305 search & AI ranking signal sets, mapped to what your code controls
Modern search and AI answer engines do not score a page with one monolithic formula. They run semantic retrieval, salient term scoring, category-specific quality twiddlers, URL-prefix sitechunks and cross-source AI corroboration. Explore every documented system below or inspect a scanned site.
305Documented signal sets167 per result, 138 per query across all verticals
48Web-page signal sets31 active web scoring sets once non-web verticals are filtered out
34AI search & grounding setsmagi, AI Mode, FoundInSource, Corroboration and Web Guide
17Page-query twin pairsDocument signals weighed only when the matching query demands them
19Actionable page systemsMapped directly to deterministic clickd rules and agent prompts
What do Google’s 305 ranking signals actually measure?
Out of 305 documented RankLab signal sets surfaced through public API schemas and antitrust disclosures, most handle specialized surfaces such as video, maps, local packs or internal logging. Roughly 48 sets evaluate ordinary web pages and 34 govern AI answer synthesis, grouping cleanly into eight engineering families.
Signal family
Documented sets
What it governs
Your engineering lever
Retrieval & crawl eligibility
6
Before any quality twiddler or AI summarizer sees a page, crawlers must fetch server-rendered HTML, obey canonicals, and retrieve passages by semantic embeddings and lexical terms.
Mustang base scoring, salient term extraction from titles, headings, body copy and internal link anchors, plus multi-pass semantic re-ranking.
title, one-h1, description
Quality, expertise & authority twiddlers
13
Separate post-retrieval adjustments for expertise, topic authority, first-hand human voices, hidden gems, authenticity, page category (Orbit) and YMYL scrutiny.
structured-data, schema-fits, named-publisher
Site sections & citation originality
8
Index-level quality attributes attached per URL prefix (sitechunks) and citation-based originality that distinguishes primary data from derivative rewrites.
pages-are-distinct, depth, no-orphans
Page and query twins
17
Systems recorded on both the document and the query so a page is judged against what that specific search demands, such as topic authority or freshness.
linked-in-context, no-orphans, breadcrumb
AI Overviews, AI Mode & corroboration
34
Generative search sets (magi, AI Mode, FoundInSource and Web Guide) that ground answers on retrieved web results and check cross-source corroboration.
Language profiles, geographic service fit and user topic affinity that match a page to who is asking and where they are.
phone-link, next-step, structured-data
How do the four ranking passes map to clickd’s stages?
A page never reaches quality twiddlers or AI Mode citations if it fails early retrieval. Search engines narrow millions of documents down to a few dozen candidates before running expensive neural re-rankers, category classifiers and source corroboration checks.
Pass 1
Crawl & semantic retrieval
Fetches raw HTML, resolves canonicals, indexes passages as meaning vectors and pulls candidate pages into the pool.
Systems: Googlebot / AI bots, PerDocData, RankEmbedclickd stage: Access (35%)
Pass 2
Base scoring & salient terms
Scores lexical and structural relevance using title, main heading, passage structure and internal anchor text.
Selects retrieved web pages whose facts corroborate cleanly and lifts concise answers, tables and lists into AI Overviews and AI Mode.
Systems: magi, AimResultSignals, Corroboration, FoundInSource, WebGuideclickd stage: All four stages + citations
Which ranking and AI corroboration signals apply to your pages?
Filter the 19 page-applicable signal systems below by pipeline family, direct code control or page-query twin behavior. Every card pairs the documented schema family with live failure rates across 2,226,142 scanned sites and a prompt you can hand your coding agent.
19 of 19 systems
Pass 1: RetrievalDirect code control
Indexability & document envelope
PerDocData / indexing envelope
Whether crawlers and AI search bots can fetch a clean 200 HTML document with self-referencing canonicals and real body words before client scripts run.
What moves it: Serve complete HTML from the server, keep robots.txt open to search and AI crawlers, settle http/https/www to one origin, and set self canonicals. Across sites clickd has read, 56% miss “Set each page's canonical address”.
Retrieves candidate pages and passages by vector meaning even when a searcher phrases the problem with different words than the page title.
What moves it: Write clear, self-contained 80-to-150-word passages under specific subheadings so each section embeds cleanly around one concrete concept. Across sites clickd has read, 49% miss “Add depth to thin pages”.
Mustang computes each candidate page’s foundational query score from document zones (title, headings, body text), internal link graph depth and anchor text.
