How Prism Writing verifies, sources, and corrects content.
Purpose
Prism Writing is a research and education platform. We publish long-form articles, breaking news summaries, and explanatory resources. This policy sets the minimum standard for sourcing, attribution, and corrections across all content types.
These standards apply to every piece of content we publish, whether it is a 10,000-word investigative report, a 300-word breaking news brief, a data visualization, or an educational module. They are binding on all editors, contributors, and freelancers who create content under the Prism Writing banner.
Our editorial philosophy is rooted in the belief that readers deserve transparency about how information is gathered, verified, and presented. We treat our audience as intelligent participants in the information ecosystem, not passive consumers.
Sourcing & Attribution
Every factual claim should be supported by at least one primary or high-quality secondary source.
All articles and news summaries must include direct source links where feasible.
We label anonymous or unverified sources clearly and avoid unsupported claims.
Images must include proper licensing, attribution, or confirmed public-domain usage.
We maintain a four-tier source hierarchy for evaluating reliability:
Tier 1 — Primary Sources: Official government records, court filings, peer-reviewed academic research, original datasets, and direct firsthand accounts. These are given the highest weight in any factual determination.
Tier 2 — Authoritative Secondary Sources: Reporting from established news organizations with strong editorial standards, published books from reputable academic or commercial publishers, institutional reports from recognized bodies (e.g., WHO, IPCC, CBO).
Tier 3 — Contextual Sources: Expert commentary, opinion columns from credentialed specialists, policy analyses from think tanks (with ideological positioning disclosed), and historical accounts. These sources help contextualize a story but are not relied upon as sole evidence for factual claims.
Tier 4 — Unverified or Social Sources: Social media posts, press releases, anonymous tips, and user-generated content. These are never presented as fact without independent corroboration. If used, they are explicitly labeled as unverified.
When multiple sources conflict, we present the strongest evidence and note the disagreement rather than selecting the most dramatic version of events. If a story cannot be adequately sourced, we do not publish it until sourcing improves.
Fact-Checking
Claims of law, policy, or regulation are reviewed against official government or statutory sources.
Quantitative statements are cross-checked against at least two independent sources when available.
Breaking news items are flagged as developing until independently corroborated.
Our fact-checking process operates at three stages:
Pre-publication review: All factual claims are extracted and individually verified before an article goes live. Statistical claims require at least two independent data sources. Legal claims are checked against statutory text, case law, or official regulatory guidance.
Post-publication monitoring: Published articles are monitored for reader corrections, source updates, and breaking developments that may affect accuracy. Articles on developing stories are revisited within 48 hours of publication.
Periodic re-verification: Evergreen and high-traffic articles are re-checked on a quarterly basis to ensure continued accuracy as new information becomes available, laws change, or statistics are updated.
When we cannot independently verify a claim but believe it is newsworthy, we publish it with clear caveats: "According to [source], [claim]—Prism Writing has not independently verified this statement." We never present unverified claims as established fact.
Editorial Independence
We do not accept paid placement in news or research. Sponsorships, if any, are labeled clearly and separated from editorial judgment.
Our editorial decisions are made solely on the basis of news value, public interest, and educational merit. No advertiser, donor, sponsor, political organization, or external entity has the ability to influence what we cover, how we cover it, or what conclusions we draw.
As a worker-owned cooperative, our governance structure ensures that editorial independence is protected institutionally, not merely by policy. Editors cannot be overruled by business interests because the editorial team participates in cooperative governance on equal footing.
Conflict of Interest Policy
All contributors, editors, and staff are required to disclose any financial, personal, or professional relationships that could reasonably be perceived as influencing their editorial judgment.
Financial interests: Contributors may not write about companies, organizations, or financial instruments in which they hold a material financial interest (stocks, consulting contracts, board positions) without disclosure. If disclosure would compromise the article's credibility, a different contributor will be assigned.
Personal relationships: Contributors must disclose personal or family relationships with subjects of their reporting. Close personal relationships with sources in a story may require recusal.
Political activity: Contributors covering politics or policy must disclose active political involvement (campaign donations above $200, volunteer work, party membership) relevant to their coverage area. We do not prohibit political participation, but we require transparency.
Prior employment: Contributors who previously worked for an organization they are now covering must disclose that relationship in their byline or author note.
