The SEO Playbook & Working Checklist¶

The canonical working document for our SEO projects — every line verified against official Google documentation¶

Verification status: double-audited. Round 1 (Aug 2026): 160 claims checked against official Google sources. Round 2: every cited URL re-fetched live, every quote grepped against page HTML, error-level findings independently confirmed by a second reviewer, plus a completeness audit. Full evidence trail in seo-verification-report.md.

Legend — where each item actually comes from: - 🟢 Google-documented — stated in an official Google source (linked) - 🟡 Google-backed in spirit — Google supports the principle; the specifics (numbers, thresholds, "always/never") are our own working standard - ⚪ Industry practice — sensible and widely used, but Google is silent; never cite it as "Google says" - ⚠️ — a correction or trap worth special attention

How we use this together¶

Tell me the project type and I'll run the right phases with you:

Project	Phases to run
New site build	0 → 1 → 2+3 per page → 4 if local → 5+6 as programs
Existing site audit	1 → 3 on a sample of key pages → 4 if local → 7 to baseline
Local business	0 (light) → 1 (light) → 4 → 7 monthly
Content sprint	0 (cluster planning) → 2+3 per piece
Small ecommerce	1 + Ecommerce module → 2 per template → 7
Traffic drop emergency	Phase 7 diagnosis sequence, in order
Domain change / replatform	Migration playbook (event-driven section)
Hacked site	Hacked-site playbook (event-driven section)

The mental model (read once — everything below hangs off this)¶

1. Search is a pipeline, and the checklist follows it. Google crawls (Googlebot fetches pages; robots.txt gates this) → renders (executes JavaScript) → indexes (picks canonicals; noindex removes pages here) → ranks (many systems score indexed pages per query) → displays (titles, snippets, rich results, AI Overviews). Phase 1 covers crawl→index. Phases 2–3 cover rank. Titles/snippets/structured data/AI features are the display layer — they change clicks, not rankings. A problem at any stage makes everything downstream irrelevant, which is why the phases run in this order.

2. There are three different rulebooks. Confusing them is the most common SEO mistake. - Technical requirements — binary. Page is indexable or it isn't. No judgment involved. - Spam policies — named violations (see ⛔ list). Enforced by spam updates and manual actions (a human at Google flags your site; you see it in Search Console and must fix + file reconsideration). Consequence: "rank lower or not appear in results at all." - Quality systems — gradients, not violations. Core updates, E-E-A-T weighting, helpfulness signals. There is no "penalty" to lift and no manual action — your content simply loses to better content, and the fix is being better, which takes months, not days.

3. Ranking is page-level, with site-wide signals on top. Google: "Our ranking systems are designed to work on the page level... Site-wide signals and classifiers are also used and contribute to our understanding of pages." A weak section can drag on the whole site — but good site-wide signals don't guarantee every page ranks, either. This is why Phase 0 asks whether planned content sits inside the site's genuine expertise.

4. Impressions vs clicks is the master diagnostic. Impressions = you appeared. Clicks = you got chosen. Impressions steady + clicks down → a display-layer problem (snippet appeal, rival rich results, AI Overviews absorbing clicks). Both down → a ranking or indexing problem. Phase 7 is built on this split.

5. The AI layer sits on the same index. AI Overviews and AI Mode are Search features: eligibility = indexed + snippet-eligible, full stop. No special markup, no AI files, no separate crawler for Google's AI surfaces (Google-Extended is about Gemini, not Search). Normal SEO is the AI strategy; the extra work is measurement (new GSC reports) and content positioning (original, first-hand material that a summary can't replace).

6. Know your source hierarchy. Google documentation (the 🟢 items) > Googler informal statements (X posts, office hours — real but unofficial, and they change) > industry studies and lore (⚪ — often useful, never "Google says"). This checklist labels every line so you always know which one you're holding.

Phase 0 — Strategy & Intent (before touching the site)¶

Why this phase exists: every later decision gets judged against what the site is for and what it can credibly claim expertise in. Skipping this is how sites end up publishing outside their lane and tripping site-wide quality signals.

