Eight common source evaluation mistakes when checking a claim

What to take away

Mistake 1: rating the outlet instead of the claim

Calling a publisher reliable or unreliable skips the work. Strong outlets publish corrections. Advocacy groups publish authentic documents. Agencies describe their own actions accurately and their motives contestably.

Score each material claim on its own: who made it, by what method, on what evidence, as of what date, and for whose benefit. Then write a conclusion about the evidence, not a permanent verdict on the source. Map the claim to its origin first, using the claim mapping guide.

Mistake 2: treating authority as universal

A title proves a role, not access to this event. A chief executive knows company policy and may never have visited the site in the complaint. A scientist qualified in one specialty can be a layperson in the next.

A national study of students' civic online reasoning put 3,446 high school students in front of unfamiliar sites. Most judged by surface markers: design, top-level domain, or the site's own About page. Ninety-six percent did not treat a climate site's fossil fuel funding as relevant to its credibility.

For each claim, ask what the person directly observed, which records they read, and what their method cannot see. Then compare that against the types of evidence sources available to you.

Mistake 3: counting copies as corroboration

A local story, a national story, a blog post, a video and a social thread can all trace to one wire report or one press release. Agreement across formats is not independent confirmation.

Trace each item backward to the document, interview, dataset or observation beneath it. Count origins, not appearances. If every path ends at one unnamed source, write that down.

Mistake 4: searching only the claim's wording

Exact phrasing finds copies, and it also traps you inside the promoter's vocabulary. The record may sit under a legal name, a technical term, a different unit, or an earlier project title.

Change the search terms with the record type:

Scan the results page before clicking. When the record type changes, change databases.

Mistake 5: inspecting only the original website

About pages, logos, footnotes and staff biographies are written by the publisher. A fraudulent or partisan site can imitate every scholarly signal on that list.

Stanford Graduate School of Education reported that lateral-reading instruction improved online source judgments among students taught a civic online reasoning curriculum. The habit that moved the needle was leaving the page to investigate its publisher elsewhere.

Open independent tabs early. Search the publisher, the author, the ownership, the funding and the quoted material. Come back to the page once you have outside context.

Mistake 6: confusing authenticity with truth

A genuine document can carry a false allegation. An unedited video can carry a false date. A real dataset can be paired with the wrong denominator.

Run three separate checks and record which question each one answers:

A pass on the first is not a pass on the third.

Worked example: one number, five pages

A claim says a city's water tests found unsafe lead levels. The search returns a local news story, a national aggregation, a blog post, a video and a thread. All five cite the same number.

Pull the original post or upload, not the screenshot. The number traces to one press release from an advocacy group. The release cites a monitoring report. The report lists a detection level, not a violation, and covers one building, not the system.

Now the claim has three parts: the detection (supported), the violation (not supported by that report), and the system-wide reading (unsupported). Each part gets its own line. The checklist for an unfamiliar source walks through that separation.

Mistake 7: seeking only confirmation

Once a conclusion forms, searches start defending it. Contradictory records become exceptions, and ambiguous wording gets read one way.

Write a disconfirmation question before you finish: what evidence would make this wrong, or narrower? Search for that. Note the conflicts in the draft. Ask a second reviewer to attack the chain.

Mistake 8: forcing a rating

Some claims are vague, subjective, private, future-facing, or simply not covered by any available record. A tidy rating can pretend to settle what the evidence cannot.

Classify honestly: unsupported, unresolved, not checkable, or outside scope. Say what you examined and what new evidence would change the answer.

Common questions

Is Wikipedia useful during verification?

Yes, for orientation and for its reference list. Treat its citations as leads to the records underneath, not as the evidence itself.

Does bias make a source useless?

No. Interests shape what a source knows and stresses. Identify them, test the evidence, and do not ask a source to prove a claim outside its access.

Are search snippets evidence?

No. They are partial previews, sometimes outdated, sometimes assembled from text that does not support the apparent claim. Open the page.

What is the fastest useful fix?

Split the sentence into separate claims, then trace every apparent confirmation back to its origin. Most errors surface in that first pass.