Paraluman Verify checks public claims, monitors government commitments, investigates misinformation, examines AI-generated and manipulated content, and compares public promises with real-world results. This methodology turns those stated functions into a transparent and repeatable editorial process.

​

It is designed around the International Fact-Checking Network’s commitments to nonpartisanship, fairness, source transparency, organizational transparency, methodological transparency, and open corrections. Alignment with these principles is an editorial objective and does not, by itself, represent or imply IFCN certification.

1. Our purpose

​

Paraluman Verify exists to help the public determine:

• whether a factual claim is accurate;

• whether important context has been omitted or distorted;

• whether an image, video, recording, document, or online post is authentic;

• whether a public promise has been delivered;

• what reliable evidence shows—and what remains unknown.

​

Our role is not to tell people what political position to take. We assess evidence, explain uncertainty, expose manipulation, and hold every claimant to the same standard.

Accuracy before speed

We aim to publish promptly, especially when misinformation may cause immediate harm. However, we will not assign a verdict until the available evidence has been examined and the work has passed editorial review.

​

Nonpartisanship and fairness

We apply the same standards to equivalent claims regardless of the claimant’s political affiliation, public status, popularity, ideology, nationality, institutional power, or relationship with Paraluman Verify.

We do not select claims merely to create artificial political balance. Selection is based on public importance, reach, potential harm, accountability value, and whether the claim can be meaningfully verified.

​

Evidence over authority

No claim is accepted solely because it comes from a government agency, political leader, corporation, academic, media organization, influencer, or other authority. Official records may be strong evidence, but their accuracy, scope, definitions, and limitations must still be examined.

​

Transparency and replicability

Whenever possible, we identify and link to the documents, datasets, recordings, research, correspondence, and other evidence used. Readers should be able to understand how we reached our conclusion and reproduce the central steps of our analysis.

​

Context

A statement can contain accurate words or numbers while creating a materially false impression. We therefore assess not only isolated wording, but also relevant context, definitions, time periods, comparisons, omissions, and reasonable implications.

​

Right of reply

Where practical, we contact the person or organization responsible for a claim and ask for supporting evidence or clarification. A failure to respond is disclosed but is not treated as proof that the claim is false.

​

Independence

Editorial conclusions are made by the Paraluman Verify editorial team. Funders, advertisers, partners, technology providers, claimants, political actors, and outside organizations may not determine or approve our verdicts.

​

Open corrections

We correct material errors promptly, visibly, and honestly. We do not conceal substantive changes through silent editing.

Paraluman Verify may examine factual claims made or distributed by:

• elected officials, candidates, political parties, and government agencies;

• public institutions and state-owned entities;

• businesses, professional associations, and civil-society organizations;

• news organizations, commentators, celebrities, and influencers;

• websites, social-media accounts, messaging channels, and online communities;

• images, videos, audio recordings, documents, advertisements, and AI-generated content;

• members of the public when a claim has substantial reach or public significance.

 

We may also investigate recurring narratives rather than a single post when several pieces of content advance the same identifiable factual proposition.

 

Claims we normally do not rate

​

We generally do not assign factual verdicts to:

• personal opinions, preferences, or value judgments;

• statements whose meaning is too vague to test;

• purely hypothetical claims;

• forecasts that cannot yet be evaluated;

• campaign rhetoric without a concrete factual proposition;

• obvious jokes, parody, or satire that is not reasonably likely to be mistaken for genuine information;

• minor slips that have little public significance;

• private disputes with no clear public-interest justification.

​

An opinion may still be examined when it depends on a specific factual premise. A prediction may later become checkable once the relevant date or event has passed. Satirical material may receive an explanatory label when it is widely presented as authentic.

Paraluman Verify monitors public speeches, news interviews, official statements, legislation, government publications, advertisements, websites, social platforms, messaging applications, viral posts, reader submissions, and other publicly available sources.

​

Editors prioritize claims by considering the following factors.

