Features

Part of Source evaluation and claim mapping: a practical guide

Fictional case: correcting a circular citation that became a fact

Circular citation correction case showing how one unsupported claim spread across reports, passed weak checks, and required a transparent article-wide repair.

What to take away

  • Multiple articles can repeat one unsupported origin.
  • A real document may be cited for a claim it never makes.
  • Corrections must reach headlines, graphics, newsletters, and partner copies.
  • Editors should preserve the failed evidence chain and explain the material change.
  • Claim ledgers and source-origin maps prevent the same error from returning.

This fictional case shows a correction workflow. The organization, program, people, and figures are invented. The verification principles are real.

The published claim

News Cod reported that the fictional Northbridge Youth News Project had trained 18,000 students since 2019. The number appeared in the headline, a chart, a newsletter, and a partner excerpt. The article linked to a university center's report and cited three news stories that repeated the same total.

A former program coordinator wrote to say the project had counted workshop registrations, including repeat attendance and canceled sessions, rather than distinct students trained. The coordinator also said the university report did not contain the 18,000 figure.

Freeze the record

The editor paused promotion and saved the published page, chart data, newsletter, partner copy, analytics screenshot, reporter notes, source emails, and every cited page. A temporary note told readers that the participation figure was under review. The article remained visible so the record of publication was not silently erased.

The team rewrote the sentence as separate claims:

  1. Northbridge operated from 2019 through 2026.
  2. Its records contained 18,000 workshop registrations.
  3. Those registrations represented 18,000 distinct students.
  4. The university center independently verified the total.
  5. The program "trained" everyone counted.

Only the first two had immediate documentary support.

Examine what the records say

The Digital Inquiry Group's lesson on using historical thinking skills has students source, contextualize, and corroborate a document, and pairs each activity with lateral reading about an internet source connected to it. Corroboration is the step that failed here: the university report described program goals and sample interviews but offered no cumulative participant total. A genuine report had been linked beside a sentence it did not support.

The project's spreadsheet contained rows labeled registrations, not people. Some email addresses appeared more than once, several group bookings lacked individual names, and canceled sessions remained in the export. The data could establish recorded registrations under an unstable administrative process. It could not establish unique students or completed training.

Map the citation loop

The first local story attributed 18,000 students to a project presentation. A later trade article cited the local story. The university center's web summary cited that trade article for background, without using the number in its report. A third story cited the center's summary and supplied the 18,000 figure from the earlier trade piece. News Cod had treated these pages as separate confirmation.

The team contacted each publisher and obtained the presentation slides. A slide said "18K engagements," with a note defining engagement as a recorded workshop registration or digital activity. The project director confirmed that a later press release had replaced "engagements" with "students trained" without a deduplication study.

Recheck the claim, not the wording

UNESCO's media literacy fact-checking unit asks learners to distinguish fact-checkable claims from opinion, document the checking steps, and critique the research process. That approach kept the review focused on what could be established rather than on whether the promotional phrase sounded plausible.

News Cod recalculated only defensible categories. It reported 12,460 noncanceled workshop registrations from 2019 through 2026, plus 3,120 recorded digital activities. It did not add the categories into people. Because identifiers were incomplete and inconsistent, the number of distinct students remained unknown. "Training" was replaced with "registrations" unless attendance and program completion were separately documented.

Publish the correction

The revised headline removed the 18,000-student claim. The chart was replaced with a table that named each unit and period. The article explained that the earlier figure conflated registrations, digital activity, repeat participation, cancellations, and distinct people.

The correction note included the original wording, corrected wording, date, reason, and scope. Editors updated the newsletter archive, partner excerpt, social cards, chart alt text, and structured data. They notified the project and the publishers whose articles had been used in the loop.

Prevent recurrence

The desk added an origin column to its source ledger, required a quoted passage or table cell for each material citation, and prohibited counting pages as independent sources until their evidence chains were mapped. Participation claims now require a unit, period, inclusion rule, repeat-person treatment, and completion definition.

Common questions

Why not delete the article?

A visible correction preserves accountability and gives people who saw the error a route to the accurate record. Legal or safety needs may require a different decision.

Was the university report false?

No. It was authentic and did not make the disputed numerical claim. The article misused it.

Could 18,000 still be a fair estimate?

Not from the available records. The mixed units and incomplete identifiers did not support a unique-person estimate.

What was the central failure?

The newsroom counted repeated citations instead of tracing them to one altered promotional phrase and checking the underlying unit.

More in Features

Guides

Source evaluation and claim mapping: a practical guide

Source evaluation guide for separating claims, finding original evidence, testing authority and independence, recording limits, and writing accurate conclusions.

Rules

Common source-evaluation mistakes and how to fix them

Source-evaluation mistakes with fixes for surface trust, authority mismatch, circular citations, weak corroboration, bad search habits, and overstated conclusions.

Guides

How to check an unfamiliar online source before trusting a claim

Online source verification workflow for preserving a page, identifying its publisher, reading laterally, tracing evidence, comparing reports, and recording limits.

Reviews

Evidence source types compared: firsthand accounts, records, experts and reports

Evidence source types compared by proximity, scope, independence, strengths, limits, and best uses for checking events, numbers, causes, and quotations.

Latest from Practice Desk

Maintenance

Account and viral-claim verification checklist

Account and viral-claim checklist for preservation, identity, control, badges, origin, reposts, links, evidence, coordination, safety, publication, and correction.

Guides

How to audit a poll or survey story before you cite it

Poll audit workflow for sponsor, population, sample, recruitment, fieldwork, questions, weighting, uncertainty, subgroup comparisons, headlines, and corrections.

Guides

How to check an unfamiliar account and viral post

Account and viral-post verification workflow for confirming identity, tracing origin, checking links and claims, contacting owners, and reporting safely.

Guides

How to geolocate and chronolocate visual media

Visual geolocation and chronolocation workflow using source history, landmarks, maps, weather, shadows, event records, uncertainty ranges, and peer review.