Gifts and hospitality: Contributors may not accept gifts, travel, or hospitality valued above $50 from sources or subjects of coverage. Any complimentary access required for reporting (e.g., event tickets for review) must be disclosed.
Secondary employment: Freelancers and contributors with other professional commitments must disclose them if they intersect with Prism Writing coverage areas.
Conflicts are reviewed by the editorial team. Depending on severity, the resolution may range from in-article disclosure to reassignment to a different contributor. Undisclosed conflicts discovered after publication will result in a published correction noting the conflict, and may result in disciplinary action up to termination of the contributor relationship.
AI Use & Disclosure
Transparency commitment: We use AI tools (including large language models like Claude and GPT) to assist with research synthesis, fact-finding, and initial drafting. However:
Human oversight: Every article is reviewed, fact-checked, and edited by a human editor before publication.
Source verification: All factual claims are verified against primary sources, regardless of how the article was drafted.
Editorial control: AI suggests, humans decide. Final editorial judgment rests with human reviewers.
No autonomous publishing: No AI-generated content is published without human review and approval.
AI Transparency Policy — Detailed Disclosure
We believe readers have a right to understand exactly how AI is used in the creation of our content. This section provides granular detail about our AI practices.
Where We Use AI
Research synthesis: AI helps summarize large volumes of source material (government reports, academic papers, legal filings) into structured research notes for human journalists.
Initial drafts: For certain article types (news summaries, data-driven reports), AI may produce a first draft that is then substantially rewritten, fact-checked, and edited by humans.
Headline and summary generation: AI may suggest headlines, meta descriptions, and article summaries. All suggestions are reviewed and approved or rewritten by editors.
Data analysis: AI assists in processing datasets, identifying statistical patterns, and generating visualizations that are then reviewed by editors for accuracy.
Translation and localization: When we translate content, AI translation serves as a starting point that is reviewed by human linguists.
Where We Do Not Use AI
Editorial judgment: Decisions about what to cover, what angle to take, and what conclusions to draw are always made by humans.
Source evaluation: Assessing source credibility and deciding whether a source is reliable enough to cite is a human responsibility.
Ethical judgments: Decisions about whether to name individuals, publish sensitive information, or use graphic content are made by human editors following our ethics guidelines.
Final publication approval: No article is published without explicit human sign-off.
AI Disclosure Labels
Every article published on Prism Writing carries one of the following AI involvement labels:
AI-Assisted Research: AI tools were used to gather, organize, or summarize research materials, but the article was written and edited by humans.
AI-Assisted Draft: An AI-generated draft served as a starting point, but the final article was substantially rewritten and verified by human editors.
Human-Only: No AI tools were used in the research, writing, or editing of this article.
Why we use AI: As a small independent publication, AI tools allow us to research and synthesize complex topics more efficiently, freeing resources for deeper investigative work and fact-checking. This is similar to how traditional newsrooms use research assistants or wire services—AI accelerates research, but humans verify every claim.
Limitations: AI can make mistakes, hallucinate sources, or misinterpret context. That's why our verification methodology requires cross-checking all claims against primary sources, regardless of how the article was researched.
Content Workflow & Quality Assurance
We maintain professional editorial standards through systematic content workflow:
Draft Management: All articles undergo draft review with version control before publication.
Scheduling System: Content moves through defined stages (DRAFT → SCHEDULED → PUBLISHED → ARCHIVED) with editorial checkpoints.
Quality Metrics: Articles display transparent quality indicators including reading time, source count, verification status, and depth level.
Version History: Major revisions are tracked with timestamps to maintain accountability.
Minimum Standards: Published articles must meet word count thresholds, cite credible sources, and pass editorial review.
Every article passes through the following checkpoints before publication:
Pitch approval — Topic is evaluated for news value, public interest, and fit with our editorial mission.
Research completion — All primary and secondary sources are gathered and documented.
Draft completion — Initial draft is complete with all claims sourced inline.
Fact-check pass — Independent verification of all factual claims, statistics, and quotes.
Copy edit — Grammar, style, clarity, and consistency review.
Legal review — For sensitive topics: review for defamation risk, privacy concerns, and fair use compliance.
Final sign-off — Senior editor approval before scheduling.