 

Public consequence

We consider whether the claim could influence elections, public policy, public spending, access to services, public safety, health decisions, legal rights, or democratic participation.

 

Potential harm

We consider whether an inaccurate claim could cause physical, financial, electoral, reputational, institutional, or social harm.

​

Reach

We examine whether the claim has already reached a substantial audience or is spreading quickly across several platforms.

​

Potential reach

A claim may be prioritized before it becomes widely viral when it is likely to be repeated by influential accounts, public officials, media organizations, or coordinated networks.

 

Accountability value

Claims made by people and institutions with public power or responsibility generally receive greater scrutiny.

​

Checkability

The statement must contain a sufficiently specific factual proposition that can be tested using evidence.

​

Timeliness

We consider whether clarification is needed while the topic remains relevant and before the inaccurate information causes additional harm.

​

Persistence

Claims that repeatedly circulate after correction may be checked again, especially when they appear in a new context or reach a different audience.

 

Information gap

We prioritize work that provides evidence, explanation, or context not readily available to the public.

​

Available evidence

We consider whether the claim can be investigated responsibly using accessible records, qualified expertise, appropriate tools, and safe reporting practices.

 

Claims are not selected because a person or organization has paid Paraluman Verify, donated to it, advertised with it, or requested a favorable conclusion.

Step 1: Capture and preserve the claim

​

We record the claim as it appeared when identified.

​

Depending on the format, we may preserve:

• the exact quotation;

• the original post or publication;

• the relevant URL;

• the name of the account, person, or organization responsible;

• the publication date and time;

• screenshots;

• archived copies;

• the full speech, interview, article, or recording;

• visible engagement figures;

• copies of relevant documents or media files;

• available metadata.

​

For content that may be altered or deleted, we may preserve a screen recording, archived version, file hash, or other evidentiary copy.

​

Step 2: Define the checkable proposition

​

We identify the central factual statement that a reasonable audience would understand the claimant to be making.

​

The proposition should specify, where relevant:

• who or what the claim concerns;

• what allegedly happened;

• the relevant location;

• the relevant period;

• the quantity, percentage, or comparison involved;

• the definitions being used;

• the outcome or causal relationship being asserted.

​

A statement containing several independent factual claims may be divided into separate fact-checks.

​

We do not combine unrelated propositions into a single averaged verdict.

​

Step 3: Identify the original source

​

We attempt to locate the earliest available version of the claim.

 

This helps determine whether the material has been:

• shortened;

• mistranslated;

• selectively quoted;

• edited;

• reposted without context;

• falsely attributed;

• modified after publication.

​

When a claim originates from a speech, interview, report, law, study, dataset, or recording, we review the complete source whenever possible.

​

Step 4: Examine the surrounding context

​

We assess the statement within its original context rather than relying only on a screenshot, headline, or short excerpt.

​

This may require reviewing:

• the full conversation;

• the complete speech or interview;

• the full article or report;

• the relevant section of a law or policy;

• the complete dataset;

• earlier and later statements by the claimant;

• the historical background of the issue.

​

Step 5: Review previous reporting and research

​

We search for previous fact-checks, investigations, official corrections, court decisions, academic research, technical analyses, and relevant datasets.

 

Previous reporting may provide useful leads, but we do not treat another article’s verdict as proof. We examine the underlying evidence whenever it is available.

 

Step 6: Contact the claimant

​

Where practical and safe, we contact the person or organization responsible for the claim.

​

We provide:

• the exact statement being examined;

• our understanding of what it means;

• focused questions;

• a request for supporting evidence;

• a reasonable response deadline.

​

The response period may be shorter for breaking news or rapidly spreading misinformation. The urgency and deadline are communicated clearly.

 

Publication may proceed when the claimant does not respond before the deadline. Relevant evidence received after publication will be reviewed under our updates and corrections policy.

​

Step 7: Gather supporting and contradictory evidence

​

Researchers must search for evidence that could support the claim as well as evidence that could contradict it.

