The Percentage
The academic literature on conspiracy theories spans psychology, political science, philosophy, and sociology. It includes hundreds of peer-reviewed studies, multiple meta-analyses, and entire journals devoted to the subject. It examines who believes, why they believe, how beliefs spread, and how they might be corrected. It does not include a single systematic study of how often the label turns out to be wrong.
The Question
Of all the claims that have been labeled “conspiracy theory” by mainstream institutions, media outlets, and academic researchers, what percentage were later confirmed as substantially true by government documents, court proceedings, congressional hearings, or declassified records?
It is not a complicated question. It has a numerical answer. Somebody could calculate it.
Nobody has.
What the Literature Contains
The academic study of conspiracy theories has grown from a fringe interest into a substantial research field. A 2019 systematic review in Frontiers in Psychology by Goreis and Voracek identified 96 psychological studies on conspiracy beliefs published before early 2018. A follow-up review covering 2018 through 2021, published in the same journal by Szymaniak and colleagues in 2023, identified 274 more. The COVID-19 pandemic alone produced dozens of additional studies on conspiracy belief and health behavior.
The field now has multiple meta-analyses. Stasielowicz (2022) conducted a meta-analysis of personality correlates of conspiracy belief across 127 independent samples. Biddlestone and colleagues (2025) analyzed 971 effect sizes from 279 studies examining the psychological motives behind conspiracy beliefs. A 2026 meta-analysis in the European Journal of Social Psychology, also by Stasielowicz, synthesized 273 effect sizes from 56 samples examining interventions designed to reduce conspiracy beliefs. The Conspiratorial Mind, a meta-analytic review published by the American Psychological Association, spanned 170 studies, 257 samples, 1,429 effect sizes, and 158,473 participants.
Here is what the combined literature studies: who believes conspiracy theories. Why they believe them. What personality traits correlate with belief. How beliefs spread through networks. Whether beliefs increased during the pandemic. How beliefs relate to vaccine hesitancy. How beliefs relate to political orientation. Whether interventions can reduce belief. What cognitive biases are associated with belief. Whether conspiratorial thinking is a stable disposition. How age, education, gender, and ethnicity relate to belief.
Here is what it does not study: whether the beliefs are true.
The funding sources for the major conspiracy theory studies are documented in their disclosures and follow a consistent pattern. The foundational review by Douglas, Sutton, and Cichocka (2017) was funded by the Centre for Research and Evidence on Security Threats (CREST), which is funded by the UK Home Office and UK intelligence and security agencies, with a total budget of £5.3 million. CREST has since commissioned a dedicated project titled “Conspiracy Theories and Extremism,” classified under national security threat research. The Uscinski et al. (2022) study on whether conspiracy beliefs have increased over time was funded by National Science Foundation Grant #2123635, a “Secure and Trustworthy Cyberspace” (SaTC) award. Additional Uscinski research was supported by the Leverhulme Trust’s “Conspiracy and Democracy” project. Other studies in the field are funded by the Irish Research Council, the Economic and Social Research Council, and similar government research bodies. To state the structural observation plainly: the agencies whose predecessors conducted MKUltra, COINTELPRO, and mass surveillance are funding the research that studies belief in conspiracy theories as a security threat. The research itself discloses these funding sources. Nobody has remarked on the circularity. The field studies the people who distrust institutions, and the institutions fund the field.
What the Literature Assumes
The field does not ignore the existence of confirmed conspiracies. It acknowledges them. Briefly, in passing, as a throat-clearing exercise before moving to its actual subject.
The pattern is consistent across the literature. A study will note that confirmed conspiracies exist: MKUltra, COINTELPRO, the Tuskegee syphilis experiment. It will then state that its focus is on “unfounded” or “implausible” conspiracy theories. It will not explain how the determination of “unfounded” was made for the specific theories under examination. It will not address the fact that every confirmed conspiracy was, at some point prior to confirmation, categorized as unfounded.
This throat-clearing follows the temporal pattern documented in The List. The confirmed conspiracies cited are almost invariably the mid-20th-century cases exposed by the Church Committee in 1975-1976. They are distant. Historical. Safely resolved. The implicit message is the same one the discovery pipeline produces: yes, that happened, but it was a long time ago, and the system corrected itself. The admission inoculates against further inquiry, just as the passage of time softens moral judgment. The academics cite the same cases the institutions have already processed and rendered inert.
