The Citation Score: a per-URL scoring methodology

Citation Score is a 0–100 score Lantern assigns to every individual URL that gets cited by an AI provider in a tracked answer. It is a per-URL number, not a brand-level rollup — and that granularity is the point.

The four inputs

• Provider count — how many of the five tracked consumer providers (ChatGPT, Claude, Perplexity, Gemini, Amazon Rufus) have cited the URL in the last 30 days.
• Query count — how many distinct tracked prompts the URL has been cited in.
• Position in answer — where in the model's answer the URL appears. Citations in the first third of an answer contribute roughly 1.8× the weight of the same URL cited in the bottom third.
• Recency — how recently the URL was cited. URLs cited inside the trailing 7 days carry full weight; older citations decay with a 30-day half-life.

The formula

formulaCitation_Score(URL) = round(
    25 * provider_weight        // 0 .. 5 providers, scaled to 0..25
  + 25 * query_weight           // 0 .. tracked-prompt-set max, scaled to 0..25
  + 25 * position_weight        // 1.8x for top-third, 1.0x mid, 0.55x bottom
  + 25 * recency_weight         // 1.0 in last 7d, half-life 30d thereafter
)

provider_weight   = providers_citing_URL / 5
query_weight      = log(1 + queries_URL_appears_in)
                  / log(1 + max_queries_in_set)
position_weight   = sum_over_citations(position_multiplier) /
                    sum_over_citations(1.0)
recency_weight    = sum_over_citations(0.5 ^ (age_days / 30)) /
                    n_citations_30d

The Own / Earned / Competitor / Social classifier

Citation Score is computed independently of the citation-type classifier — but the two run together in the Citations surface because the same score means different things on different source types. Own is the brand's primary registered domain. Earned is a third-party site that is neither competitor nor social. Competitor is a tracked competitor's domain. Social is a maintained list of UGC and community-platform domains.