 

We do not begin with a preferred verdict and collect only evidence that confirms it.

 

Where possible, important findings are checked against more than one independent source. An exception may apply when a single record is uniquely authoritative, such as an original law, certified

election result, court judgment, audited financial statement, or complete recording.

 

Step 8: Evaluate source quality

 

Each source is assessed according to:

• its direct knowledge of the issue;

• its relevance to the precise claim;

• its independence;

• its expertise;

• its methodology;

• its proximity to the event;

• its date;

• its completeness;

• its consistency with other reliable evidence;

• its potential conflicts of interest;

• whether the underlying evidence can be inspected.

​

The number of sources does not substitute for source quality.

 

Several websites repeating the same unsupported report do not constitute independent confirmation.

​

Step 9: Verify calculations and comparisons

​

When a claim involves numbers, we reproduce the relevant calculation whenever possible.

 

We examine:

• the numerator and denominator;

• the starting and ending periods;

• the chosen baseline;

• percentages compared with percentage-point changes;

• nominal figures compared with inflation-adjusted figures;

• totals compared with per-capita measurements;

• national figures compared with local data;

• preliminary data compared with revised data;

• sample size and margin of error;

• changes in definitions or reporting practices;

• whether unlike categories are being compared;

• whether charts or graphs use misleading scales.

​

The published fact-check should show enough of the calculation for readers to understand how the result was reached.

 

Step 10: Assess the overall evidence

​

We weigh the evidence according to its quality, relevance, directness, consistency, and completeness.

​

We distinguish between:

• evidence supporting the claim;

• evidence contradicting the claim;

• evidence that is uncertain;

• information that is missing;

• reasonable alternative explanations.

​

Step 11: Assign a verdict

​

The proposed verdict must correspond to the weight of the evidence and the published verdict definitions.

​

A verdict is not determined by the researcher’s opinion of the claimant or the political consequences of the conclusion.

 

Step 12: Conduct editorial review

​

Every rated fact-check is reviewed by an editor who did not conduct the primary research.

​

The editor examines the claim, sources, calculations, reasoning, context, wording, and proposed verdict.

​

High-risk investigations may also receive legal, specialist, security, or senior editorial review.

​

Step 13: Publish and monitor

 

After publication, Paraluman Verify monitors credible challenges, new records, corrected datasets, claimant responses, court decisions, and other developments that may affect the conclusion.

Primary evidence

​

Whenever possible, we rely on original or firsthand materials, including:

• legislation and regulations;

• official records;

• court documents;

• raw datasets;

• full recordings;

• original photographs and videos;

• contracts;

• audited financial statements;

• public expenditure records;

• scientific papers;

• election records;

• direct observations.

​

Primary evidence is not automatically reliable simply because it is original. It must still be examined for authenticity, completeness, methodology, interpretation, and possible bias.

 

Independent expert evidence

​

We may consult qualified specialists when a claim involves technical knowledge.

​

Experts are selected according to their relevant expertise rather than their title, public profile, or institutional prestige.

 

We consider:

• whether their training directly relates to the question;

• whether they have conducted relevant research or professional work;

• whether their interpretation is supported by evidence;

• whether they have financial, political, professional, or personal conflicts of interest.

​

Expert opinion does not replace underlying evidence when relevant data are available.

​

Secondary evidence

 

Responsible news reporting, documented investigations, institutional summaries, and reputable databases may provide context or confirmation.

 

When secondary reporting relies on a primary document, we seek the original document whenever possible.

 

User-generated and open-source material

​

Social-media posts, anonymous tips, screenshots, crowd-sourced databases, and unverified files are generally treated as investigative leads rather than confirmed evidence.

 

They may contribute to a conclusion only after appropriate authentication and corroboration.

 

Anonymous sources

​

Anonymous information may be used when:

• the information serves a clear public interest;

• the source is in a position to know the information;

• an editor knows the source’s identity;

• identification could expose the source to a credible risk;

• the information has been corroborated wherever possible.