The 2019 Goreis and Voracek review noted that the field uses measurement instruments that presuppose conspiracy theories are false. Scales ask respondents to rate agreement with statements framed as examples of irrational belief. The Generic Conspiracist Beliefs Scale includes items such as “experiments involving new drugs or technologies are routinely carried out on the public without their knowledge or consent.” This statement is, as a historical matter, true. It describes MKUltra. But in the context of the measurement instrument, agreeing with it is scored as evidence of conspiratorial ideation.
The word “beliefs” is worth noting. The field studies conspiracy beliefs. Not conspiracy claims, not conspiracy hypotheses, not conspiracy allegations. Beliefs. The word carries a specific epistemological weight: a belief is something held by a mind, not something verified by evidence. You study someone’s beliefs when you are interested in their psychology, not when you are interested in whether they are right. Nobody studies a physicist’s “beliefs” about gravity. They study the physicist’s findings. The framing is the diagnosis. By the time you have categorized the subject matter as a set of beliefs to be explained, you have already answered the accuracy question. You have decided it does not need asking.
Joseph Uscinski of the University of Miami, one of the most cited scholars in the field, has noted this problem. In a 2018 paper in Argumenta, he cautioned against labeling conspiracy theories using a true-or-false dichotomy. He argued that conspiracy theories should be treated as relatively more or less suspect based on the amount of verifiable evidence in their favor. He also warned that conspiracy theories have “unique epistemological properties” that shield them from easy falsification, and that treating them categorically as misinformation risks suppressing a form of political dissent.
Uscinski’s caution has not become the field’s practice. The overwhelming majority of studies treat conspiracy belief as the dependent variable to be explained, not as a set of claims to be evaluated. The researchers study the believers. Nobody studies the researchers’ beliefs about the believers. The direction of the diagnostic gaze has never been reversed.
What Comes Close but Misses
There are studies that approach the accuracy question without quite asking it.
In 2016, David Robert Grimes of the University of Oxford published “On the Viability of Conspiratorial Beliefs” in PLOS ONE. The paper constructed a mathematical model to estimate how long large-scale conspiracies could remain secret, given the number of participants involved and a calculated probability of individual leaks. Grimes concluded that conspiracies involving more than 1,000 agents would become untenable quickly.
The model was elegant. It was also the wrong question. It asked how long conspiracies could remain hidden. It did not ask how many labeled conspiracy theories turned out to be true. It calibrated its leak probability using historical cases of exposed conspiracies, which means the model’s own parameters were derived from confirmed conspiracy theories: the very category whose existence it was implicitly arguing against. And the model did not account for what The List documents in detail: that the average gap between a confirmed conspiracy’s operation and its public confirmation exceeds thirty years, that confirmation consistently requires extraordinary external pressure, and that the mechanism for confirmation (congressional investigation of the Church Committee’s scope) was itself dismantled after the 1970s while the mechanism for concealment was not. Grimes modeled leak probability. He did not model the probability that a leak would be believed.
In 2022, Uscinski, Klofstad, and colleagues published “Have Beliefs in Conspiracy Theories Increased Over Time?” in PLOS ONE. The study tracked public endorsement of 37 conspiracy theories across multiple time periods. It found that of the 37 beliefs examined, only six showed a significant increase over time. Fifteen showed a significant decrease. The study’s definition of “conspiracy theory” explicitly included an epistemological criterion: the theories “have not been judged as likely accurate by the appropriate epistemological bodies.”
This is the closest the empirical literature comes to tracking outcomes. But the study measured changes in belief levels over time, not which specific theories were later confirmed. And its definition pre-excluded confirmed theories by design. Once a conspiracy theory is judged accurate by “appropriate epistemological bodies,” it exits the category. The confirmed ones disappear from the dataset automatically.