What moves it: Keep important pages within three clicks of home, use one clear h1, write focused 15-to-65-character titles, and link related pages inside body prose. Across sites clickd has read, 87% miss “Fix description lengths”.
Run clickd_signals on example.com and fix pages buried deeper than three clicks or missing contextual internal links.
Pass 2: Base & salienceDirect code controlPage-query twin
Salient terms & anchor coherence
quality_salient_terms.SalientTerm
Summarizes what a page is truly about from its title, headings, body terms and the anchor words other pages use when linking to it.
What moves it: Align your title tag, h1, opening paragraph and incoming internal link anchors on the same core subject terms instead of generic "learn more" links. Across sites clickd has read, 79% miss “Fix title lengths”.
clickd rules:titleone-h1linked-in-context
Run clickd_signals({ site: "example.com" }) to find pages whose title, h1 and incoming link anchors disagree on salient terms, then align them in code.
Transformer and concept passes revisit the top 20 to 30 retrieved documents and score whether specific passages actually satisfy the question behind the search.
What moves it: Phrase subheadings as real user questions and place a direct 25-to-90-word answer immediately underneath before expanding into edge cases. Across sites clickd has read, 36% miss “Phrase headings as questions”.
Re-orders an already scored result set by comparing candidates against top-performing reference results and query-specific layout settings.
What moves it: Cover the complete task a searcher expects for that intent (comparison table, clear pricing/steps, working action) so visitors do not pogo-stick back. Across sites clickd has read, 43% miss “Add a way to get in touch”.
clickd rules:has-tableliftable-blocksnext-step
Compare our key landing pages against rivals with clickd_rivals and clickd_signals to add missing comparison tables and clear next steps.
Classifies every document by format and risk level (tutorial, product, forum, local service, or YMYL health, finance and legal) so the right twiddlers fire.
What moves it: Match your JSON-LD schema type and page structure to the page’s real job, and apply strict author and citation rigor on any YMYL money, health or legal page. Across sites clickd has read, 61% miss “Add structured data”.
Run clickd_signals on example.com to inspect each page’s Orbit role classification and fix every YMYL or commercial page missing its required trust signals.
Pass 3: Quality twiddlersDirect code control
Expertise twiddler
twiddler.expertise.ExpertiseTwiddlerDocFeatures
Adjusts rankings after base scoring using document-level signals of domain rigor: named authorship, verifiable primary citations and precise technical detail.
What moves it: Name the author or publishing organization in visible bylines and Article schema, link primary external sources in body copy, and cut vague sales puffery. Across sites clickd has read, 49% miss “Add depth to thin pages”.
Evaluates authority topic by topic rather than domain-wide, pairing the site’s depth on a subject cluster with the query’s topic settings.
What moves it: Build tightly interlinked topic hubs with breadcrumbs and contextual cross-links instead of scattering orphan posts across unrelated subjects. Across sites clickd has read, 20% miss “Add breadcrumb data”.
Boosts first-hand practitioner experience, real photographs, concrete measurements and genuine human writing over generic, mass-generated summaries.
What moves it: Include real content images with descriptive alt text, concrete measured figures, and strip stock AI phrasing, hidden watermarks and unrendered markdown. Across sites clickd has read, 68% miss “Describe images”.
clickd rules:has-imageimage-altplain-claims
Run clickd_tells and clickd_signals on example.com to strip stock AI filler and add real tables, measured examples and described images.
Index & sitechunksDirect code control
Section-level site quality (NSR Sitechunks)
QualityNsrNsrData / NsrSitechunkSignals
Computes site quality scores by URL directory prefix (sitechunks) as well as domain-wide, so a bloated folder of thin or near-copy pages drags down that section.
What moves it: Audit quality folder by folder (/blog, /locations, /integrations); merge or rewrite doorway templates that share over half their text. Across sites clickd has read, 49% miss “Add depth to thin pages”.
clickd rules:pages-are-distinctdepthno-orphans
Inspect sitechunk health on example.com with clickd_signals and consolidate or rewrite any URL directory weakened by near-copy or thin pages.
Distinguishes pages that publish original figures, benchmarks, datasets and primary reporting from derivative pages that merely restate others.
What moves it: Put original figures in HTML tables, show methodology and dates clearly, cite primary sources, and avoid duplicating your own pages. Across sites clickd has read, 31% miss “Put figures in tables”.
Pairs a document’s publish and update dates against whether the query actually demands recent information. Evergreen queries hold multi-year-old pages.