​

An anonymous allegation will not normally serve as the sole basis for a definitive verdict.

 

We explain why anonymity was granted without publishing information that could expose the source.

​

Missing evidence
 

A claimant’s failure to provide proof does not automatically establish that a claim is false.

A claim may be rated Unsupported when no adequate evidence can be found after a reasonable search.

A claim is rated False only when reliable evidence affirmatively contradicts it or demonstrates that it cannot be correct.

Claims involving health, medicine, science, engineering, statistics, or other technical fields require additional care.

 

We give greater weight to:

 

• systematic reviews;

• well-designed studies;

• peer-reviewed research;

• transparent methodologies;

• sufficiently large and representative samples;

• findings replicated by independent researchers;

• evidence reflecting the current state of knowledge;

• recognized technical standards;

• specialists with directly relevant expertise.

​

We explain significant limitations involving:

​

• sample size;

• study design;

• confounding factors;

• uncertainty;

• preliminary findings;

• preprints;

• lack of replication;

• conflicts of interest;

• differences between laboratory and real-world results.

​

A single study is not presented as settled evidence when the broader research remains uncertain or contradictory.

 

We distinguish between:

​

• absence of evidence;

• evidence showing no effect;

• insufficient evidence;

• an emerging finding;

• a well-established conclusion.

​

We do not create false balance by presenting an unsupported fringe position as equal to a large and consistent body of reliable evidence.

Claims about trends, increases, decreases, or causation receive additional scrutiny.

 

For trend claims, we examine:

• the selected starting and ending dates;

• whether the period is representative;

• seasonal patterns;

• population changes;

• missing observations;

• revisions to data;

• changes in definitions;

• changes in collection methods;

• the appropriateness of the comparison group.

 

For causal claims, we do not assume that one event caused another merely because they occurred at the same time or one followed the other.

 

We consider:

• alternative explanations;

• confounding variables;

• study design;

• comparison groups;

• the proposed mechanism;

• consistency across studies;

• replication;

• the strength of the overall evidence.

​

Correlation alone is not sufficient to prove causation.

Paraluman Verify may use a combination of open-source and forensic techniques to assess digital content.

 

These may include:

• reverse-image searches;

• keyframe searches;

• identification of the earliest available upload;

• metadata examination;

• file-format analysis;

• geolocation;

• comparison of landmarks and terrain;

• weather and environmental analysis;

• shadow and lighting analysis;

• chronolocation;

• frame-by-frame review;

• audio waveform and spectral analysis;

• comparison with original broadcasts;

• typography and document-format checks;

• review of account history;

• provenance records;

• content credentials;

• consultation with forensic specialists.

​

No single automated detection tool is treated as conclusive evidence.

 

Automated systems may produce false positives or false negatives because of compression, editing, file conversion, poor image quality, or changes in detection technology.

 

When the available evidence cannot determine authenticity reliably, we state that the result is inconclusive.

 

Authentic content

​

Content may be described as authentic when the available evidence supports that it is genuine and has been represented accurately.

 

Miscaptioned content

Content is miscaptioned when the image, video, or recording is genuine but the accompanying description materially misstates the date, place, person, event, or circumstances.

 

Altered content

Content is altered when genuine material has been edited, combined, cropped, slowed, accelerated, or otherwise modified in a way that creates a materially misleading impression.

 

Routine brightness correction, color balancing, or cropping is not considered deceptive unless it changes the meaning of the content.

 

Synthetic or AI-generated content

​

Content may be identified as synthetic or AI-generated when sufficient forensic, provenance, contextual, or source evidence supports that conclusion.

 

We do not label content as AI-generated solely because an automated detector assigns it a high probability score.

 

Impostor content

​

Content is considered impostor content when it falsely presents itself as originating from a real person, organization, publication, government agency, or authentic account.

 

Staged content

Content is staged when an event has been arranged or performed but is presented as spontaneous, documentary, or authentic.

