Media Watch Methodology

Relationship to the main methodology

The main methodology page describes three principles that govern all parliamentary data on this site: official sources first, facts not interpretations, and full provenance. Media Watch operates within those constraints but extends them. Reporting quality cannot be verified by inspecting an official parliamentary record — it requires a third-party assessment. We therefore apply one additional rule: every case must rest on a named external assessor whose verdict we record and attribute, never adopt in our own voice.

The practical consequence is that our summaries describe what an assessor found, not what the site believes. If a court ruled a claim false, we write "the court ruled the claim false." If a fact-checker rated a headline misleading, we write "the fact-checker rated the headline misleading." We do not paraphrase those findings into characterisations of the outlet's character, motives, or political orientation.

1. Inclusion criterion

A case is included only if an independent third party has assessed the reporting. Qualifying assessors are:

• A national or regional press council (Presserat ruling, Rüge, or Missbilligung);
• A court with jurisdiction over the claim (injunction, Urteil, or helyreigazítás order);
• A professional fact-checking organisation (CORRECTIV, dpa-Faktencheck, Lakmusz, Mimikama, and equivalents with documented editorial standards);
• A peer-reviewed academic study that assesses specific named reporting.

Party complaints, social-media reactions, and opinion articles — however widely cited — are not qualifying assessments. A case where only the affected party called the reporting false is excluded until an independent body makes a finding.

Cases where the qualifying URL is missing or no longer resolvable are moved to a quarantine file and removed from the live catalogue until a working reference is restored.

2. What the assessment labels mean

Each case carries one of four assessment labels. The label reflects the assessor's finding, not our own reading of the material.

incorrect

The assessor found a factual statement that was demonstrably false — a fabricated quote, a non-existent law, a statistic that contradicts its stated source. The distinguishing mark is that the claim fails on verifiable facts alone, regardless of framing.

Example: Bild's 2023 headline "Habeck will Energie-Stasi einsetzen" asserted that the Gebäudeenergiegesetz would create inspectors entering private homes. No such enforcement mechanism existed in the draft law. The Otto Brenner Stiftung media study documented the claim as factually absent from the legislation.

misleading

The assessor found that individual statements were technically accurate but that the overall framing, omission of context, or presentation created a false impression. The distinction from incorrect: the underlying facts are real, but their arrangement deceives.

Example: Bild's 2023 headline "Habeck will Öl- und Gas-Heizungen verbieten" implied an immediate blanket ban. The GEG amendment was actually a phased transition for new installations from 2024 with long transition periods for existing systems. CORRECTIV rated the framing misleading because it omitted the phase-in timeline and exemptions.

correct

An assessor reviewed the reporting and found it accurate, or a court dismissed a challenge to reporting that an actor had contested. These cases appear in the catalogue to document independently confirmed accurate reporting.

Example: ARD's 2024 reporting on the Potsdam meeting (CORRECTIV investigation) was challenged by AfD politicians as distorted. The Hamburg District Court confirmed CORRECTIV's factual reporting was accurate and dismissed the AfD's legal challenge.

opinion

A court, press council, or assessor found that the content was protected opinion or commentary, not a factual claim capable of being true or false. These cases are documented where the opinion/fact distinction itself was adjudicated.

3. Evidence tiers A / B / C

Evidence tiers reflect the institutional weight of the assessor, not the severity of the finding. A Tier-C case can document a serious factual error; the tier only tells you how much independent scrutiny the assessor itself has been subject to.

Tier A — formal adjudication

A court ruling (LG, OLG, BGH, BVerfG, Törvényszék, ECHR) or a press council ruling (Deutscher Presserat Rüge, Österreichischer Presserat, formal finding). These bodies have defined procedures, opportunity-to-respond requirements, and public accountability. The ruling itself is the public record.

