Editorial standards and AI use
This page explains, in plain terms, how The Stovall Report collects, scores, ranks and explains the news, where AI is used, and how errors are corrected. It describes the system as it actually runs.
Last reviewed
How stories are collected
Stories come from publishers' own RSS feeds — several hundred active feeds covering national and international news, policy, economics, science and technology, religion, and Texas regional and local outlets. An automated pipeline reads the feeds every night, with an additional midday refresh.
Each story's address is normalized and de-duplicated so the same article is not counted twice. Text is read only so the scoring models can evaluate it; it is never published here.
How stories are scored and ranked
Similar coverage of the same event is first grouped together. A smaller AI model screens one representative story from each group for relevance; groups that clear a relevance threshold go to a larger model, which applies the full editorial rubric. Stories that do not clear the threshold keep their relevance score and a one-line note, and are rarely ranked.
The rubric produces two scores from 1 to 100. Importance measures consequence: from nation-altering events at the top, through major policy and geopolitical developments and significant but routine news, down to marginal items. Sports, entertainment, celebrity and lifestyle stories fall outside the mandate and are scored low. Credibility measures the reporting itself: how specific the evidence is, named versus anonymous sourcing, transparency about data, internal logic, and sensational tone. A story cannot score above 95 for credibility without primary sourcing. Credibility rates the article in front of the model, not the outlet's reputation.
The composite score that orders the front page is 0.6 × importance plus 0.2 × credibility, plus up to 20 points for reader engagement on this site (aggregated click counts from the last 48 hours, scaled against the most-read story). After 48 hours a story's composite declines by 3% per hour, down to a floor of 10% of its value, so the page stays current.
The Morning Edition is the ten highest-composite stories from the previous 72 hours, subject to diversity rules: no single topic may take more than 20% of the slots, no single publisher more than two, and at least 15% of the slots go to technology and science. The edition is then frozen: its order, scores and reasoning are stored with the date, so later re-scoring of a story does not alter a published edition.
What the “board” is
“The board” and “board reasoning” refer to a written editorial rubric that the AI models follow, not to a panel of people who meet and vote. The rubric asks the model to weigh each story through several lenses — truth and consequences, virtue and moral order, liberty and constitutional order, civilizational stakes, strategic clarity, economic reality, and whether a story is being overlooked — and to reason in the manner of a roster of historical thinkers and statesmen, with St. Thomas Aquinas as the lead perspective and figures such as St. Augustine, Pope Leo XIII, George Washington, Alexander Hamilton, C.S. Lewis and Fulton Sheen among the others.
These names describe an intellectual tradition the rubric draws on. None of them, and no living person, reviews or endorses individual scores. The rubric instructs the model to evaluate only the article provided, to say “Not established in the reporting” where evidence is missing, and to state uncertainty plainly.
AI use disclosure
AI models draft the following: importance and credibility scores; board reasoning; historical context; what-to-watch notes; topic tags, related threads and story flags; the Morning Edition narration script; the synthetic voice that reads it; the Saint of the Day profile; the TX-23 resident briefs; and the extraction of the national screwworm detection total from published sources.
The models run on the publisher's own hardware. The publisher sets the rubric and the rules, oversees the pipeline, reviews the output, and corrects errors. AI drafts are not individually line-edited before publication, which is why every score is shown with its reasoning and why the corrections log exists.
Headlines, publisher names, author bylines and publication times come from the publishers' feeds and are shown as published.
Editorial lens
The Stovall Report judges importance from a classical-conservative and Catholic point of view: Thomist natural law, the Catholic social teaching that begins with Pope Leo XIII's Rerum Novarum (subsidiarity, the dignity of labor, the common good), ordered liberty under the Constitution, and fiscal discipline. The lens shapes which stories are ranked as consequential and how they are explained. It is stated openly so readers can weigh it.
The lens does not license changing facts. Credibility is scored on the evidence in the reporting, whether or not its conclusions agree with the lens, and stories that run against institutional consensus are marked as contrarian rather than being promoted or hidden for that reason.
Sourcing and attribution
Every ranked story is credited to its publisher and links to the original article. The Stovall Report shows only a story's headline, publisher, byline and time, alongside its own scores and analysis; it does not republish the publishers' article text. Read the original reporting for the full facts.
Inclusion in the rankings is not an endorsement of a publisher, and a low score is a judgment about one article's consequence or evidence, not about its outlet.
Screwworm figures
The Screwworm Watch combines a hand-verified dossier of official sources with live state data. The national detection total is extracted automatically each night from recent published coverage and official pages; a figure is recorded only when a source states the number explicitly, and it is shown with its source and a confidence rating. If no figure is found for 14 days, the page falls back to the last hand-verified count.
Corrections policy
When a factual error is found in our own text — a score explanation, context note, narration, brief or figure — it is corrected in place, and a dated note describing the correction is added to the public corrections log.
Contact
Report an error or ask about sourcing at [email protected]. Please include the page address and, for a Morning Edition, the date and story number.