 

Inconclusive content

Content is labeled inconclusive when the available evidence is insufficient to determine its authenticity, origin, or method of creation.

Readers should be able to tell what kind of content they are consuming.

​

We clearly distinguish among:

 

• News: Verified factual reporting in which the journalist’s personal opinion is not presented as fact.

• Analysis: Evidence-based interpretation that helps explain the meaning or implications of reported facts.

• Opinion or Commentary: A named author’s argument, judgment, or personal viewpoint.

• Explainer: Contextual material intended to clarify a topic, process, or event.

• Fact Check: A structured assessment of a verifiable claim using the Paraluman Verify methodology.

• First-Person or Eyewitness Account: An account based primarily on the contributor’s personal experience or observation.

• Press Release: Material supplied by an organization and published substantially in its original form.

• Wire, Syndicated, or Partner Report: Material originating from another identified news provider.

• Sponsored or Paid Content: Material created or distributed as part of a commercial arrangement.

• Satire, Fiction, Reconstruction, or Simulation: Non-literal material that could otherwise be mistaken for factual reporting.

​

Opinion columns must be labeled prominently. The views of a columnist or contributor do not necessarily represent Paraluman News.

​

Opinion and analysis are not exempt from factual standards. Claims presented as facts must be supportable, significant interests must be disclosed, and commentary must not rely on fabrication, plagiarism, unlawful defamation, dehumanization, or deliberate misinformation.

True

The claim is supported by the strongest available evidence and contains no material error or omission.

​

Minor wording differences that do not affect the central meaning do not prevent a True verdict.

​

Mostly True

The central claim is accurate, but it contains a minor error, imprecision, exaggeration, or omission that does not substantially change the overall conclusion.

 

Misleading

The claim contains accurate information but creates a materially distorted impression through missing context, selective evidence, inappropriate comparison, framing, exaggeration, implication, or presentation.

 

The label describes the effect of the content. It does not require proof that the claimant intended to deceive.

​

Mostly False

The claim contains a limited element of truth, but its central assertion or overall impression is contradicted by the evidence.

 

False

The central claim is contradicted by clear, reliable, and sufficiently direct evidence.

 

Unsupported

No adequate credible evidence was found to substantiate the claim after a reasonable search, but the available evidence is not strong enough to establish that it is false.

 

Inconclusive

The available evidence is inaccessible, incomplete, genuinely conflicting, or otherwise insufficient to reach a responsible conclusion.

 

Explainer

An article may be labeled an Explainer when it clarifies a complex or developing issue without assigning a verdict to one central proposition.

 

What We Know

This label may be used during breaking or developing events when Paraluman Verify can confirm some information but important facts remain uncertain.

Verdicts are assigned according to the following principles.

​

Materiality

The verdict is based on the central meaning of the claim.

​

A minor mistake does not necessarily make an otherwise accurate claim false. Similarly, a small accurate detail does not make an otherwise false claim true.

 

Context

A statement may be literally accurate while still being misleading when it omits information necessary for a reasonable interpretation.

 

Overall impression

We consider how an ordinary audience would reasonably understand the claim, not only whether each individual word could be defended in isolation.

Intent

We do not need to prove that a claimant intended to deceive before rating content Misleading, Mostly False, or False.

Original publication

A claim is evaluated as it appeared when published or spoken.

 

A later clarification does not erase the original version, although the clarification will be reported fairly.

 

Uncertainty

We use Unsupported or Inconclusive when the evidence does not justify a definitive True or False verdict.

 

New evidence

A verdict may be updated when new authoritative evidence becomes available, a dataset is corrected, a relevant court decision is issued, or an error is discovered.

 

Any verdict change is explained prominently.

Promise tracking is separate from ordinary fact-checking.

​

A promise may have been sincerely made but not delivered. A promise may also be delivered even when the claimant used inaccurate supporting arguments.

