Content analyzer
What makes a passage citable (self-contained)?
A citable passage answers a question in its first sentence, stands alone without prior context, runs 100–300 words, and carries at least 1 concrete fact with a named source. Seeklyn scores each passage 0–100 on 4 criteria — answer-first, self-contained, in-band length, fact-present — and flags the 2 structural issues that break extractability: ANAPHORA_OPEN (pronoun opener) and NO_ANSWER_FIRST (first sentence lacks a direct declarative answer).
Each passage in a Seeklyn content report shows its citability score (0–100) and a self-contained badge — run your page free at /dashboard/content.
What is passage citability?
Passage citability is a 0–100 score Seeklyn assigns each 100–300-word passage, measuring how likely an AI answer engine is to lift that passage verbatim into a synthesized answer. AI engines like Google AI Overviews, Perplexity, and ChatGPT retrieve individual passages from your page — not the whole page — and the passage either passes 4 self-containment criteria or it does not.
The 4 criteria are: (1) the first sentence directly answers the question the section is about, (2) the passage stands alone — no pronouns like "this" or "it" that demand prior context, (3) word count falls in the 100–300-word retrieval-sized band, and (4) at least 1 concrete fact (a number, date, or unit) appears in the passage, ideally with an attributed source. A passage that clears all 4 can be extracted cleanly; one that fails any criterion breaks when an engine tries to quote it out of context.
How does Seeklyn determine whether a passage is self-contained?
Seeklyn marks a passage self-contained when none of its sentences fire either of the 2 structural flags: ANAPHORA_OPEN (a sentence opens with a pronoun — "This shows…", "It means…", "They found…") or NO_ANSWER_FIRST (the first sentence is not a direct declarative answer). Only these 2 flags break standalone extractability. The remaining 6 flags (HEDGE, RUN_ON, LOW_FACT_DENSITY, NO_ATTRIBUTION, ENTITY_SPARSE, KEYWORD_STUFFING) reduce the passage score without voiding self-containment status.
Seeklyn's `isSelfContained` function (in `analyzecontent.go`) returns false if ANAPHORA_OPEN or NO_ANSWER_FIRST is present in any sentence. Every other flag is scored against the 0–100 citability scale without flipping the self-contained badge. A passage can score 45/100 yet remain self-contained — it has weaknesses like hedging or missing attribution, but can still be extracted cleanly.
What does the 0–100 citability score measure?
The citability score is the passage-level number that feeds Seeklyn's Citation stage of the 3-stage citation funnel (Retrieval → Citation → Overall). A score near 100 means the passage answers first, names entities, carries facts, cites them, and sits in the ideal 134–167-word band. A low score usually means 1 or more of the 8 extractability flags fired: the 2 structural flags (ANAPHORA_OPEN, NO_ANSWER_FIRST) have the heaviest weight, and the 4 moderate flags (HEDGE, RUN_ON, LOW_FACT_DENSITY, NO_ATTRIBUTION, ENTITY_SPARSE) each reduce it further.
The score is monotonic under edits: applying a fix that clears a flag can only raise or hold the citability score, never lower it. Seeklyn enforces this invariant so you can iterate with confidence — each edit either helps or does nothing, and the number is comparable across runs for the same content.
Why does answer-first structure matter so much?
Answer-first structure matters because AI engines retrieve passages in milliseconds and then generate an answer — the engine's ranking model strongly favors passages whose first sentence already contains the answer. A passage that buries the key fact in sentence 3 after 2 sentences of setup is often passed over: the first sentence is what the engine reads to decide relevance, and a vague opener fails that test.
Princeton's GEO research (KDD 2024) identified answer-first structure as one of the 6 core AEO levers that correlate with being cited. Seeklyn operationalizes it with the NO_ANSWER_FIRST flag (evidence: strong): the first sentence must be a declarative statement, must not open with a hedge or pronoun, and must contain at least 1 fact token (digit, percent, or unit). A sentence that says "Content marketing is important" passes the structure test but not the fact-token test, so it still fires NO_ANSWER_FIRST.
Frequently asked questions
Can a passage score 100/100 on citability?
Yes. A passage scores 100 when it fires none of the 8 extractability flags, sits in the 134–167-word ideal band, answers first with a concrete fact, names entities, and attributes its claims. It is achievable — most informational how-to or definition passages do it naturally.
Is self-contained the same as high citability?
No — self-contained is a binary badge (true/false) based on just 2 structural flags, while citability is a 0–100 score across all 8 flags. A passage can be self-contained (no pronoun opener, answer-first) and still score 40/100 if it has no facts, no attribution, and hedged language. Self-contained is a necessary condition, not a sufficient one.
What is the ideal passage length for citability?
The ideal band is 134–167 words — the sweet spot within the retriever-sized 100–300-word range. Seeklyn flags passages as in the ideal band at that length. Shorter passages may lack enough context; longer ones risk being too dense or getting split mid-thought during chunking.
Does citability affect the Retrieval score?
No — citability feeds the Citation stage of the funnel, not Retrieval. Retrieval is determined by BM25 query-term overlap: whether the passage contains your target query terms. A passage can score 90/100 on citability but still be orphaned (zero BM25 score) if it shares no words with the query.