Several scholars have noted what amounts to a definitional escape hatch in conspiracy theory research. When a conspiracy theory is confirmed, it is reclassified. It becomes “investigative journalism” or “historical analysis.” The Wikipedia article on conspiracy theories states this explicitly: theories that are “proven to be correct such as the Watergate scandal are usually referred to as investigative journalism or historical analysis rather than conspiracy theory.” The Uscinski definition encodes the same logic: if the appropriate bodies have judged a theory accurate, it ceases to be a conspiracy theory. This means the category “conspiracy theory” is defined so that it can never include confirmed cases. The false positive rate of the label is, by construction, zero. Not because no conspiracy theories are ever confirmed, but because confirmed ones are reclassified on exit. The field has built an instrument that cannot register its own errors.
What the Philosophers See
The philosophers who study conspiracy theories have noticed the missing question. They have not been quiet about it.
Charles Pigden of the University of Otago, in a 2023 paper in Social Epistemology, described “conspiracy theory” as a “Tonkish term,” borrowing a concept from philosopher Arthur Prior. A Tonkish term is a word whose usage rules license inferences from truths to falsehoods. Pigden’s argument: the phrase “conspiracy theory” lets you infer from the true premise “this is a theory about a conspiracy” to the false conclusion “this theory is false, crazy, or unbelievable.” The term, as commonly deployed, smuggles its verdict into its label. Generalizations about “conspiracy theories” therefore do not have determinate truth-values, because the category does not have a stable extension. In Pigden’s assessment, psychological and social scientific research built on this term is “often about as intellectually respectable as an enquiry into bastards and what makes them so mean.”
Lee Basham and M R. X. Dentith, in a 2016 paper in the Social Epistemology Review and Reply Collective, were more direct. They argued that a group of social psychologists had created what amounted to a “conspiracy theory panic,” proposing to develop scientific techniques that would prevent people from even recognizing conspiracy as an explanatory option. Basham and Dentith pointed out that governments and corporations routinely conspire, that this fact is neither startling nor controversial to anyone who is historically literate, and that the social science literature was proposing to treat awareness of this fact as a pathology requiring intervention.
The word they used was “cure.” The social scientists wanted to cure everyone. What the philosophers noticed is that the cure was being designed without first establishing that the patients were sick, without measuring whether the beliefs being targeted were accurate or inaccurate, and without disclosing that the research was funded by the same category of institutions whose confirmed conspiracies appear on The List.
David Coady of the University of Tasmania, responding to Sunstein and Vermeule’s proposal for “cognitive infiltration” of conspiracy theory groups, pointed out that their assertion that conspiracy theories in open societies are usually unjustified conflicts directly with documented cases like Watergate, MKUltra, and COINTELPRO. Coady’s position, developed across a series of papers and his 2012 book What To Believe Now, is that the truly irrational position is not conspiracy belief but its categorical dismissal.
These are not marginal voices. Pigden’s work is published in Social Epistemology. Coady edited the first philosophical anthology on the subject. Dentith’s 2023 special issue on “Conspiracy Theory Theory” in Social Epistemology included contributions from across the philosophical spectrum. The philosophers have identified the gap. The empirical researchers have not filled it. The question is whether they have been funded to fill it.
Charles Pigden, “‘Conspiracy Theory’ as a Tonkish Term,” Social Epistemology, 2023
The Search
To confirm the absence, TFRi conducted a structured search for any peer-reviewed study that systematically measures the accuracy rate of claims labeled “conspiracy theory.”
Databases searched: Google Scholar, PsycINFO (via APA PsycNet), PubMed, PhilPapers, JSTOR. Search terms included: “conspiracy theory” combined with “accuracy rate,” “confirmation rate,” “true positive rate,” “verified,” “substantiated,” “base rate accuracy,” “meta-analysis confirmed,” and “hit rate.” Date range: unrestricted through March 2026.
Results: zero.
No study was found that systematically identifies a sample of claims labeled “conspiracy theory” at a defined point in time and tracks which were subsequently confirmed by primary source evidence. No study was found that calculates a ratio. No study was found that treats accuracy as a dependent variable.
