Editorial Process

Our articles are produced by an AI-assisted editorial pipeline. We say that out loud because readers deserve to know, and because pretending otherwise would be the one thing that actually undermines trust.

AI is a drafting and research tool, not the editor. Every piece of content goes through the same five stages, in order:

1. Ideation.A topic is selected based on the sports calendar (upcoming games, rivalries, seasonal milestones), real reader search intent from our own data, and gaps in our city/team coverage. No clickbait topic generation, no "whatever is trending."
2. Research. A research agent pulls structured facts from our internal data stores and a curated list of external sources (see the Sources section below). The research brief is saved and attached to the article for provenance.
3. Drafting. An AI writer assembles a draft from the research brief under the voice of a named persona. The draft is grounded in the research — not freeform generation — which is how we minimize hallucinations.
4. Humanization and quality scoring. The draft passes through multiple editing stages that smooth voice, remove AI tells, and verify internal consistency. An automated quality score is computed. Drafts below a minimum bar are rejected and rewritten.
5. Publication decision. Every draft that clears the quality bar passes through AI detection checks and automated scoring. Articles above our threshold publish automatically; those that fall short are flagged for additional review or rejected entirely.

We keep a full provenance trail for every article — the original research brief, the first draft, every humanization pass, the quality score, the detection score, and the final editorial decision. That audit log exists so we can improve the pipeline, catch recurring failure modes, and answer the question "where did this sentence come from?" when it matters.

An article is only as honest as its sources. Here's what grounds our content:

• Our own venue directory.Everything we say about a bar — address, hours, TV count, team affiliation, vibe — comes from the verified records in our database. See the About page for how those records are built from Overture Maps and public business data.
• Game schedules and scoreboard data.We pull from official league schedules via ESPN's public endpoints. Dates, matchups, venues, and broadcast networks come from that source and update nightly.
• Live odds. Betting odds are provided by The Odds API, cached on our own infrastructure and refreshed multiple times per day. Odds are labeled with a timestamp.
• Team records and historical matchups. Season-to-date records, head-to-head history, and streaks come from our internal intelligence store, which is built from the same league feeds above.
• External league and team news. A curated list of ~2,000 team-level and national sports sources is monitored nightly. Facts pulled from those sources are attributed internally and cross-checked before they land in a draft.

What we do notdo: invent sources, attribute quotes to wire services that didn't say them, or paraphrase scraped content and pass it off as reporting. If we don't have a fact, we leave it out.

Articles on Game Day Bars are signed by named personas — for example, Jay Backfield, The Chalk, Maplewood. These are editorial voices, not real individuals, and we want to be explicit about that because clarity beats cleverness.

Each persona has a consistent beat (analysis, betting, local color, etc.) and a consistent writing voice. Treating them as voices rather than ghost authors lets us be transparent about how content is produced without losing the craft of having a point of view. The persona is the tone, not the author of record — every article is produced by our editorial pipeline, and we own the output.

The author of record for everything on this site is Game Day Bars (operated by Secret Lab, LLC). Questions about attribution or voice? Email [email protected].

Mistakes happen. Here's how we handle them:

• Report it. Email [email protected] with the article URL and what's wrong. A person reads every one.
• We verify.We re-check the claim against the same sources the article used — or better ones if the correction itself points us somewhere new.
• We correct.Factual errors are fixed promptly. Meaningful corrections are noted on the article so readers know what changed and when. We don't ghost-edit articles to hide the original error.
• We learn.If a correction reveals a systematic issue — a bad source, a research gap, a voice problem — the pipeline gets updated so that class of mistake doesn't repeat.

• Publish an article we wouldn't stand behind.
• Take money to change a ranking, a review, or a factual claim.
• Invent sources or quotes. If you see a name attributed in a GDB article, that name exists.
• Pretend a persona is a real person. Personas are voices, and we say so on this page.
• Push sportsbook content in states where online sports betting is not legal. Our CTAs are geo-fenced for that reason.