How to Tell Which News Facts Are Confirmed—and What Remains Unresolved

To separate confirmed facts from unresolved claims, identify exactly what is being asserted, trace it to the original source, and check what that source actually establishes. Compare independent evidence, dates and corrections. A credible report can confirm that an allegation was made without proving the allegation itself.

Six useful evidence labels

Confirmed: the specific fact is established by evidence appropriate to that fact. An official record may establish that an agency issued an order; it does not automatically establish every claim in the order.

Reported: a named outlet or source says something happened. Assess its access to the evidence and whether the underlying material is available. Attribution tells you who is making the claim, not that it is independently verified.

Disputed: identified parties or evidence conflict. Look for what each side actually contests and whether their accounts refer to the same event, time and measure.

Inference: an interpretation drawn from established facts. A sequence of events may support a possible explanation without proving cause.

Forecast: a projection about the future. Identify who made it, when, its assumptions and the measure predicted; it is not an observed result or an official decision.

Unresolved: evidence does not yet settle a material question. An absence of evidence is not proof that an event, loss or cause did not exist.

Trace the claim, not just the headline

Find the original document, interview, dataset or first-hand reporting. Check whether later articles supply independent evidence or repeat the same source. Five copies of one report are still one evidentiary chain. A primary source can establish its own decision or statement, but its account may need independent scrutiny.

Compare the event date with the publication and update dates. Check whether figures describe the same population, unit, rate and period. Read the qualifications and look for a later correction. The Associated Press explains its attribution and corrections practices; those are AP's standards, not a claim about Espresso's review process.

Example: the OpenAI mathematics manuscripts

On October 6, 2026, an OpenAI community announcement said its public repository contained 722 manuscripts in 372 result families. The announcement supports the claim that OpenAI reported releasing that collection at that time. It does not by itself establish that every mathematical result is correct or independently verified.

Espresso's coverage of the verification gap illustrates the question to keep open: which individual results have been checked, by whom, and with what limits? Check the underlying repository and subsequent expert assessments before treating a specific proof as settled.

When the story changes

Record the claim and its source, then look for a dated update, correction, new primary record or independent finding. A later report can narrow, contradict or supersede an earlier account. If no reliable update resolves the question, keep the status unresolved rather than filling the gap.

Espresso uses AI to help find connections and draft content. Its editorial standards, AI-use and corrections page describes how people spot-check coverage, investigate error reports and correct substantive errors. Not every article receives human review before publication.

Methodology and update history

Published November 3, 2026. This guide draws on AP's public sourcing and correction guidance and the linked OpenAI announcement. The example describes the announcement's stated scope, not an independent validation of its mathematical results. Updated November 3, 2026.