Build SEC filing URLs and find the actual evidence

Keep three identifiers separate

A CIK identifies the filer. An accession identifies the submission. A document filename identifies one file within that submission. Many filings contain a primary report plus attachments; the primary document is not always the document containing the fact you need.

For example, an earnings-related current report may refer to an exhibit. Retrieve the filing’s document index and inspect attachment descriptions instead of assuming the primary HTML contains the full release. This is a reading workflow, not a claim that every filing has the same exhibit structure.

CIK in data API:     0000320193
CIK in archive:      320193
Accession retained: 0000320193-YY-NNNNNN
Directory segment:  0000320193YYNNNNNN

https://www.sec.gov/Archives/edgar/data/{unpadded_cik}/{accession_without_dashes}/{primaryDocument}

Path template. YY and NNNNNN are placeholders; obtain real accession and document values from SEC metadata.

A citation should resolve the exact document

• Retain CIK, dashed accession, form, filing date and the primary or exhibit filename.
• Keep the original document URL beside any cleaned text or chunk ID.
• Record the extraction time and parser version so a later reviewer can reproduce the text view.
• When quoting a passage, identify the section or item and validate it against the document body.
• Keep an amended submission as a distinct accession and explicitly record any replacement policy.

HTML extraction has failure modes

Tables, inline XBRL, nested layout and a repeated table of contents can confuse a plain-text extractor. A heading match alone may return the table of contents rather than the actual section. Inspect section boundaries and verify numbers in the rendered table or tagged facts.

Download only the documents you need and honor the SEC access rules. Preserve a permitted raw copy or content hash for your audit trail. If an extraction step fails, return the document reference and an explicit parse failure rather than filling in a summary from a different report.

Use the correct route for the next step

For structured financial observations, start with XBRL APIs when their coverage fits. For incoming filing notifications, use the stream event as a discovery trigger and retain the official document as evidence. These two paths can share a filing identifier without being interchangeable.