Methodology

How FactCheckRadar checks claims.

FactCheckRadar checks public claims, viral posts, screenshots, articles, images, and fast-moving online rumors against available evidence. The goal is simple: identify the claim, find relevant sources, explain what the evidence shows, and publish a clear verdict with visible sourcing.

What we check

We focus on factual claims that can reasonably be checked with public evidence. This includes social media posts, headlines, article claims, public statements, screenshots, image captions, and rumors connected to breaking or renewed news interest.

We do not try to fact-check pure opinions, satire, vague predictions, private claims, or statements where no meaningful public evidence is available.

How claims are selected

Claims can enter FactCheckRadar through the Chrome extension, a reader-submitted source, a traffic spike on an existing report, or operator review.

Research process

For each check, FactCheckRadar searches for relevant public sources and reads accessible page content. Search snippets are treated only as leads. Final reports are based on reviewed source content, not unsupported model memory.

Sources are preferred when they are directly relevant to the claim, current enough for the topic, transparent about dates and authorship, close to the original evidence, and corroborated by other reliable material. For country-specific claims, searches may include the local language when that is more likely to surface primary or authoritative evidence.

Source review

The system separates sources that were merely found from sources that support the final report. Published pages show the sources used for the assessment. Duplicate pages, irrelevant keyword matches, stale unrelated articles, and pages that do not materially address the claim do not drive the verdict.

Verdicts

Verdicts summarize the best available evidence at the time of publication or review. Common outcomes include supported, contradicted, mostly true, misleading, and unverified. A verdict can change if stronger evidence appears later.

Writing and review

FactCheckRadar uses automation and AI-assisted workflows for research, drafting, source organization, and structured publishing. The standard is journalistic: the report should explain the claim, the evidence, the context, the uncertainty, and why the cited sources matter.

AI assistance is used inside a controlled pipeline with source validation and structured database publishing. It is not a license to publish unsupported claims or rely on fluent text without evidence.

Reader-submitted sources

Readers can submit sources for existing reports. Submitted sources are reviewed before they affect a report. A source is useful only if it materially helps confirm, contradict, update, or contextualize the claim.

Reviews and updates

A published report may be reviewed again because a reader submitted a source, the article is receiving renewed attention, a traffic spike suggests the topic is active again, or an operator manually triggers a review. A review can update the verdict, add context, add or remove sources, or confirm that the original assessment still stands.

Image checks

Image-related checks look for source pages, earlier appearances, captions, and context around the image. FactCheckRadar avoids treating visual similarity alone as proof of authenticity, staging, location, or date. The strongest image checks are tied to traceable appearances and source context.

Corrections

Corrections are appropriate when a verdict, source, date, claim description, or important context is wrong or incomplete. Readers can report issues or submit sources, and updated reports may be refreshed when new evidence changes the assessment.

Limits

Fact-checking depends on available evidence. Some claims remain unresolved because evidence is missing, inaccessible, contradictory, or still developing. A FactCheckRadar verdict is a sourced assessment at a specific point in time, not a permanent declaration that can never change.

Founder and editorial responsibility

FactCheckRadar is operated by Roman Polach, a Czech journalist and software developer. Roman previously worked as a freelance and full-time journalist for Czech publications, specializing in science, health, and technology coverage. That background informs the site's emphasis on source transparency, evidence-based conclusions, and clear handling of uncertainty.

Editorial selection and homepage placement

Fact-checks shown on the homepage are currently selected by Roman Polach. Users can also publish fact-checks on the website, but user-published articles are not automatically listed on the homepage. This keeps the homepage editorially selected while still allowing user-generated reports to exist on the site.