Individual examples of confirmed conspiracy theories appear frequently in the literature. They appear in introductory paragraphs, in footnotes, in parenthetical acknowledgments. MKUltra, COINTELPRO, the Gulf of Tonkin fabrication, Tuskegee, NSA mass surveillance, the tobacco industry cover-up: these are cited routinely as historical context. But they are never aggregated. They are never counted. They are never compared against the total population of claims that received the label. And as The List documents, the examples that appear in the throat-clearing are overwhelmingly the safely historical ones: the Church Committee revelations of the 1970s, processed by time into institutional trivia. The more recent confirmations, the ones that implicate living institutions, appear less often.
The label is applied with the frequency of a diagnostic test. Nobody has measured the test’s false positive rate.
Why It Does Not Exist
There are methodological reasons why such a study would be difficult. They are real, and they are worth stating honestly.
First, the sample frame is unclear. There is no authoritative registry of claims that have been officially labeled “conspiracy theory.” The label is applied informally by journalists, politicians, academics, and institutions. Defining the denominator requires decisions about what counts as a labeling event and who counts as a labeling authority.
Second, confirmation is not binary. Some conspiracy theories are confirmed in whole. Some are confirmed in part: the core claim was accurate, but specific details were wrong. Some remain in an indeterminate state where evidence is suggestive but not conclusive. Any systematic study would need a rubric for these gradations.
Third, the timeline is open-ended. A claim labeled “conspiracy theory” in 1975 might be confirmed by declassified documents in 2005. Any study would need to define a follow-up period, and any such period would be arbitrary.
These are genuine methodological challenges. They are also the kind of challenges that empirical researchers solve routinely in other fields. Epidemiology tracks diagnostic accuracy across uncertain timelines. Legal scholarship tracks wrongful conviction rates despite ambiguous case boundaries. Intelligence analysis has entire frameworks for measuring predictive accuracy under uncertainty. The Brier score, calibration curves, base rate analysis: the tools exist.
The methodological challenges explain why the study would be hard. They do not explain why nobody has attempted it. And they do not explain why nobody has funded it.
The more likely explanation for the absence is structural, and it operates on two levels. The first is professional. The academic field of conspiracy theory research is organized around explaining belief, not evaluating claims. Its journal categories, its tenure incentives, and its measurement instruments are all oriented toward the psychology of belief, not the epistemology of accuracy. A researcher who proposed to systematically track the accuracy rate of claims labeled “conspiracy theory” would be proposing to evaluate the reliability of the very label that defines the field’s subject matter. If the accuracy rate turned out to be nontrivial, the implications for decades of research that treats the label as epistemically settled would be considerable. The second level is financial. The field’s primary funding sources are government security agencies and national research councils operating under security mandates. The NSF SaTC program, the UK Home Office via CREST, and similar bodies fund research that treats conspiracy belief as a threat to be understood and countered. An accuracy study would, by design, evaluate whether some of those “threats” were accurate descriptions of reality. It is difficult to secure government security funding for a study whose potential finding is that people who distrust the government are sometimes right. This is not a conspiracy. It is an incentive structure. It is the kind of structural pressure that produces blind spots in any field, and it operates without anyone needing to plan it.
The Analogy
Consider a medical diagnostic test. The test identifies a condition. It is applied to millions of patients annually. It has been in use for decades. Academic literature on the test spans hundreds of studies. Researchers have studied the demographics of patients who test positive. They have studied the psychological characteristics of patients who resist the diagnosis. They have studied interventions to increase compliance with treatment following a positive result.
Now suppose that in all those years, nobody has ever measured the test’s false positive rate. Nobody has tracked a cohort of positive results forward in time to see how many turned out to be correct. Nobody has calculated sensitivity, specificity, or positive predictive value. The test has never been validated against outcomes.
In medicine, that test would be withdrawn pending validation. In academic research on conspiracy theories, that test is the entire field’s organizing principle.
What a Study Would Require
The study that does not exist would need the following components:
A defined sample frame: a set of claims that were labeled “conspiracy theory” by identifiable institutional sources (major newspapers, government officials, academic publications) during a defined period. The period would need to be long enough to capture the confirmation lag documented in The List, where the average gap between a conspiracy’s operation and its public confirmation exceeds thirty years. It would also need to avoid the temporal trap that the discovery pipeline sets: anchoring the sample in safely historical cases that the system has already processed. Any start date chosen will itself reflect assumptions about when the label became active, and any end date will exclude cases still awaiting confirmation. These are design decisions, not obstacles.