What moves it: Show accurate visible update dates and matching Article dateModified schema when facts truly change; never auto-bump dates without updating substance. Across sites clickd has read, 12% miss “Show when guides were updated”.
clickd rules:dated
Check our guides with clickd_signals and clickd_traffic to refresh stale facts and set accurate visible <time datetime> and dateModified schema.
Ties a domain to a known organization, local business or publisher across the web so search engines and assistants resolve who stands behind the site.
What moves it: Declare Organization, Person or LocalBusiness JSON-LD on the home page with sameAs links to official social profiles and link those profiles visibly. Across sites clickd has read, 61% miss “Add structured data”.
Selects and orders retrieved web documents to ground AI Overviews and AI Mode answers, including dedicated list and shopping classifiers.
What moves it: Pass web retrieval first (server HTML + open AI bots + top organic relevance), then structure facts in clean lists, tables and Product/Service JSON-LD. Across sites clickd has read, 33% miss “Put the text in the HTML”.
Checks whether specific claims in a generated answer are directly supported in the source page’s text and consistent with structured data and peer sources.
What moves it: Keep visible numbers, prices, FAQs and reviews identical to JSON-LD fields, state unambiguous facts under question headings, and cite primary sources. Across sites clickd has read, 32% miss “Answer the first question right under it”.
Groups results into sub-topic clusters for exploratory queries and runs a relevance oracle pass to drop pages that drift from their promised heading.
What moves it: Keep each page focused on one clear intent, match title and h1 tightly to body sections, and use BreadcrumbList schema to expose hierarchy. Across sites clickd has read, 79% miss “Fix title lengths”.
Demotes doorway template clones, self-serving rating stars with no visible reviews, deceptive claims and compromised markup.
What moves it: Eliminate city-swapped doorway clones, remove AggregateRating schema where reviews are not shown on the page, and replace hype claims with checkable facts. Across sites clickd has read, 50% miss “Link straight to the final address”.
Matches language declarations, geographic service evidence and actionable local contact paths when a query carries local or language intent.
What moves it: Declare accurate LocalBusiness schema, publish crawlable phone and address details, and provide working tap-to-call and quote forms. Across sites clickd has read, 61% miss “Add structured data”.
Run clickd_signals and clickd_check({ kind: "locations", url: "example.com" }) to verify local business schema, phone links and contact paths.
No signal systems match that filter. Clear the search box or pick All.
Which signals fire for your page role (Orbit)?
Not every page is judged by the same twiddlers. Document intent classifiers (Orbit) separate YMYL medical, financial and legal pages from practitioner guides, storefronts, first-hand reviews and developer docs so each format is held to the signals its searcher actually needs.
Named author or specialist organization in both visible byline and Article schema
Visible links to primary statutory, clinical or regulatory sources in the main copy
Direct, hedged factual answers with zero unsupported superlatives or guarantees
Accurate visible review/update date when regulations or thresholds change
Active ranking & AI systems
Orbit YMYL classifier
Expertise twiddler
Fringe content check
Web health & citations
AI Mode corroboration
Skip the busywork: Do not chase conversational slang or publish thin FAQ stubs without primary citations; YMYL queries weight institutional expertise and verifiable sources.
Find every YMYL page on example.com with clickd_signals and ensure each has a named author/publisher, visible primary source links and zero unsupported claims.
Practitioner & technical guide
Examples: How-to tutorials, engineering deep dives, buying guides, industry benchmarks
What the page must show
Question-phrased h2 headings with a 25-to-90-word direct answer right beneath the first question
HTML comparison tables, numbered steps or specification lists that assistants can lift cleanly
Contextual internal links connecting the guide to its parent topic hub and sibling guides
Original measured numbers, code or worked examples rather than generic AI prose
Active ranking & AI systems
Topic authority twiddler
RankBrain & DeepRank
Salient terms
Citation-based originality
FoundInSource
Skip the busywork: Do not auto-bump dateModified every week if the guide’s facts did not change; evergreen technical pages rank for years when substance holds.
Matching Product, Service or LocalBusiness JSON-LD whose prices, availability and ratings match visible HTML
Focused title and single h1 aligned with the exact service or product terms used in internal link anchors
Distinct, specific substance on every location or variant page (never city-swapped doorway templates)
Working conversion path: checkout, quote form with real action, or tap-to-call link for local trades
Active ranking & AI systems
Ascorer base ranker
Shopping & Magi classifiers
Salient terms
Sitechunk quality
Local & action checks
Skip the busywork: Do not stuff 2,000 words of blog fluff below a product grid or spin 50 near-identical town pages that drag down your /locations sitechunk.