Tier B — professional fact-check

An assessment by an organisation with documented editorial standards for fact-checking: CORRECTIV, dpa-Faktencheck, Lakmusz, Mimikama, Volksverpetzer, Übermedien, tagesschau Faktenfinder, MDR Faktencheck, AFP, and equivalents. These are organisations that publish their methods, accept corrections, and apply stated criteria. Tier B is included and labelled as a professional assessment; it is not equivalent to a court ruling.

Tier C — academic study or documented media critique

A peer-reviewed study, institutional research report, or documented media critique (Bildblog, Kobuk, apabiz, Otto Brenner Stiftung, OSCE/ODIHR election observation reports). Tier C is included but carries the lowest institutional weight: the study may document a pattern rather than a specific error, or the research body may not apply the same procedural safeguards as a court or press council. Every Tier-C case is labelled as such in the interface.

4. Selection-bias disclosure

The cases in this catalogue are those that a press council chose to rule on, a fact-checker chose to investigate, or a court was asked to adjudicate. That set is shaped by:

• Which outlets attract press council complaints (complaints must be filed; not all affected parties file);
• Which claims fact-checkers chose to investigate (they have limited capacity and tend to focus on high-circulation, viral, or politically salient claims);
• Which parties or individuals had the resources and legal standing to pursue court rulings;
• The countries and time periods our ingestion pipeline covers.

A party that appears frequently in the "effect" column may simply have been the subject of more assessor attention during the period covered, not necessarily the target of more false claims in absolute terms. The inverse is also true: absence from the catalogue does not mean an outlet has a clean record — it may mean its coverage was not assessed by a qualifying body in our dataset.

The country-level tables in the Media Watch section make clear which date range and which assessors are included. See the Media Watch tables for the current coverage scope.

5. Effect vs. intent

Each case records a claim_effect: whether the assessed claim harmed, benefited, or had a neutral effect on the party it mentioned. This field records the effect of the claim as established by the assessor — not an inference about the newsroom's intent.

We do not state or imply that a journalist or outlet intended to deceive or harm a party. If the assessor's own published finding uses language about intent ("deliberately misleading," "designed to deceive"), we quote those words and attribute them to the assessor. We never add motive language in our own voice.

The wording "harms [party]" in the effect field means: the assessed claim, if accepted by a reader, would reduce that party's standing. It describes the directional consequence of the false or misleading content, not the outlet's motivation for publishing it.

6. Corrections and vindications

Reporting that is later independently confirmed as accurate is catalogued with the label correct. There is no target quota for any label. The distribution of labels in the database is whatever the assessed evidence yields — we do not add correct cases to balance incorrect cases, and we do not suppress incorrect findings to make the overall picture look more balanced.

When an outlet issues a formal correction for a case already in the database, the corrected flag is set to true and the correction date is recorded. The original finding remains in the record; the correction is an additional data point, not a retroactive removal.

Cases where a legal challenge to reporting was dismissed — confirming the reporting was permissible — are documented as correct with the court ruling as the assessor. This is not an editorial endorsement of the reporting; it is a record of the judicial outcome.

7. How to challenge a case

If you believe a case in this catalogue is factually wrong, misattributed, or should be removed, contact us at patrick.fodor.sc@gmail.com.

To be considered, a challenge must include:

1. The case ID (e.g., mr-128) and the specific factual claim you contest;
2. A resolvable source (URL, court Aktenzeichen, Presserat file number, or equivalent) that contradicts or corrects the information in our record;
3. If you believe the assessor's finding itself is wrong: the name of the assessor, the finding you dispute, and a primary source that contradicts it.

We do not remove cases based on the affected outlet's objection alone, because the outlet being assessed is not an independent assessor of its own accuracy. Challenges from third parties with independent evidence are treated the same way as original case submissions.

All corrections and removals are logged in the public correction record committed to the repository. When a case is amended, the before-and-after change and the reason for the change are documented. When a case is removed, the reason and the source that prompted removal are recorded.

The human-readable correction log is published on the Media Literacy page (correction log section).