About

Editorial policy

The site is built to help readers inspect the record. It avoids argumentative framing, exposes AI usage, and treats corrections as part of the research workflow.

37,141
Indexed records

The current corpus is a curated subset, not the complete collection.

2,165
OCR records

Full-text search depends on OCR availability and quality.

neutral
Posture

The site surfaces sources and tensions without endorsing a theory.

Neutrality

This site does not endorse any theory of the assassination. It indexes primary-source records and a small set of allowlisted secondary references; it does not argue for or against the Warren Commission's conclusions, the HSCA's conclusions, or any other account. The site's job is to make the documentary record reachable and to show where the record is internally consistent, where it is in tension, and where it is incomplete.

To enforce that posture in practice, two surfaces are maintained side-by-side: Open Questions surfaces tensions, contradictions, redaction patterns, and unresolved threads; Established Facts catalogs findings that official investigations substantially agree on, plus a tagged Contested tier for claims where the official record is internally inconsistent. Neither surface is the final word.

Source allowlist

AI-generated content is grounded in a curated citation registry of primary sources and allowlisted reference works. The allowlist includes the Warren Commission Report and Hearings volumes, the HSCA Final Report and appendix volumes, the ARRB Final Report, Church Committee Book V, NARA finding aids, the FBI Records Vault, and relevant court records.

The allowlist excludes partisan blogs, self-published books, and speculative secondary works - not because those works have no value to researchers, but because citing them on a neutral site would push the editorial posture.

Banned language in AI-generated content

Prompts driving the site's Gemini calls bar a standing list of editorializing words: remarkably, curiously, suspiciously, conveniently, shadowy, mysterious, allegedly except inside direct quotes, plus so-called, mere, obviously, clearly, undoubtedly, tellingly, revealingly. Generated output is spot-checked against this list. When a banned word slips through, the containing passage is flagged for review and regeneration.

AI disclosure

Every AI-generated panel on the site carries a footer line naming the model, generation date, and record count the output was grounded in. Users should treat AI summaries as starting points: they speed primary-source research, but they do not replace it. Where a claim matters, read the underlying records.

Corrections

Factual errors in entity bios, timeline entries, AI summaries, or document metadata can be reported via the corrections form. Entity and topic pages also carry a report link that pre-fills the form with the surface and target id. Submissions queue for editorial review; cross-checking against linked sources is the strongest correction mechanism.

What this site is not

It is not a complete mirror of the JFK Assassination Records Collection; it is a curated subset of 37,141 records, with full-text OCR on 2,165 of them. The methodology page details scope, pipeline, and known limitations. It is not a conspiracy site, and it is not a defense of orthodoxy; generalizing from what this site does or does not contain to what is hidden or proven is beyond what the documentary record supports.

Back to About
Editorial policy · JFK Research Center