$ cite-engine docs methodology

how the cite score works.

the cite score is a proprietary 0-100 weighted composite. five inputs. one number. anchored to the lighthouse-score pattern — easy to talk about, hard to game without actually improving.

the formula

cite_score = (
  0.40 * citation_share_of_voice +
  0.20 * bot_crawl_health        +
  0.20 * answer_readiness        +
  0.10 * content_provenance      +
  0.10 * agent_commerce_readiness
) clamped to [0, 100]

the v0.2 and v0.3 buckets contribute 0 in v0.1. the effective denominator is 80 — we normalize back to 100 so the user-facing number always reads 0-100. as v0.2 and v0.3 ship, their weights re-engage.

the weights

  • 40%

    citation share-of-voice

    how often the site is cited across chatgpt, claude, and perplexity for its top 50 keywords. measured by sampling each engine via the api gateway and parsing citation links.

  • 20%

    bot-crawl health

    frequency, recency, and coverage of ai bots (gptbot, claudebot, perplexitybot, applebot-ai) relative to googlebot. logged via the cite engine worker.

  • 20%

    answer-readiness

    llms.txt present + non-trivial, ai-citation schema emitted, structured data quality (json-ld parses cleanly, articles have author + datepublished, products have offers + price).

  • 10%

    content provenance (v0.2)

    c2pa content credentials signed. zero in v0.1 — reserved for the q3 release.

  • 10%

    agent-commerce readiness (v0.3)

    product schema + acp compatibility + stripe shared payment token wiring. zero in v0.1.

the data sources

ai gateway citation sampling
chatgpt, claude, perplexity queried via the official api gateways. citation links parsed from responses. samples are rate-limited and rotated; engine providers are NOT scraped.
bot-visit logs
cirv sight customers run the cirv sight worker on cloudflare. logs are anonymized, aggregated to the bot-type + path level, never per-user.
schema audit (live)
the cirv sight schema audit tool re-fetches the home + 5 sampled internal pages every 24 hours, parses json-ld and microdata, and emits the answer-readiness score.

update cadence

citation samples refresh every 15 minutes. bot-visit logs ingest in near-realtime (p95 under 60 seconds). schema-audit re-fetches run every 24 hours per ranked site. the leaderboard you see is the most recent 15-minute snapshot.

how to claim your rank

  1. find your domain on the leaderboard or via search.
  2. click claim → on your row.
  3. verify ownership — either via a one-line dns txt record, an html meta tag, or by uploading a 32-character key file to your site root. takes about 30 seconds.
  4. once claimed, you get the embeddable cite badge, daily rank deltas across all three engines, and a private dashboard.

why this matters

ai search is taking citations that organic search used to convert into clicks. semrush, ahrefs, and yoast measure the old game. the cite score measures the new one. owning a top-100 rank today is the easiest way to get on the rolling list of sites journalists, agencies, and brands cite as "ai-cite-worthy." the moat is the rank itself.