Editorial methodology

We select sources based on expertise in the subject: ANTT for transport regulation, tax administrations for documents, ANP for fuels, and IBGE and Central Bank for economic series. An official source can be a navigation index; this alone does not justify a news item.

The collection records origin and evidence. The extraction identifies claims and verifiable excerpts. The analysis must explain the fact, the affected public, the operational context, and practical guidance supported by the document. Consultation, publication, and validity dates are different and should not be interchanged.

AI assists classification, research, writing, and checking. Model verification is not human review. Internal records maintain steps and traceability; technical evidence has limited retention. We do not invent interviews, authors, experts, reviewers, or dates to fill the editorial model.

On 09/22/2026, we reviewed the collection of 43 publications: 16 contents were deepened and 27 related summaries were incorporated into the corresponding pages, with URL redirection and history preservation. The edition was supported by AI and source verification, with no record of individual human review. The auditor signals absence of sources, context, links, and very short texts; word count does not prove quality nor is it a Google rule.

New drafts need to present context and practical action. Factual regulatory reports can be published automatically with evaluation of the exact version, excerpts proven in primary source, and absence of individual advice. Individual legal guidance remains withheld. Before/after comparisons and economic impacts should only appear with foundation. When it is not possible to conclude, we specify what remains to be confirmed.

The planning considers aggregated themes from the assistant, utility evaluations, and tool usage. These signals guide research; they do not prove a fact nor dispense with sources. An isolated vote does not alter a publication. The production follows frequency limits, seeks evidence, and checks existing coverage to avoid repetition.

Publication, update, and correction preserve versions and reasons in the administration. We do not change dates to simulate novelty. The radar reproduces the registered documentary status and does not automatically classify the level of impact. The observatory explains the assumptions of each series and their calculations.

The specialized press is presented as external reading attributed. To correct an error, use Contact with the URL, the excerpt, and a source that allows verifying the discrepancy. We evaluate the evidence and the need for review, correction, or removal.