR Writer generates a section, two background tasks activate automatically: citation tracing anchors each claim to a source document with superscript markers, and ICH E3 validation scores the section for regulatory compliance. If you later upload a revised protocol, any section that relied on the old version is instantly flagged.

  • Citation [N] markers with hover tooltips showing exact source and quote
  • ICH E3 section score (0–100) with pass / warn / fail per checklist item
  • Orange 'Source Updated' banner when any source document changes
  • Manual re-annotate and re-validate buttons on any section
Try the CSR Writer
§2 SynopsisCompletePass 88%5 sources Source Updated

Protocol v3.2 (synopsis) was updated — this section may be outdated. Re-generate or review manually.

This Phase III, randomised, double-blind, placebo-controlled study evaluated Testinib 100 mg once daily in adults with Type 2 Diabetes Mellitus (T2DM).[1] The primary endpoint, change from baseline in HbA1c at Week 26, was met with statistical significance (p<0.001).[2] Safety was consistent with the known profile; no new signals were identified.[3]

ICH E3 Checklist

Study rationale and objectives statedPASS
Primary endpoint definedPASS
Sample size justification referencedWARN
Safety monitoring plan describedPASS
Patient disposition table cross-referencedFAIL

Sources

[1]Protocol v3.2·Section 2 — Synopsis
[2]SAP v2.0·Section 5.1 — Primary Endpoint
[3]Investigator's Brochure·Section 9 — Safety Summary

13 Modules

Everything a medical writer needs

From first draft to regulatory submission — all in one workspace, built for the way medical writers actually work.

Core

QC Engine

6-category AI quality check — language, scientific, references, data, formatting, compliance — plus deterministic rule engine. Embedded inline in every writer module.

Manuscript Builder

Upload TLF tables → publication-ready clinical narrative in medical writing style. Study-type aware, journal-formatted.

Trust Layer

Clinical Study Report

All 10 ICH E3 sections generated with your source docs as context. Citation tracing, ICH E3 validation scores, and source staleness detection on every section.

SLR / Meta-Analysis

PICOS → Boolean search strings → PRISMA flow — end-to-end systematic review with AI screening and data extraction.

Investigator's Brochure

ICH E6 compliant IB for any phase — all 11 sections, version-controlled, submission-ready.

Informed Consent Form

Generate or review ICFs — plain-language, GCP-compliant, multi-region, fully editable.

CTD / Dossier

ICH M4 Module 2 summaries and Module 5 CSRs for NDA, BLA, and MAA submissions.

PSUR / PBRER

ICH E2C(R2) periodic safety reports — signal evaluation, benefit-risk assessment, CIOMS tables.

Clinical Trial Protocol

Full protocols or amendments — ICH E6(R2) compliant, GCP-ready, study design through statistical section.

Slide Decks

Generate CME decks or condense 150–300 slides to target count — with slide map for full audit trail.

Reference & Claim Checker

Map every claim to its citation — flags unsupported statements, partial support, and contradictions before submission.

Submission Pack

Journal-specific formatting, ICMJE checklist, cover letter generator, and ZIP export in the target format.

New

DSUR

Development Safety Update Reports aligned with ICH E2F — aggregate safety analysis across the development programme.

Workflow

How Learned Writers works

AI generates, the trust layer verifies, you review, workflow approves. Nothing reaches regulators without your sign-off.

1

Upload your source material

Upload study data, TLF tables, reference PDFs, protocols, or manuscript drafts. Knowledge Sources lets you set org-level documents (SOPs, style guides) that auto-apply across all projects.

2

AI generates a structured draft

Our AI — powered by your own LLM key (OpenAI, Anthropic, Azure, Gemini, and more) — generates regulatory-aligned content following ICH, FDA, and EMA guidelines, drawing on your source documents as priority context.

3

Trust layer activates automatically

Citation tracing runs a second pass and anchors every statement to a source document with [N] markers. ICH E3 validation scores the section for compliance. If a source is later revised, affected sections are flagged stale.

4

Review, approve, and export

Submit for review, get sign-off through the built-in workflow, then download Word, PowerPoint, or Excel — formatted to agency or journal specs.

Inline QC — on every module

QC your output before you leave the page

Every writer module embeds a QC panel directly below the generated output. Pick your check categories — language, scientific, regulatory — run the analysis, watch issues stream live, resolve them, and export. No tab-switching. No copy-paste.

  • Works on all 13 writer modules
  • Pick exactly which categories to check
  • Issues stream in real time as the AI checks
  • Mark resolved, add comments, then export
Try it now

QC Analysis — Clinical Study Report

Score: 84/100
criticalscientific

P-value reported as p<0.05 — exact value required

majorcompliance

Missing study limitation disclosure in Section 9.3

minorlanguage

Inconsistent tense in Results paragraph 4

minorformatting

Table 3 referenced before Table 2 in text

4 issues · 2 critical · 1 major · 1 minor
Knowledge Sources

Your organisation's knowledge, always in context

Upload your SOPs, writing style guides, and regulatory templates once at the workspace level. They become the highest-priority context for every AI generation across all projects — so the AI writes to your standards, not generic ones.

P1
Org LevelCompany SOPs, house style, regulatory templates
P2
Project LevelProtocol, SAP, Investigator's Brochure, TLF data
P3
LLM BaselineModel knowledge — lowest priority, overridden by your docs

Knowledge Sources

+ Add document

Org Level — P1

Company Writing SOP v4245 KB
Regulatory Style Guide88 KB

Project Level — P2

Protocol v3.21.2 MB
Statistical Analysis Plan890 KB
TLF Tables Workbook3.4 MB
LLM Baseline — used when no source doc matchesP3
Compliance Rule Engine

Two layers of QC — AI and deterministic rules

LLM checks find nuanced issues — unclear phrasing, statistical gaps, compliance risks. Deterministic rules catch patterns that should always be flagged: missing required sections, overclaim phrases, abbreviation errors, URL citations in text.

R01Missing mandatory section per template
R02Abbreviation used without first definition
R12Overclaiming efficacy — promotional tone
CustomUpload your own rules via Excel template

Rule Engine

Upload Excel
R01templatecritical
R02abbreviationminor
R03referencemajor
R12overclaimcritical
R14regexcritical

13 built-in rules · Upload custom rules via Excel

Bring Your Own LLM

Your data never leaves your infrastructure

Connect your own API key from any provider — Anthropic Claude, OpenAI GPT-4o, Azure OpenAI, Google Gemini, Ollama (on-premise), Groq, Bedrock, and more. Every LLM call goes directly from your key to the model. We never see your prompts or outputs.

Anthropic ClaudeOpenAI GPT-4oAzure OpenAIGoogle GeminiOllama (on-prem)AWS BedrockGroqMistral
Get started
Regulatory Compliance Built In

Aligned with international guidelines

Every module is structured to comply with the relevant ICH, FDA, EMA, CDSCO, and reporting guidelines — so you don't have to start from scratch.

ICH E3 (CSR)ICH E6(R2) (GCP)ICH M4 (CTD)ICH E2C(R2) (PSUR)ICH E2F (DSUR)FDA 21 CFR 50EMA GuidelinesCDSCOPRISMA 2020CONSORTSTROBEICMJEVancouver Style

Ready to write faster?

Your admin sets up a workspace in minutes. Log in and start your first CSR section or QC run today — citation tracing and ICH E3 validation run automatically on every output.

Sign In