Content Rating System
All content published on Prism Writing is classified using a maturity rating system and quality indicators to help readers make informed decisions about what they read.
Maturity Ratings
GENERAL
Suitable for all audiences. No content warnings required. Covers all standard news, education, and research content.
TEEN
May contain mild language, discussion of violence, or sensitive topics such as war, crime, or substance use. A content advisory is displayed at the top of the article.
MATURE
Contains graphic descriptions, strong language, or in-depth discussion of adult themes. Requires age verification (18+) and explicit opt-in through account settings.
Quality Indicators
Each article displays the following transparent quality signals:
Source count: The number of distinct sources cited in the article.
Verification status: Whether the article has undergone full fact-check review, partial review, or is still in the initial verification stage.
Reading time: Estimated reading time based on word count and content complexity.
Depth level: Whether the article is a brief summary, standard report, deep dive, or comprehensive investigation.
Last verified date: When the article's factual claims were most recently re-checked.
Corrections and Retractions
Substantive errors are corrected promptly. Significant corrections are logged on our Corrections page with a timestamp and explanation.
Correction Categories
Minor corrections: Typographical errors, broken links, formatting issues, or minor factual clarifications that do not change the substance of the article. These are fixed silently but logged internally.
Standard corrections: Factual errors that affect a specific claim but do not alter the overall conclusions of the article (e.g., incorrect date, misstated statistic, wrong title for a source). These are corrected with an appended correction notice at the bottom of the article and logged on our public Corrections page.
Major corrections: Errors that materially affect the article's conclusions, fairness, or reliability. These are corrected with a prominent notice at the top of the article, a detailed explanation of what was wrong and what is now correct, and a public log entry.
Retractions: If an article is found to be fundamentally flawed, based on fabricated sources, or so inaccurate that correction is insufficient, we retract the article entirely. Retracted articles are replaced with a retraction notice explaining why the content was withdrawn. The original text is preserved in our internal archives for accountability but is no longer publicly accessible.
Correction Process
Error is reported (by reader, editor, or automated monitoring).
Editorial team verifies whether an error exists and assesses severity.
Correction is drafted and reviewed by at least one editor who was not involved in the original article.
Correction is published with a clear timestamp and categorization.
Original author is notified and given the opportunity to review the correction.
Correction is added to the public Corrections log.
Reader-Submitted Corrections
We actively encourage readers to report errors. Corrections can be submitted through our contact page, by emailing corrections@prismwriting.com, or through the "Report an error" link available on every article. We acknowledge receipt of error reports within 48 hours and publish corrections within 5 business days of verification.
Source Protection and Anonymity
Protecting confidential sources is a fundamental principle of journalism and a commitment we take seriously. We will not reveal the identity of a confidential source without their explicit consent, even under legal pressure.
When We Grant Anonymity
Anonymous sourcing is used sparingly and only when all of the following conditions are met:
The information is of significant public interest and cannot be obtained through on-the-record sources.
The source faces credible risk of retaliation, harm, or legal jeopardy if identified.
The source is in a position to have direct knowledge of the information they are providing.
The information has been corroborated by at least one additional source or piece of documentary evidence.
How We Describe Anonymous Sources
When granting anonymity, we provide readers with as much context as possible about why the source is credible without revealing their identity. We describe the source's basis for knowledge (e.g., "a senior official in the department with direct knowledge of the decision") and explain why anonymity was granted.
Technical Protections
We offer secure communication channels (encrypted email, Signal) for sources who need to communicate confidentially.
Notes and communications with confidential sources are stored securely with access limited to the reporter and their editor.
We do not log metadata (IP addresses, timestamps) for communications received through our secure tip line.
Source identities are never stored in our content management system or shared with third-party services.
Reader Trust Principles
Trust is earned through consistent, transparent practice. These principles guide every interaction between Prism Writing and our readers:
Show our work: We link to sources, explain our methodology, and provide enough context for readers to evaluate our claims independently. We do not ask readers to trust us blindly.
Admit what we don't know: When evidence is incomplete or ambiguous, we say so explicitly rather than overstating certainty. Uncertainty is not a weakness; pretending certainty is.
Separate fact from analysis: News reporting and opinion/analysis are clearly labeled and visually distinct. Readers should always be able to tell whether they are reading a factual report or an interpretive piece.
