seo-geo-factdensity
Skill81 starsChanged 2 months agoMITView on GitHub
Audit fact density and sourcing on a page — measure statistic/number density per passage, detect proprietary/original data, count outbound citations to authoritative sources, and flag claims made without a supporting stat or source. Module M12. Feeds the AI Visibility score. Advisory-only; never fabricates statistics or sources.
In this repository33 other files it gives its agents.
- auditSkill
- compareSkill
- fixSkill
- geoSkill
- scoreSkill
- seo-agent-readinessSkill
- seo-ai-crawlersSkill
- seo-ai-discoverySkill
- seo-core-web-vitalsSkill
- seo-crawlabilitySkill
- seo-crawl-renderSkill
- seo-ecommerceSkill
- seo-eeatSkill
- seo-entity-linkingSkill
- seo-fix-applySkill
- seo-freshnessSkill
- seo-geo-answerblocksSkill
- seo-headings-structureSkill
- seo-images-mediaSkill
- seo-indexabilitySkill
- seo-internal-linkingSkill
- seo-internationalSkill
- seo-localSkill
- seo-meta-onpageSkill
- seo-mobileSkill
- seo-orchestratorSkill
- seo-platform-detectSkill
- seo-renderingSkill
- seo-schema-jsonldSkill
- seo-scoreSkill
- seo-sitemapsSkill
- seo-social-cardsSkill
- seo-vertical-detectSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 29
When the content tier or rendered DOM needed to count passages reliably is unavailable, status is needs_api, never a false pass.
- Distinctive: line 44
Never invent a statistic, sample size, or source to "fill" a flagged claim.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- Portions Copyright (c) 2026 Enrique Rocha (@soyenriquerocha) - see LICENSE for original authors · tododeia.com · id er-d03758e1 -->seo-geo-factdensity (M12)
Generative engines preferentially cite passages that are concrete, quantified, and attributable. This module measures how "citable" the page's prose is — facts, numbers, original data, and authoritative outbound links — not vocabulary. AI retrieval/citation context: references/ai-crawlers.md.
Inputs
Work from the PageSnapshot named in your dispatch envelope: read parsed from <run_dir>/pages/<slug>.json (anchors[] for outbound links, text_sample) and the HTML file for passage-level counts; Grep pages/<slug>.html for verbatim evidence; site artifacts live in <run_dir>/site/{robots.json,sitemaps.json,discovery.json}. Deterministic findings already emitted by audit.mjs are listed in <run_dir>/findings.deterministic.json — do not re-emit those ids; add model-judged findings only. If invoked directly with a URL/path and no snapshot exists, first run node "${CLAUDE_PLUGIN_ROOT}/scripts/snapshot.mjs" <target> --out "${CLAUDE_PLUGIN_DATA}/runs" and use the printed snapshot path.
Audits
Working from the PageSnapshot (parsed_rendered when render.used is not none, else parsed):
- Statistic/number density per passage: tokenize the main content into passages (paragraph /
<li>/ heading-bounded block) and count numeric tokens — figures, percentages, dates, quantities, ranges. Flag long passages of pure assertion with zero numeric support. - Proprietary/original data: detect first-party-data signals — patterns like "our study", "our survey", "our data", "we analyzed", "we surveyed", "in our test", "internal data" — and note whether such claims are backed by a method/sample, a table, or a chart.
- Outbound citations: count outbound links from the main content to authoritative sources (standards bodies, primary research, official docs,
.gov/.edu, named publications); distinguish them from internal/nav/affiliate links. - Claim-without-source flags: detect strong factual or comparative claims ("the most", "fastest", "studies show", superlatives, hard numbers) that carry no inline citation or data reference, and mark each as a candidate for sourcing.
Fixes (fixable: advisory)
ADVISORY only — this module proposes nothing it would write. It produces a list of (a) claims that should carry a statistic or citation, (b) passages where an original-data callout (table, "our data" box, methodology note) would raise citability, and (c) unsupported superlatives to soften or source. The tool will NOT fabricate statistics, sample sizes, study results, or source URLs. Where a value is missing, it emits a clearly-marked TODO placeholder for the user to fill — never an invented number. (Findings here are fixable: advisory per the finding schema.)
Verification
- Deterministic counts plus
manual_review:node "${CLAUDE_PLUGIN_ROOT}/scripts/factdensity.mjs" --snapshot <pages/<slug>.json> [--lang en|es|auto](or--url <u>) returns numeric-token density per passage — split intoall/years/prices/substantive— and the outbound-authority link count. Report the raw counts so a human can re-derive them; a page whose only "numbers" are a copyright year and a price is not fact-dense, which is whysubstantiveis the count that matters. - Language coverage: authority-host rules, superlative lexicons, and original-data phrases ship for EN and ES (including
.gob.mx,.gov.uk,scielo.org). Outside those languages the language-dependent fields come backnull, never 0, and the finding ismanual_review. - Judgment calls (is this claim "strong"? is this source "authoritative"?) require
manual_review; do not auto-passthem. - When the content tier or rendered DOM needed to count passages reliably is unavailable, status is
needs_api, never a falsepass.
Findings
Emit findings per schema/finding.schema.json; axis ai; fixable: advisory throughout. Severity policy: 5 is reserved for catastrophic, eligibility-killing facts at site/template scope; 4 major · 3 moderate · 2 minor · 1 cosmetic · 0 informational — and only an established severity-5 fail in an active category can cap a score (references/scoring-model.md). Nothing in M12 is established, so nothing here ever caps.
M12.density.low_numeric_passages— multiple main-content passages with zero substantive numeric support (warn, severity 3, confidencedirectional).M12.citations.no_outbound_authority— the main content makes factual claims and links to no authoritative outbound source (warn, severity 3, confidencedirectional).M12.original_data.missing— a topic that invites first-party evidence (benchmark, survey, pricing study) carries none (warn, severity 3, confidencedirectional).M12.original_data.present— a first-party data claim backed by a method, sample, table, or chart (pass, severity 3, confidencedirectional).M12.claim.unsourced_superlative— a superlative/comparative claim with no inline citation or data (warn, severity 1, confidencespeculative).M12.lang.unsupported— the page language is outside the EN/ES lexicons, so density and authority scoring cannot decide (manual_review, severity 0; language-dependent fieldsnull). Each finding:evidence.observedquotes the exact passage/claim from the page;verification.reproduceis the runnable count above;expected_impactis banded + confidence-tagged (no naked %).
Honesty
- What Google actually asks for is unique, non-commodity content — information that is not already available on every other page on the topic, produced by people with first-hand experience. Numeric density and outbound citations are our proxies for that, not the thing itself; they are
directionaland they are neverestablished. A page can be dense with numbers and still be commodity content. - Refuse "AI-specific keyword" rewrites — there is no magic vocabulary that wins citations. Citability comes from extractable structure, verifiable facts, and demonstrable authority, not phrasing tricks.
- Never invent a statistic, sample size, or source to "fill" a flagged claim. Quantification only helps if it is true and attributable; a fabricated number is a liability, not a win.
- Density is a means, not an end — stuffing numbers into prose that does not warrant them is its own anti-pattern, and this module flags it as one.
- Outside EN/ES the honest answer is
manual_review, not a pass: a zero count in a language we cannot tokenize is a gap in the tool, not a finding about the page.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

