How to Research a Conspiracy Theory (Without Falling Into Bad Sources).

Doing this well is a skill. Doing it badly is easy. The difference is almost entirely in the quality of sources you anchor to and the discipline with which you track them.

Pick the right question

Conspiracy theories have vague shape and sharp sub-questions. "Was the moon landing fake?" is too broad; "Did NASA have the thermal-protection capability in 1969 to return a crewed vehicle from translunar trajectory?" is researchable. The JFK assassination covers 63 years of questions; asking "Was there a second shooter at Dealey Plaza?" narrows to a specific evidentiary question.

The rule: your question should be answerable with evidence, not with opinion. If you cannot imagine what finding would change your mind, you are asking an identity question, not a research question.

Find the primary sources that actually exist

Most conspiracy topics have more primary-source material than people realize. For the topics we cover on this site:

• Operation Northwoods: the full 15-page JCS 1969/321 document, declassified 1997.
• JFK assassination: the Warren Commission Report, the HSCA 1979 final report, the Zapruder film, the 2025 release of 80,000 additional pages.
• Chemtrails: HR 2977 as introduced in 2001, 1974 Congressional testimony on Operation Popeye, Harvard SCoPEx public documentation, Tennessee SB 2691.
• HAARP: the Eastlund patents (US 4,686,605 and successors), the 1999 EU Parliament resolution A4-0005.
• MH370: the Malaysian 2018 final report, the 2017 ATSB operational search report, the Inmarsat technical papers.

Each of these is available, free, from official archives. Most of them are PDF-downloadable and most of them are readable without technical training. If a research topic has no accessible primary material, that's a red flag about the topic — and often about what people are claiming it contains.

Build a credibility tier for every source

As you collect material, place each source into one of three tiers:

• Strong: declassified government documents, court records, named eyewitness testimony, peer-reviewed scientific literature on the specific question, primary video with verifiable provenance.
• Interesting: investigative journalism by named reporters with established track records, academic books, independent research with documented methodology.
• Weak: anonymous claims, unverified video, sources without identifiable provenance, "I heard from a friend who works at..." — useful as leads, not as load-bearing evidence.

Classified's credibility-rating feature is built around exactly this distinction. Every piece of evidence gets marked Strong, Interesting, or Weak. When you present a case, you see the tier distribution — and so does anyone you show it to.

Save everything locally, immediately

Online conspiracy-research material has a half-life. Videos are pulled. Accounts are suspended. Articles are quietly edited. URLs become 404s. If you find a piece of primary material you intend to use, save it locally on the first visit — the file, the URL, the access date, the creator.

Classified is built for this: tap Share → Classified from any app, and the video is saved to your iPhone with URL, creator, and timestamp attached. But any local-saving tool is better than trusting platform-side saves alone.

Common failure modes

The most common ways conspiracy research goes wrong:

• Source laundering. You find claim X in a YouTube video that cites a blog post that cites a forum thread that cites "reports." When you chase it all the way back, the original source is anonymous and unverifiable. The claim has been passed hand-to-hand until it appeared authoritative. Do this tracing before you use the claim.
• Composite sourcing. A researcher combines real fact A with real fact B and implies a connection that neither source actually makes. If the connection is load-bearing, you need a source that makes the connection directly — not two sources and your inference.
• Pattern-matching without mechanism. "These events all happened in the same year" is a starting point, not a conclusion. Before you treat correlation as connected, you need a plausible mechanism and, ideally, evidence of that mechanism operating.
• Unfalsifiable theories. If no observation you could make would change your view of the theory, you have a closed system — not a research question. Good research has an answer to "what would make me update?"
• Bad-faith sourcing. Some sources are known to fabricate. Some are known to reliably misquote. You need to know the track record of the sources you cite — not just whether they agree with your conclusion.

Engage the counter-argument seriously

The discipline that separates serious research from confirmation: deliberately searching for and steel-manning the best counter-argument. Not the weakest version of the skeptical case — the strongest one. If you cannot restate the mainstream position on the topic in a form a mainstream analyst would accept, you are not yet done researching.

Good practitioners of this: the moon-landing hoax researchers who engage Van Allen's own published work; the JFK researchers who engage Gerald Posner's pro-Warren-Commission arguments; the chemtrails researchers who acknowledge atmospheric-chemistry basics while arguing they are incomplete.

Organize the case before you present it

A case for someone else needs structure: specific claims, each supported by specific evidence, each with a credibility tier. "Here is a lot of stuff" is not a case. "Here is claim X, supported by Strong sources A, B, C and Interesting source D; here is claim Y, supported by..." is a case.

This is what the case-and-argument structure in Classified is built for. But any structured workspace — Notion, Obsidian, a paper notebook — works as long as the structure is actually there.

Update when evidence changes

If a declassification, a new witness, or a primary-source discovery changes your view on one element of the case, update. Research that never updates is advocacy. The honest researcher is willing to hold some elements of a case strongly, some weakly, and some as open questions — and to move pieces between those categories as new material arrives.

Related reading

For examples of research built to this standard on specific topics:

• Operation Northwoods — built almost entirely on one declassified primary document
• JFK assassination — the balance between two conflicting official findings (Warren vs HSCA)
• Chemtrails — named laws, declassified programs, state legislation, independent research
• MH370 — satellite technical analysis, physical debris evidence, ongoing investigation
• Do your own research: a practical guide — the companion methodology page
• The biggest conspiracy theories — every topic, with full research pages