Check commercial and service pages on example.com with clickd_signals to fix schema mismatches, doorway overlap and missing conversion actions.
First-hand case study, review & comparison
Examples: Tested product reviews, /vs/ competitor pages, customer case studies, field notes
What the page must show
Real photographs, screenshots or charts in HTML img/figure tags with descriptive alt text
Side-by-side HTML comparison tables stating honest trade-offs, including when an alternative fits better
Specific tested versions, dates and measured outcomes instead of stock AI adjectives
Active ranking & AI systems
Authenticity twiddler
Human voices
Hidden gems (Natural Pearls)
For Loop Ranker
Web Guide
Skip the busywork: Never fabricate AggregateRating schema or write one-sided /vs/ hit pieces; unbacked claims fail both user trust and corroboration checks.
Audit our /vs/ and review pages with clickd_signals and clickd_tells to strip stock phrasing and add honest comparison tables and described media.
Documentation, API & specification
Examples: /docs/*, API endpoints, CLI flags, integration setup guides, changelogs
What the page must show
Server-rendered HTML (not an empty client-rendered docs shell) and an up-to-date /llms.txt index
Exact parameter tables, copyable code blocks and one unambiguous h1 per endpoint or concept
BreadcrumbList schema and clean canonical URLs without duplicate version or trailing-slash traps
Active ranking & AI systems
RankEmbed
Salient terms
FoundInSource
Breadcrumb hierarchy
Skip the busywork: Do not pad concise API reference pages to hit arbitrary blog word counts; reference pages win by precision, tables and clean structure.
Run clickd_signals on our /docs section to verify server-rendered HTML, BreadcrumbList schema and salient term alignment.
How do I run signal diagnostics from an AI coding agent?
Install clickd’s open agent skills or connect over MCP, then ask your agent to run clickd_signals on your domain or a single staging/live URL. Your agent inspects weak URL sitechunks, misaligned salient terms and missing Orbit role checks, edits your repository, and rescans after deploy.
1. Add the skillsnpx skills add subinternal/clickd
2. Tell your agentFind and fix the biggest SEO problems on yoursite.com with clickd.
Frequently asked questions
Does Google really use 300+ ranking factors on every web page?
No. Documented RankLab schemas list 305 signal sets across Google's wider search stack (167 per result and 138 per query), including video, maps, shopping, news and query-understanding modules. Roughly 48 signal sets apply directly to standard web pages (31 scoring sets), alongside 34 sets for AI Overviews, AI Mode and source corroboration.
Does this signal lens change clickd’s 0-to-100 score?
No. clickd’s 45-rule 0-to-100 score remains stable across Access (35%), Understand (30%), Trust (20%) and Act (15%). Ranking signal diagnostics sit alongside the rulebook as a companion lens (via clickd_signals) to inspect URL-prefix sitechunks, Orbit page-role fit, salient term alignment and AI corroboration.
What is a sitechunk and why can one folder drag down rankings?
Quality systems such as NSR compute quality attributes by URL directory prefix (for example /blog, /locations or /integrations) as well as domain-wide. When a folder accumulates dozens of city-swapped doorway clones, thin stubs or unedited AI prose, that section’s sitechunk weakens even if your home page is strong.
What are page-query twins in modern search ranking?
At least 17 signal families record matching structures on both the document and the query, including topic authority, freshness and category intent. A page is not scored in a vacuum: freshness boosts matter only when the query demands recent facts, while YMYL expertise checks fire hardest when the query touches health, money or law.
How do AI Overviews and AI Mode decide which pages to cite?
Generative search pipelines retrieve candidate pages from the core web index first, then run dedicated ranking and corroboration passes (such as magi, Corroboration and FoundInSource) that check whether concise claims, numbers, tables and JSON-LD fields match visible source text and peer sources.
How can my coding agent check these signals on our site?
Connect your agent to clickd over MCP at clickd.com/mcp/all (or install the open skills with npx skills add subinternal/clickd) and call clickd_signals with { site: "example.com" } for a full sitechunk and page-role audit, or { url: "https://example.com/page" } to test any single live URL.
Sources
Signal schema names reflect public Google Content Warehouse / RankLab protocol buffer definitions and U.S. antitrust trial disclosures, paired with official Google Search Central and crawler specifications.
Signal diagnostics inspect only what your own HTML, links and site sections put on the page. Search engines and assistants do not publish weights, and passing every check does not promise rankings, traffic or citations.