 

Which promises qualify

​

A commitment may be included in the Paraluman Verify Promise Tracker when it is:

• attributable to an identifiable official, candidate, administration, agency, or institution;

• made publicly;

• preserved in a reliable record;

• sufficiently specific to assess;

• within the claimant’s authority or presented as something the claimant would pursue;

• relevant to public policy, governance, public spending, rights, services, or community welfare;

• measurable through an action, output, outcome, milestone, or deadline.

 

Broad aspirations such as “improve the economy” do not normally qualify unless they are accompanied by a measurable target or defined action.

 

Promise documentation

​

For each tracked promise, we record:

​

• the original wording;

• who made it;

• the date;

• the setting;

• the original source;

• the policy area;

• the promised action or outcome;

• the people or place affected;

• the relevant baseline;

• the measurement indicator;

• the stated or implied deadline;

• the responsible institution;

• important conditions or dependencies.

​

Later changes or qualifications are documented but do not silently replace the original promise.

 

Promise Recorded

This status means that the commitment has been documented but implementation has not yet been meaningfully assessed.

 

Not Started

This status means that no verified material action toward delivery has occurred.

 

In Progress

This status means that concrete implementation has begun and remains consistent with the commitment, but delivery is incomplete.

 

Partly Delivered

This status means that a substantial part of the commitment has been completed, but one or more important elements remain unmet.

 

Delivered

This status means that the commitment has been fulfilled in substance and at the scale, scope, and standard reasonably established by the original promise.

 

Stalled

This status means that implementation began but meaningful progress has stopped, significantly slowed, or departed from the expected pathway.

 

Broken

This status means that the deadline has passed without material fulfillment, the commitment has been abandoned, or the responsible authority has taken action materially contrary to the promise.

 

Not Yet Assessable

This status means that the deadline has not arrived or the evidence required for a fair assessment is not yet available.

 

Superseded or No Longer Applicable

This status means that the commitment was formally replaced or ceased to apply because of a significant legal, institutional, or factual change.

 

This status is explained and is not treated as successful delivery.

An announcement is not the same as implementation.

​

A proposal is not the same as enactment.

​

A budget request is not necessarily an appropriation.

​

An appropriation is not necessarily an actual disbursement.

​

A construction launch is not the same as completion.

​

A completed facility is not necessarily operational.

​

Promises are assessed according to their wording.

​

A promise to propose or file something may be fulfilled when the proposal is formally submitted.

 

A promise to pass or enact something requires completion of the relevant legislative or legal process.

 

A promise to implement a program requires functioning execution rather than the publication of a plan alone.

 

A promise to fund a program requires the promised level of financing to be made meaningfully available and, where relevant, disbursed.

 

A promise to build or complete a facility generally requires physical completion. When public use was part of the commitment, the facility should also be operational.

 

A promise to create jobs, deliver housing units, or provide services requires verified outputs. We examine possible double-counting, reclassification, and changes in measurement.

 

A promise to increase or reduce a particular outcome is assessed against an appropriate baseline and indicator. Taking action does not constitute delivery when the promise specifically concerned the result.

​

When no deadline was stated, we may use the relevant term of office, official program schedule, fiscal period, or another clearly explained time horizon.

 

External crises, legal barriers, political opposition, or institutional constraints are reported as context. They do not automatically convert non-delivery into delivery.

 

Promise statuses are reviewed when significant milestones occur, when new authoritative evidence appears, and when the relevant deadline arrives.

A standard Paraluman Verify fact-check should clearly present the following elements.

 

The claim

The article identifies the exact or faithfully summarized proposition, who made it, where it appeared, and when it was published or spoken.

 

Why we checked it

The article explains the claim’s public importance, reach, potential harm, or accountability value.

 

The verdict

The verdict appears prominently and is accompanied by a concise explanation.

 

What the evidence shows

The article presents the central evidence and reasoning in clear language.

It includes relevant information that supports the claim as well as evidence that contradicts or limits it.

 

Context

The article explains definitions, chronology, comparisons, limitations, and omitted information needed to understand the issue fairly.

 

The claimant’s response

​

Relevant evidence or explanation from the claimant is included.