A confirmation rubric: a set of criteria for what counts as substantiation. Primary source documents (declassified files, court records, congressional testimony, corporate internal documents released through litigation) would constitute strong confirmation. Credible investigative reporting with named sources would constitute partial confirmation. Absence of disconfirming evidence would not constitute confirmation.
A classification scheme: confirmed, partially confirmed, unconfirmed, disconfirmed, indeterminate. Each claim would be coded by at least two independent researchers with inter-rater reliability measured.
A base rate calculation: the number of confirmed and partially confirmed cases divided by the total sample, with confidence intervals.
This is a standard research design. It could be executed by a graduate student with access to a university library and a newspaper archive. It would not require novel methodology. It would not require proprietary data. It would require funding from a source that does not have a stake in the outcome. That may be the hardest requirement of all.
What the Absence Means
The absence of measurement does not prove that the conspiracy theory label is inaccurate. It might be highly accurate. It might correctly identify baseless claims 99 percent of the time. It might correctly identify them 70 percent of the time. It might correctly identify them 50 percent of the time. Nobody knows. That is the point.
A label that is applied categorically but has never been validated empirically is not a finding. It is a convention. It may be a useful convention. It may be a harmful one. Without measurement, there is no way to know.
What can be stated factually is this: a substantial number of claims that were labeled “conspiracy theory” by mainstream institutions have been subsequently confirmed by primary source evidence. The List documents ten. TFRi’s False Negative Registry documents more. The exact ratio has never been calculated. The academic field devoted to studying conspiracy theories has produced hundreds of studies on the psychology of belief, funded largely by government security agencies, and zero studies on the accuracy of the label those agencies helped popularize.
The field studies the thermometer. It has never checked the thermometer against the patient’s actual temperature.
If someone were to construct the study described above, even a preliminary version using only the most unambiguous cases, the result would be a number. A percentage. A base rate. That number, whatever it turned out to be, would transform the conversation. If the accuracy rate were 2 percent, it would validate the current practice of categorical dismissal. If it were 20 percent, it would raise serious questions about a label that functions as a verdict. If it were higher, the implications would be extraordinary. The fact that nobody in the history of the field has produced this number is itself a finding. The most important statistic in conspiracy theory research is the one that does not exist. The Tinfoil Research Institute maintains an open invitation to any researcher, institution, or funding body interested in producing it. The methodology is described above. The data is in the public record. The contact is signal@tinfoil.wtf.
This dispatch cites 19 sources across two active categories. Estimated breakdown: academic and scholarly (peer-reviewed meta-analyses, systematic reviews, philosophical papers, and published books by Goreis, Szymaniak, Stasielowicz, Biddlestone, Bowes, Douglas, Uscinski, Grimes, Pigden, Basham, Dentith, Coady, Keeley, Sunstein) ~90%; platform self-reporting (CREST’s own institutional disclosures about its funding and mandate, UCL press release describing CREST-commissioned projects) ~10%.
These percentages are editorial estimates, not computed metrics. A source may appear in more than one category. A dispatch that examines an academic field will necessarily cite that field’s own literature as its primary evidence: the papers are both the subject and the source. No primary documents, independent journalism, or subject’s own statements are cited because the dispatch is not about a person, an event, or a declassified record. It is about a body of research and what that body of research has not studied. The relevant question is whether independent sources corroborate the factual claims. In this dispatch, all factual claims about the literature’s content, scope, and funding are independently verifiable through the cited papers’ own abstracts, methodology sections, and funding disclosures. The full source list follows.
Sources
Goreis, Andreas and Martin Voracek, “A Systematic Review and Meta-Analysis of Psychological Research on Conspiracy Beliefs: Field Characteristics, Measurement Instruments, and Associations With Personality Traits,” Frontiers in Psychology, Vol. 10, Article 205, 2019.
Szymaniak, Katarzyna, Monika Kaczmarek, and Roland Imhoff, “Contemporary Trends in Psychological Research on Conspiracy Beliefs: A Systematic Review,” Frontiers in Psychology, Vol. 14, Article 1075779, 2023.
