EDGAR and Investor Relations Pages: Where Filings Live

Scope. This article describes where United States filings are published, how the search systems differ, and what the limits of each are. It recommends no data vendor, names no company, and does not tell a reader what to search for. Jurisdiction: United States federal filings, checked 14 August 2026 and re-confirmed at publication QA on 18 August 2026. Other markets publish through their own regulators, exchanges or national storage mechanisms, on different terms and with different retention.

That fact is worth stating baldly, because a substantial industry exists between the reader and the primary source, and much of what it sells is convenience rather than access.

EDGAR — the Electronic Data Gathering, Analysis, and Retrieval system — is the SEC's repository, and it is the authoritative record. An investor relations page is a copy maintained by the filer. When the two disagree, or when the IR page has quietly reorganised, EDGAR is the version that stands.

Three ways in, answering three different questions

Company search answers: what did this filer file? Given a company, it lists submissions by form type and date. This is the route for reading a known company's history and it reaches back to the beginning of electronic filing in the early 1990s.

Full-text search answers: who said this? It indexes the words inside filing documents rather than the filing metadata, which inverts the question — a phrase goes in and a list of filers comes out.

The structured data endpoints answer: what values were tagged? Filings carry machine-readable financial data, and the SEC exposes it through public interfaces returning a company's reported facts, a single concept's history, or one concept across many filers for a period.

Using the wrong one is the most common reason EDGAR feels harder than it is. A keyword typed into company search finds nothing; a company name typed into full-text search finds every filing that happens to mention it.

The boundary that produces false conclusions, and it is invisible unless you know about it. EDGAR full-text search covers filings from 2001 onward. Documents filed between the start of electronic filing in the early 1990s and the end of 2000 are present in EDGAR and can be browsed by company and form type — but they were never text-indexed, so a keyword search restricted to those years returns nothing. An empty result set for that period is not evidence that nobody said it. This is the single most consequential limitation of the system for anyone researching a historical question, because the search returns zero results rather than a warning, and zero results reads exactly like an answer. For pre-2001 material, the companies have to be identified first by other means and their documents read directly.

The identifier that actually matters

Every filer has a CIK — a Central Index Key — and it is the primary identifier in the system. Names change, tickers change and get reassigned, and several distinct entities can share a similar name. The CIK does not change, which makes it the right thing to anchor a research record to.

Ticker and name lookups exist and are conveniences built on top of the CIK. A research process that records the CIK and the accession number of the specific filing consulted can be reproduced exactly, by anyone, years later. One that records "the 2024 annual report" cannot, because there may be an amendment.

Structured data, and its trap

The tagged data behind the filings makes bulk comparison possible, and it removes transcription error from any figure a reader would otherwise retype.

The trap is tag selection. A concept a reader thinks of as one number frequently has several tags in the taxonomy, applied differently by different filers, and some concepts appear under a taxonomy for domestic filers and others under the one used by foreign private issuers reporting under IFRS. Pulling a plausible tag and treating it as the figure is how a naive extraction produces numbers that are internally consistent and wrong. The remedy is dull and reliable: reconcile any extracted value back to the face of the statement in the filing itself before using it.

Investor relations pages: useful, and not the record

An IR page carries the same filings plus material that is not filed at all — presentations, supplemental schedules, non-GAAP reconciliations, historical financial summaries, and often a longer archive of press releases than anywhere else.

Three limitations follow from who maintains it. Curation: what appears is chosen by the company, and a summary table on an IR page is a presentation of the filings rather than the filings. Impermanence: pages are reorganised, and material that existed at one time can stop existing without notice. No standard structure: every IR site is laid out differently, which is precisely the problem EDGAR's fixed form types solve.

The reasonable division of labour is that the IR page is where the supplementary material lives and EDGAR is where the record lives, and anything a reader intends to rely on should be traced back to the filing.