​

When no response was received before publication, that may be stated.

​

Sources

The article links to or identifies the most important evidence used.

​

Methods and limitations

The article explains specialized calculations, forensic techniques, assumptions, and significant uncertainties.

 

Publication information

​

The article identifies the author, editor, publication date, updates, corrections, and method for requesting a review.

 

Headlines, captions, graphics, videos, and social-media posts must accurately reflect the article.

 

A nuanced conclusion must not be reduced to a more absolute or sensational statement during distribution.

Every rated fact-check must receive review by at least one editor who did not perform the primary research.

 

• The editor independently checks:

• the wording of the claim;

• the original context;

• the quality of the sources;

• whether contrary evidence was considered;

• calculations and statistical reasoning;

• quotations;

• translations;

• media-authenticity findings;

• potential conflicts of interest;

• whether the verdict matches the published definitions;

• the headline and summary;

• graphics and social-media language;

• whether the claimant was contacted where appropriate;

• whether limitations and uncertainty are clearly explained.

​

The editor should inspect the most important sources directly rather than relying only on the researcher’s summary.

 

Additional review may be required for:

• elections;

• active political campaigns;

• allegations involving identifiable individuals;

• public-health claims;

• legal proceedings;

• national security;

• conflict and war;

• leaked or confidential material;

• synthetic-media findings;

• claims that could create a serious risk of harm.

​

Final editorial authority rests with the designated editor.

​

Commercial, funding, partnership, or technical representatives may not overturn an editorial verdict.

Paraluman Verify distinguishes among minor edits, clarifications, updates, corrections, verdict changes, and retractions.

 

Minor edit

A minor edit corrects spelling, punctuation, formatting, or style without changing the meaning.

 

Errors involving names, dates, figures, quotations, evidence, or attribution are not treated as merely stylistic.

 

Clarification

A clarification is issued when the original information was not materially false but wording or context could reasonably be misunderstood.

 

A dated note explains what was clarified.

​

Update

An update adds new information or subsequent developments without invalidating the original conclusion.

 

A dated note explains what was added.

​

Correction

A correction is issued when a material factual, analytical, sourcing, or presentation error affected the article.

 

A dated correction explains the error and the change.

​

The headline, verdict, graphics, and social-media distribution are corrected where necessary.

 

Verdict change

A verdict change is issued when new evidence or the discovery of an error alters the conclusion.

 

The article clearly states the former verdict, the new verdict, and the reason for the change.

 

Retraction

​

A retraction is issued when the central article cannot be supported or responsibly repaired.

 

Where legally and safely possible, the original page remains accessible with a prominent retraction notice explaining the decision.

 

We do not remove an accurate fact-check merely because the subject disagrees with or dislikes the conclusion.

Readers and subjects may request a correction or review by providing:

• the article title or URL;

• the specific statement they dispute;

• an explanation of why they believe it is inaccurate;

• supporting records, evidence, or links;

• contact information for follow-up.

​

Requests are evaluated according to the strength of the evidence, not the status, influence, or political position of the person submitting them.

 

Where practical, a substantive appeal is reviewed by an editor who was not the primary author of the original fact-check.

 

Material corrections, verdict changes, and retractions are recorded in a public corrections log.

 

When appropriate, a significant correction is also shared through the main channels used to distribute the original article.

Paraluman Verify publishes accessible information about its:

• legal status;

• ownership;

• organizational structure;

• editorial leadership;

• key editorial personnel;

• significant funding sources;

• relevant partnerships;

• contact information.

​

Funding or partnership agreements must preserve editorial independence.

 

A funder or partner may not require Paraluman Verify to:

• check a particular claim;

• assign a particular verdict;

• suppress a finding;

• provide favorable coverage;

• obtain approval before publication.

​

Sponsored or commissioned work is labeled clearly.

​

Editorial staff must disclose relevant:

• political roles;

• campaign work;

• financial interests;

• employment or consulting relationships;

• close personal or family relationships;

• gifts or benefits;

• advocacy roles;

• disputes involving a subject of coverage.

