Methodology

    We believe readers deserve to know how the content they're reading was produced. Here is the process every Dorsi article goes through.

    1. Topic selection

    Topics are selected from our keyword research pipeline — a mix of search-volume data, real questions asked in fitness and wearable-tech communities, and gaps we notice in existing coverage. We don't write about a topic unless we have access to credible primary sources for it.

    2. Source gathering

    For each article we pull from primary sources: peer-reviewed studies (PubMed, sports-medicine journals), official guidance (ACSM, AHA, NIH), and manufacturer documentation (Apple, Garmin, Polar, Whoop) for product-specific claims. For real-world experience, we pull verbatim discussion from relevant subreddits and forums, with full attribution. We do not paraphrase community discussion — we quote it.

    3. Drafting

    We use AI assistance (large language models) to draft articles from the primary sources gathered in step 2. The AI's job is to synthesize and translate the sources into accessible language — not to invent claims. Sentences that cannot be tied back to a primary source are dropped during the next step.

    4. Anti-AI quality gateway

    Each draft passes through an internal quality gateway (powered by a separate AI model) that scores it for the "AI tells" Google's scaled-content-abuse guidance flags: empty filler sentences, generic transitions, vague hedges. Drafts that score above our threshold are rewritten or blocked entirely.

    5. Human review

    Every article is reviewed by a domain editor before publication. The editor reviewing a given piece is named in the byline under the article title and is accountable for the final accuracy and framing. See our reviewer team for who reviews what.

    6. Refresh & updates

    Articles are revisited on a rolling cadence: source citations re-verified periodically, comparison tables refreshed when referenced products release significant updates, community quotes preserved as-is (they're snapshots of real discussion at a specific point in time and are not rewritten in place).

    What we don't do

    • · We don't publish articles without a human reviewer named in the byline.
    • · We don't claim medical, clinical, or institutional affiliations our editors don't actually hold.
    • · We don't paraphrase community discussion or invent attributed quotes. Real quotes only, with permalinks.
    • · We don't auto-publish content with quality-gateway scores above our internal threshold.