Tutorial — DocumentCloud annotation workflow
Last updated: October 21, 2026
Skill: document-research-specialist / public-records-research-specialist
Time: ~20 minutes per document
Goal
Publish primary sources with highlights and structure so readers can verify claims.
Steps
- Upload PDF to DocumentCloud (or your org instance).
- Run OCR if needed; fix page order for mis-scanned bundles.
- Add summary and tags (agency, date, topic).
- Create annotations tied to exact page coordinates for each cited fact.
- Link entities (people, orgs) when the platform supports it.
- Embed or link the document from your story with “primary document” labeling.
About the platform
DocumentCloud describes itself as an all-in-one platform used by newsrooms around the world to upload and organize documents, extract and search their text, annotate, redact and edit them, and embed them on a website. Accounts sign in through MuckRock, which publishes the platform’s release notes. Its public archive holds more than 7.2 million documents that anyone can browse and gather into projects, so search it before step 1: the record you need may already be there, sometimes with another newsroom’s annotations you can cite.
Access is set per document and per note. A note can be public (shown in yellow), visible to your organization (green) or private to you (blue), and notes are searchable. Two habits follow. Keep working notes private until the story runs, then make the citing notes public in the same pass that opens the document, so a reader never meets an annotation that points at a page they cannot see. And write each public note as a full sentence naming the fact it supports, because readers and search will find the note without the story around it.
The platform includes a redaction tool. After redacting, search the document’s text for a word you removed; if the search still finds it, the text layer has survived and the redaction has failed. Apply the same check to a document that arrived already redacted, and record who made those redactions.
Pitfalls
- Redacted pages — note what is withheld vs missing.
- Bulk uploads without metadata become unsearchable later.
Next steps
- Tutorial — Web archiving for evidence preservation for live web pages
Sources
- DocumentCloud, home page — platform description and public archive count (October 2026).
- MuckRock, DocumentCloud release notes: searchable notes and OCR improvements — note access levels.
Expanded with AI research and drafting assistance from the primary sources cited; editorial review pending.
