Am I cited?
How is AI citation likelihood calculated?
Citation likelihood is Seeklyn's forecast of how likely a page is to be cited by AI answer engines, a score between 0 and 1 expressed in 3 bands: likely (≥0.65), possible (≥0.35), and unlikely (<0.35). The score is a weighted mean of 4 signals: retrieval rank (45%), fan-out coverage (25%), citability (20%), and competition density (10%), renormalized over whichever signals are available so a missing signal never drags the score.
Citation likelihood bands appear in the "Am I Cited?" panel in every Seeklyn content report — run a check free at /dashboard/cited.
How is a likelihood different from an observed citation?
An observed citation is a fact: Seeklyn read the actual AI answer from ChatGPT, Google AI Overviews, Perplexity or another engine and found your page among its cited sources. A likelihood is a forecast: it estimates how likely a citation is from the signals that drive one, useful before your first check or for a draft that is not live yet.
The forecast is labeled `inferred` throughout the API and UI — kind is always "inferred", never "observed" — so it is never mistaken for a direct measurement. When the two disagree, the observed answer wins.
What are the 4 signals and their weights?
The likelihood formula uses 4 signals with locked weights, summing to 1.0 when all are present. Retrieval rank carries the most weight (45%) because whether an engine's index surfaces your page at all is the largest determinant of citation. Fan-out coverage (25%) captures whether your content covers the ~10.7 sub-queries a prompt typically fans out into. Citability (20%) reflects the quality of your best passage — how well it answers a question on its own. Competition density (10%) measures how many competing pages rank above yours, which reduces your relative citation probability.
| Signal | Weight | What it measures | Source |
|---|---|---|---|
| Retrieval rank | 45% | How highly your page ranks in web search results for the query | Web search results for the query |
| Fan-out coverage | 25% | What fraction of the query's sub-angles your content covers | Content analyzer fan-out score (0–1) |
| Citability | 20% | Quality of your best passage — answer-first, self-contained, fact-dense | Content analyzer best-chunk citability score (0–1) |
| Competition density | 10% | How many competing pages rank above yours (crowded SERPs lower likelihood) | Competing pages in the same results |
How are the 3 likelihood bands defined?
3 bands divide the 0–1 score range with thresholds locked in the codebase: likely is ≥0.65, possible is ≥0.35, and unlikely is below 0.35. The bands are labeled consistently in the API (`band: "likely" | "possible" | "unlikely"`) and in the Seeklyn UI, so the language you see always matches the underlying number. A score of 0.64 is "possible", not "likely" — the threshold is a hard cut, not fuzzy.
| Band | Score range | Interpretation |
|---|---|---|
| likely | ≥ 0.65 | Strong retrieval and content signals — citation plausible |
| possible | ≥ 0.35 and < 0.65 | Mixed signals — some factors support citation, others don't |
| unlikely | < 0.35 | Weak retrieval or content signals — citation improbable |
What happens when a signal is missing?
Missing signals renormalize out of the formula rather than being counted as zero. If fan-out coverage or citability are not available (for example, because no content was analyzed for the target URL), those weights are dropped from both the numerator and the denominator and the remaining signals are reweighted proportionally. A check run with only retrieval rank and competition available still produces a meaningful likelihood — it is just less precise than one with all 4 signals.
The `basis` field in Seeklyn's API response lists exactly which signals were used in each prediction, so the confidence behind any given likelihood is always legible. A prediction with basis `["retrieval", "competition"]` is weaker than one with `["retrieval", "competition", "fanout", "citability"]` — and the UI reflects that honestly.
Frequently asked questions
Does a "likely" band guarantee a citation?
No — a "likely" band means the 4 retrieval and content signals are strong, not that a citation is certain. AI engines control their own ranking; Seeklyn can measure the signals that correlate with being cited, but cannot guarantee any placement. The band is always labeled inferred for this reason.
Why is retrieval rank weighted 45% — more than citability?
Because a passage can be perfectly quotable yet never appear in the engine's sources if retrieval rank is too low. The retrieval stage is a hard gate: a page that is not retrieved cannot be cited, regardless of content quality. Giving retrieval the highest weight reflects that order-of-operations reality.
Can I improve my likelihood score?
Yes. The 4 signals are all content-addressable: improving BM25 retrieval rank (better keyword coverage), expanding fan-out coverage (answering more sub-queries), raising citability (more self-contained, fact-dense passages), and earning more citations than competitors (lowering competition density above you) will each raise the likelihood. Seeklyn's content analyzer measures all 4.
Should I trust a likelihood or an observed citation?
The observed citation. Seeklyn tracks the actual AI answers and citations across ChatGPT, Google AI Overviews, Perplexity and more; a likelihood is a forecast that helps you prioritize fixes before or between checks. It is always labeled inferred because that is what the data supports.