Stasielowicz, Lukasz, “Who Believes in Conspiracy Theories? A Meta-Analysis on Personality Correlates,” Journal of Research in Personality, Vol. 98, 2022.
Stasielowicz, Lukasz, “The Effectiveness of Interventions Addressing Conspiracy Beliefs: A Meta-Analysis,” European Journal of Social Psychology, Vol. 56, No. 1, 2026, pp. 275-291. DOI: 10.1002/ejsp.70041. Published online December 8, 2025.
Biddlestone, Mikey, Ricky Green, Karen M. Douglas, Flavio Azevedo, Robbie M. Sutton, and Aleksandra Cichocka, “Reasons to Believe: A Systematic Review and Meta-Analytic Synthesis of the Motives Associated with Conspiracy Beliefs,” Psychological Bulletin, Vol. 151, No. 1, 2025, pp. 48-87.
Bowes, Shauna M., Theodora Costello, et al., “The Conspiratorial Mind: A Meta-Analytic Review of Motivational and Personological Correlates,” Psychological Bulletin, American Psychological Association, 2023.
Douglas, Karen M., Robbie M. Sutton, and Aleksandra Cichocka, “The Psychology of Conspiracy Theories,” Current Directions in Psychological Science, Vol. 26, No. 6, 2017, pp. 538-542. Funded by the Centre for Research and Evidence on Security Threats (CREST), Economic and Social Research Council Award ES/N009614/1.
Centre for Research and Evidence on Security Threats (CREST), funded by the UK Home Office and security and intelligence agencies, £5.3 million total budget. Economic and Social Research Council Award ES/V002775/1. crestresearch.ac.uk.
Uscinski, Joseph E., “The Study of Conspiracy Theories,” Argumenta, 2018.
Uscinski, Joseph E., Adam M. Enders, Casey Klofstad, et al., “Have Beliefs in Conspiracy Theories Increased Over Time?”, PLOS ONE, Vol. 17, No. 7, 2022. Funded by National Science Foundation SaTC Grant #2123635 and the Leverhulme Trust “Conspiracy and Democracy” project.
Grimes, David Robert, “On the Viability of Conspiratorial Beliefs,” PLOS ONE, Vol. 11, No. 1, January 26, 2016.
Pigden, Charles, “‘Conspiracy Theory’ as a Tonkish Term: Some Runabout Inference-Tickets from Truth to Falsehood,” Social Epistemology, Vol. 37, No. 4, 2023, pp. 423-437.
Basham, Lee and Matthew R. X. Dentith, “Social Science’s Conspiracy-Theory Panic: Now They Want to Cure Everyone,” Social Epistemology Review and Reply Collective, Vol. 5, No. 10, 2016, pp. 12-19.
Coady, David, “Cass Sunstein and Adrian Vermeule on Conspiracy Theories,” Argumenta, 2017.
Coady, David, What To Believe Now: Applying Epistemology to Contemporary Issues (Chichester: Wiley-Blackwell, 2012).
Dentith, M R. X., ed., “Conspiracy Theory Theory” (special issue), Social Epistemology, Vol. 37, No. 4, 2023.
Keeley, Brian L., “Of Conspiracy Theories,” The Journal of Philosophy, Vol. 96, No. 3, March 1999, pp. 109-126.
Sunstein, Cass R. and Adrian Vermeule, “Conspiracy Theories: Causes and Cures,” Journal of Political Philosophy, Vol. 17, No. 2, 2009, pp. 202-227.
University College London, “Crime Science academics win funding to lead projects tackling national security threats,” December 21, 2021. Describes CREST-funded project “Conspiracy Theories and Extremism” by Professor Paul Gill.
Connected Research
This dispatch is part of the TINFOIL™ Consensus Machine series, an eight-part investigation into how institutional knowledge systems manage what counts as credible. Related dispatches:
The Label · The Oldest Trick in the Book · The List · The Reliable Source · The Mechanism That Predicts Its Own Dismissal · The Science
TINFOIL™ makes cognitive defense gear for people who check the instrument before trusting the reading.