1. Purpose
Sources provide the foundation through which readers can evaluate journalism and its supporting evidence. Data Watchdog identifies and attributes relevant source material wherever doing so is compatible with accuracy, safety, confidentiality, applicable law, and the public interest.
Citation is not merely a formatting requirement. A cited source must genuinely support the claim for which it is referenced.
The level of sourcing and verification required increases with the seriousness, novelty, potential risk, and possible harm associated with a claim. Significant allegations, safety claims, legal assertions, scientific findings, financial figures, and claims concerning an individual’s conduct require particularly careful examination.
2. Source Hierarchy and Assessment
Where available, we prefer direct, primary, and contemporaneous evidence.
Relevant primary material may include legislation, court records, official statistics, regulatory decisions, audited filings, contracts, technical documentation, research papers, datasets, source interviews, authenticated correspondence, photographs, video, system records, and direct observation.
Primary evidence is not automatically treated as reliable. A corporate filing may omit relevant context, a government statement may reflect an institutional interest, a technical demonstration may be staged, a witness may be mistaken, and a dataset may contain methodological limitations.
Editors therefore consider provenance, authenticity, independence, expertise, incentives, methodology, timing, completeness, and corroboration.
Secondary sources may be used for context, interpretation, comparison, background research, and reporting leads. Repeated publication does not transform a secondary claim into primary evidence.
Data Watchdog seeks to avoid circular sourcing, unsupported statistics, and claims whose original evidence cannot be reasonably traced.
3. Attribution
Facts, quotations, data, distinctive analysis, images, documents, and other material obtained from external sources should be attributed in a manner appropriate to the format and circumstances.
Where practical, online links should direct readers as closely as possible to the actual supporting source rather than to an unrelated homepage or generic search result.
Quoted material must preserve the original meaning and relevant context.
Minor edits for clarity, such as removing verbal fillers, may be made when they do not materially alter meaning.
Translations should be identified where relevant. Particular care should be taken when translating contested, technical, legal, political, or otherwise sensitive language.
4. Links, Archives, and Access Dates
When a source is publicly available, Data Watchdog aims to link to an authoritative and reasonably stable version where practical.
For sources that may change, disappear, or become inaccessible, we may preserve a lawful archive, screenshot, downloaded copy, or internal verification record for editorial purposes.
Preserving a copy does not by itself authorize republication beyond the applicable legal basis.
Publication dates, update dates, access dates, retrieval dates, or archival capture dates may be identified when they are relevant to understanding the source or interpreting the claim.
Where a material source link becomes unavailable, it may be replaced with an official archive or an equivalent authoritative source, with the change recorded where appropriate.
5. Data and Research
Data based claims should identify relevant information about the source, time period, population, unit of measurement, methodology, and significant limitations.
Correlation should not be presented as causation without sufficient supporting evidence.
Small samples, self selected surveys, unpublished methodologies, vendor funded studies, preprints, and model generated estimates should be identified and interpreted with appropriate caution.
Research findings should be described in proportion to the available evidence.
A single research paper should not automatically be presented as scientific consensus.
Where material to the story, editors may consider retractions, corrections, conflicts of interest, peer review status, methodological concerns, and replication evidence.
6. Company, Government, and Advocacy Materials
Press releases, investor presentations, product demonstrations, white papers, commissioned research, government statements, advocacy reports, and similar materials may contain useful information but are generally treated as interested sources.
Data Watchdog identifies the origin of such material and independently assesses significant claims where reasonably possible.
When a company, founder, campaign, government body, advocacy organization, or other interested party provides substantial documents, data, interviews, travel, access, or editorial materials, the relationship may be disclosed where that information would help readers evaluate the resulting work.
Material supplied by an interested party is not treated as independently verified merely because it has been published or quoted.
7. Anonymous and Confidential Sources
Anonymous sourcing is used selectively and only where there is a legitimate public interest reason.
Generally, the information should be significant enough to justify confidentiality, the source should face a credible reason for requiring protection, the information should not reasonably be obtainable on the record, and the responsible editor should know or be able to verify the source’s identity.
Where safety permits, published material should explain why anonymity was considered necessary.
A single anonymous source will rarely be sufficient for a serious and contested allegation.
Corroboration may include documents, independent witnesses, data, direct observation, or other credible evidence.
Data Watchdog does not permit anonymity to be used simply to facilitate an avoidable personal attack, advance an undisclosed interest, or avoid accountability for an ordinary opinion.
8. User Generated and Social Media Material
Posts, screenshots, direct messages, videos, audio recordings, and other online material should be verified before being relied upon in published journalism.
Depending on the circumstances, verification may include examination of original files, metadata, geolocation, chronology, account history, reverse searching, direct contact with the source, visual analysis, and comparison with independent evidence.
Virality is not evidence of authenticity.
We seek to minimize unnecessary republication of personal information and graphic material. The fact that information is publicly accessible does not remove considerations relating to privacy, safety, consent, or context.
9. Artificial Intelligence and Synthetic Material
Artificial intelligence systems may assist with discovery, research workflows, transcription, translation, data processing, or other editorial tasks, but AI output is not itself treated as a source.
Model generated answers, summaries, citations, quotations, images, or transcripts must be checked against original or authoritative material before publication.
Data Watchdog does not publish AI invented references or treat a model’s confidence as evidence.
Synthetic or materially altered media should be identified when used in a manner where its nature is relevant to readers.
Claims involving deepfakes, cloned voices, generated images, manipulated documents, or other synthetic material require technical and contextual verification proportionate to the potential risk.
10. Source Records and Protection
Data Watchdog may retain source records for as long as reasonably necessary for verification, corrections, legal obligations, security, and editorial accountability, subject to privacy and source protection considerations.
Access to source records should be limited according to the sensitivity of the material and the legitimate need for access.
Promises of confidentiality must be made carefully and only by an authorized editorial representative.
No journalist or editor can guarantee absolute protection against every possible legal demand, device compromise, security incident, or technical failure.
Where material limitations are known, sources should be informed of them where reasonably possible.
11. Corrections and Challenges
A source citation may require correction where it does not actually support the associated claim, identifies the wrong source, omits material context, or directs readers to an incorrect or inappropriate version.
Readers, cited individuals, organizations, and other relevant parties may raise sourcing concerns through:
Correction requests should identify the relevant article, claim, citation, and supporting explanation or evidence where available.
Data Watchdog will assess credible sourcing concerns and make appropriate corrections, clarifications, or updates when warranted.