​

When a reasonable reader could question a staff member’s impartiality, the person must be removed from the assignment or the conflict must be disclosed and managed by an editor.

 

Staff members retain their civic rights, including the right to vote. However, editorial personnel should avoid public partisan activity on subjects they may be assigned to fact-check when that activity could reasonably undermine confidence in the newsroom’s independence.

Paraluman Verify may use artificial-intelligence tools to assist with:

• monitoring public information;

• transcription;

• preliminary translation;

• document organization;

• identification of recurring claims;

• extraction of structured data;

• search-term development;

• detection of possible media anomalies;

• routine production support.

​

AI output is not treated as an independent source or as proof.

​

A human journalist must verify every material fact, quotation, translation, calculation, citation, and media conclusion before publication.

 

The final verdict is always made by accountable human editors.

​

Paraluman Verify will not:

• assign a verdict solely through an automated system;

• cite a chatbot response as factual evidence;

• describe content as AI-generated solely because of a detector score;

• upload confidential or sensitive material to systems without appropriate safeguards;

• publish synthetic illustrations in a way that could cause them to be mistaken for evidence.

​

Material use of AI in forensic analysis may be disclosed when it affects how readers should understand the investigation.

 

AI-assisted errors are corrected under the same standards as any other editorial error.

Fact-checking should reduce harm rather than unnecessarily spread harmful or false material.

 

• Before publication, we consider:

• whether repeating the complete false claim is necessary;

• whether including a link, username, hashtag, or screenshot could increase its reach;

• whether personal information should be redacted;

• risks to minors, victims, witnesses, and private individuals;

• the presence of graphic or traumatic material;

• the risk of harassment or retaliation;

• the security of confidential sources;

• whether the headline could spread the false claim without its correction.

​

Where appropriate, we lead with the verified information before describing the false claim.

Images and graphics must make the verdict clear and must not present deceptive content in a way that appears authentic.

Public figures are subject to greater scrutiny regarding their public actions and statements.

Private individuals are identified only when identification is necessary, proportionate, and supported by a clear public-interest justification.

Readers may submit claims for possible verification through Paraluman Verify’s designated submission channel.

A useful submission includes:

• the exact claim;

• a link, screenshot, recording, or document;

• who made or shared it;

• when and where it appeared;

• why it may be important;

• any available supporting or contradicting evidence.
   

Submitting a claim does not guarantee that it will be investigated or published.

All submitted claims are evaluated using the same editorial selection standards, regardless of who sends them.

Personal information collected through submissions is used only for legitimate editorial, safety, verification, and follow-up purposes.

Paraluman Verify reviews this methodology at least once a year and whenever significant changes occur in:

• editorial practices;

• verification technology;

• evidence standards;

• promise-tracking procedures;

• corrections practices;

• organizational structure;

• professional fact-checking standards.

​

Each published revision should include:

• a version number;

• an effective date;

• a summary of material changes;

• access to earlier versions where practical.

​

The newsroom may conduct periodic internal quality reviews examining:

• whether verdict definitions are applied consistently;

• the quality and diversity of sources;

• the completeness of claimant-response efforts;

• correction and appeal patterns;

• political and subject-matter distribution;

• the treatment of uncertainty;

• consistency in promise-tracking decisions;

• responsible use of AI;

• whether headlines and social posts accurately reflect the underlying articles.

 

The findings of these reviews should inform staff training, editorial improvements, and future revisions of this methodology.

Paraluman Verify follows a consistent, evidence-based fact-checking process. We select claims according to their public importance, reach, potential harm, accountability value, and verifiability—not according to the identity or political position of the claimant.
 

We seek primary evidence, examine supporting and contradictory information, contact claimants where practical, subject rated fact-checks to independent editorial review, disclose important sources and limitations, and correct material errors openly.
 

Our verdicts describe what the strongest available evidence supports at the time of publication.