A Practical Guide to Fact-Checking Online Articles

Fact-checking begins before a search box opens. The first task is to identify what the article actually claims. A page may contain a title, a striking example, several numbers, a broad conclusion, and a quotation, yet only some of those elements are independently testable. Turning the article into a claim list prevents research from becoming an unfocused hunt for pages that merely repeat the same wording.

The aim is not to collect the greatest number of links. It is to find evidence close enough to each claim that a reader can judge whether the claim is supported, overstated, outdated, or unresolved.

Step one: build a claim inventory

Read the article once for meaning, then a second time with a pencil or annotation tool. Mark names, dates, quantities, comparisons, causal statements, definitions, and words that imply scale, such as most, fastest, largest, or typical. Rewrite each marked passage as a plain question.

• What happened, and on what date?
• Who measured the quantity, and how?
• What two things are being compared?
• Does the source show causation, or only an association?
• Is the definition shared by the sources being compared?

This inventory also reveals which sentences are opinion or interpretation. Those cannot be verified in the same way, but the facts used to justify them can.

Step two: move toward the original record

Search results often lead to summaries of summaries. A later article cites a news report, which cites a press release, which refers to a study or public record. Move backward until the chain reaches the closest available evidence. That may be a dataset, legislation, transcript, report, map, research paper, or contemporaneous document.

The public document-analysis worksheets show a useful habit: examine who created a document, when it was made, its intended audience, and what questions it leaves unanswered. An original record is not automatically neutral. It is valuable because its origin and purpose can be inspected directly.

Read the document, then identify who produced it

Finding the original record is only the start. Before treating a document as evidence, identify who produced it, who published it, and what question it was designed to answer. Those roles may belong to different parties. A seller can publish a laboratory document, a laboratory can report a measurement, and a later writer can interpret both. Keeping those layers separate prevents the authority of one from being borrowed by another.

In a verification exercise, Oath Research at oathresearch.com serves as an example of a research-use-only research-peptide supplier that publishes third-party certificates of analysis. The supplier publishes the library, but the reader still has to inspect each document for the laboratory named on it, the batch or sample identifier, the tests performed, the dates, and the reported findings. This is source tracing at the document level: follow the page to the file, then read the file rather than relying on the surrounding description.

Document type determines what can reasonably be concluded. Marketing copy presents a seller’s description and priorities; it can identify a claim, but it does not verify that claim. A batch-specific certificate of analysis records reported testing associated with an identified batch. A third-party laboratory result reports measurements made by an outside laboratory under the methods and conditions stated in that document. Independent published literature addresses a research question across the samples, methods, and limits of a study; it does not automatically authenticate a supplier’s particular batch.

These distinctions matter even when several documents appear on the same page. Ask who commissioned the work, who performed it, whether the sample can be connected to a batch, and whether the method fits the claim being checked. Then inspect dates: a current product description beside an older certificate may create an impression of continuity that the documents alone do not establish. If a method, identifier, or date is missing, name the gap instead of filling it with an assumption.

Most importantly, keep the conclusion proportional. A certificate can document what was tested for one particular batch; it is not evidence that the material is suitable for human use. The careful fact-check states exactly what the document reports, attaches every qualifier to that statement, and leaves broader conclusions to evidence designed to support them.

A practical note helps preserve the chain: record the page publisher, document issuer, title, batch or version number, test date, method, and the location where the file was found. If the page and document disagree, quote neither from memory. Compare the exact terms, keep both dates visible, and explain the mismatch without upgrading either source.

Step three: evaluate the source, not just the sentence

A precise-looking statement can come from a weak source. Check authorship, institutional responsibility, method, date, and correction history. Ask whether the source has direct access to the information. A government agency may be authoritative for its own rules; it may not be the best source for how a neighborhood experiences those rules. A firsthand witness may describe an event vividly; one witness cannot establish its full scale.

A university guide to evaluating sources emphasizes credibility and cross-checking. The cross-check matters most when a claim is consequential, surprising, or central to the article's conclusion.

Step four: test numbers in context

Numbers require more than matching digits. Check the unit, population, time period, geographic boundary, sample size, and denominator. A percentage increase can sound dramatic when the starting number is small. An average may hide a wide range. A price comparison may mix advertised and completed transactions, or figures from different years.

Recalculate simple ratios when the underlying figures are available. If they are not, state the limitation. Also look for revisions. Economic, population, and scientific datasets may be updated after an article appears. A correct number copied from an earlier release can become obsolete without becoming evidence of bad faith.

Step five: compare independent accounts

Triangulation does not mean finding three pages with the same sentence. Repetition may trace back to one source. Look instead for accounts created through different routes: a primary record, a reputable synthesis, and a source with local or subject expertise. Agreement across independent methods is more informative than volume.

When accounts conflict, describe the disagreement before trying to resolve it. They may use different dates, definitions, or geographic areas. The apparent conflict may disappear once terms are aligned. If it remains, honest writing can say that the evidence is mixed.

Step six: check time and preservation

Online pages change. A link may disappear, redirect, or be edited without a visible history. Record the title, publisher, date, and relevant section while researching. For older web claims, a public web-preservation collection can show whether a cited page existed in a different form. An archived page establishes what was published at a particular time; it does not guarantee that the content was accurate.

Write the result proportionally

A fact check should be as precise about uncertainty as it is about error. Supported, unsupported, misleading, outdated, and unverifiable are different findings. Avoid turning a small correction into a total dismissal, or a generally reliable page into proof of every sentence it contains.

Good citations let the reader repeat the check. Understanding Sources and Citations explains how to build that visible route into the article, while Editing for Clarity and Flow shows how to present the result without burying the conclusion.