Welcome scrutiny: We publish our editorial standards, correction history, and funding sources because we believe accountability is strengthened by transparency.
Correct visibly: We do not silently alter articles to make ourselves look better. Corrections are documented and visible, because a publication that hides its mistakes cannot be trusted to report honestly.
Respect reader intelligence: We do not use manipulative headlines, misleading framing, or emotional bait to drive engagement. We trust that substantive content will find its audience.
Protect reader data: We collect only the minimum data necessary to operate the service. We never sell reader data or use it for behavioral advertising. See our Privacy Policy for full details.
Free access: All published content is accessible without paywalls. We believe information access should not be determined by ability to pay.
Data Journalism Standards
Data-driven reporting requires its own set of rigorous standards beyond traditional fact-checking. When we publish data journalism, we adhere to the following practices:
Data Sourcing and Quality
Primary data sources: We prefer data from official government databases, peer-reviewed research, and established institutions. When using alternative data sources, we disclose the source and any known limitations.
Data provenance: We document the origin, collection methodology, and date range of every dataset used in our reporting.
Data cleaning: When we modify raw data (removing outliers, normalizing values, filling missing entries), we document all transformations and explain our rationale.
Sample sizes: We always report sample sizes and confidence intervals when presenting survey or statistical data. We do not present findings from inadequate samples as definitive.
Visualization Standards
Accurate axes: Chart axes start at zero unless there is a clear editorial reason to truncate, in which case the truncation is clearly indicated visually and explained in the caption.
Color accessibility: Data visualizations use colorblind-friendly palettes and include alternative text descriptions for screen readers.
Context: Visualizations include sufficient context (title, axis labels, source citation, date range) for a reader to understand the data without reading the accompanying article.
Interactive elements: Interactive charts and graphs provide keyboard navigation and include static fallback images for assistive technology users.
Reproducibility
Where possible, we make our datasets, analysis code, and methodology available so that readers and other journalists can verify our work. Data journalism that cannot be independently reproduced is labeled as such, with an explanation of what limits reproducibility (e.g., proprietary data, source protection).
Community Content Guidelines
Prism Writing hosts community discussions, comments, and user-generated contributions. These guidelines govern all community participation on our platform.
What We Expect
Good faith engagement: Comments and discussions should contribute constructively to the conversation. Disagreement is welcome; hostility is not.
Evidence-based discussion: When making factual claims in comments, we encourage citing sources. Unsupported claims presented as fact may be flagged by moderators.
Respect for diverse perspectives: Our readership spans a wide range of political, cultural, and personal backgrounds. Engage with ideas, not identities.
Stay on topic: Comments should be relevant to the article or discussion thread. Off-topic comments may be moved or removed.
What Is Not Permitted
Hate speech, slurs, or content that attacks individuals or groups based on protected characteristics.
Harassment, threats, doxxing, or intimidation of other users, sources, or subjects of articles.
Spam, commercial solicitation, or AI-generated comments posted without disclosure.
Deliberately misleading information presented as fact (distinct from honest error or good-faith disagreement).
Personal attacks on other commenters, authors, or editors.
Content that violates any applicable law, including defamation, copyright infringement, or incitement.
Moderation Approach
We use a combination of automated filters and human moderation. Our moderation philosophy prioritizes inclusivity and civil discourse over maximum engagement. We would rather have a smaller, thoughtful community than a large, hostile one.
First-time violations typically receive a warning with an explanation.
Repeated violations may result in temporary suspension (7 days for second offense, 30 days for third).
Severe violations (threats, doxxing, hate speech) result in immediate permanent ban.
All moderation decisions can be appealed through our contact page.
Reader Analytics & Engagement
We collect anonymized engagement data to improve content quality:
Reading Behavior: Track scroll depth, reading time, and completion rates to understand which content resonates.
Privacy-First: All analytics are aggregated and anonymized. We do not create individual reader profiles or sell data.
Editorial Use: Analytics inform future topic selection and content depth, not algorithmic manipulation.
We are explicit about the boundary between analytics and surveillance: we measure content performance, not individual behavior. We do not build individual reader profiles, serve personalized advertising, or use engagement data to create addictive feedback loops.