⚪ Define what the site must actually achieve (leads / sales / ad revenue / bookings) — every SEO decision gets judged against this
🟢 Identify the site's genuine area of expertise. Google's self-assessment explicitly flags "producing lots of content on many different topics in hopes that some of it might perform well" and entering "some niche topic area without any real expertise... because you thought you'd get search traffic" as search-engine-first content — Creating helpful content
🟡 If planned content falls outside the site's expertise, flag the risk from site-wide signals/classifiers. Google confirms site-wide signals and classifiers exist and contribute, but ranking is primarily page-level, and the old helpful-content classifier no longer exists as a single system (absorbed into core, March 2024) — Ranking systems guide
🟢 Check if the niche is YMYL. Official Google concept (Quality Rater Guidelines §2.3): topics that could significantly impact "the health, financial stability, or safety of people, or the welfare or well-being of society" (legal/civic topics fall under these axes). Google's systems "give even more weight to content that aligns with strong E-E-A-T" on these topics — Creating helpful content
🟡 If YMYL: demonstrable expertise is heavily weighted — Google's rater guidelines say some YMYL advice "must come from experts" (though first-hand-experience content can also earn high E-E-A-T). Formal credentials are strongly advisable for advice content but are not a stated Google ranking requirement
⚪ Keyword/topic research: cluster by intent, one page per intent, pillar + supporting structure planned (industry methodology — Google prescribes no site architecture model)
⚪ For every target query: google it and record what page type ranks and which SERP features appear (AIO? local pack? shopping?) — useful intent recon, but reading intent from SERPs and "click ceilings" are industry practice, not Google guidance
🟡 Competitor gap analysis: what do top results cover, and what original information/value can you add? Google's self-assessment asks for "original information, reporting, research, or analysis" and "substantial value when compared to other pages in search results." ("Information gain" is a Google patent term, not documentation; "top 3" is an arbitrary convention)
🟢 Check target queries for AI Overview presence and note which sites get linked. Eligibility is simply: indexed + snippet-eligible; Google states there are no special optimizations for AI features — any "why cited" pattern analysis is your own inference — AI features and your website

Phase 1 — Technical Audit (day-one health check)¶

Why this phase exists: this is the crawl→render→index half of the pipeline. Nothing you do in content or links matters for a page Google can't crawl, render, or index. Run it before promising anyone anything.

Access & indexation - [ ] 🟢 GSC property type deliberate: a Domain property covers all subdomains and protocols (DNS verification); URL-prefix covers only the exact prefix+protocol — prefer Domain unless you need scoped data — property types. ⚪ On takeover: audit Users & permissions, remove departed owners/agencies - [ ] 🟢 robots.txt reviewed — nothing important blocked. Google: if the absence of blocked resources "make[s] the page harder for Google's crawler to understand the page, don't block them"; and robots.txt "is not a mechanism for keeping a web page out of Google" — robots.txt intro - [ ] ⚪ AI crawler policy for GPTBot, ClaudeBot, PerplexityBot is a deliberate decision (third-party crawlers — outside Google's docs by definition) - [ ] 🟢 ⚠️ Google-Extended controls Gemini training/grounding ONLY. It does not affect Google Search, AI Overviews, or AI Mode — those follow normal Googlebot indexing + snippet controls. Google: "Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal" — Google crawlers - [ ] 🟢 ⚠️ Index verification via GSC Page indexing report + URL Inspection, NOT site: counts. Google explicitly warns the site: operator "doesn't necessarily return all the URLs that are indexed" and shouldn't be used to count indexed pages — site: operator doc - [ ] 🟢 GSC connected; Page indexing report (renamed from "Index Coverage" in 2022) reviewed — report help - [ ] 🟡 "Crawled/Discovered – currently not indexed" at scale: investigate. Google's crawl-budget doc ties crawl demand to page quality for large sites, but the status itself is officially described as normal, not a quality verdict - [ ] 🟢 No stray noindex on money pages — block indexing - [ ] 🟢 No noindex + robots.txt-block combos — "For the noindex rule to be effective, the page... must not be blocked by a robots.txt file" - [ ] 🟢 XML sitemap lists the canonical URLs you want in results, submitted in GSC (200-status/indexable-only is sound practice; Google states the canonical part) — build a sitemap - [ ] 🟢 Manual Actions and Security Issues reports in GSC: clean (if Security Issues is NOT clean → Hacked-site playbook below) — manual actions, security issues

Canonicalization & duplication - [ ] 🟢 One version resolves: https + one hostname; all others permanent redirect (301 or 308 — both qualify per the redirects doc) — consolidate duplicate URLs, redirects - [ ] 🟢 rel=canonical correct sitewide (remember: Google treats it as a hint, not a directive) — canonicalization - [ ] 🟢 Parameter URLs and filters handled — Google's Dec 2024 faceted-navigation guidance: disallow crawl of filter URLs, canonical to the unfiltered version, 404 nonsense combinations — faceted navigation - [ ] 🟢 ⚠️ Pagination: each page keeps its own self-canonical — Google: "Don't use the first page of a paginated sequence as the canonical page." Sequential <a href> links between pages; rel=next/prev is ignored by Google; infinite-scroll needs paginated URL equivalents (Googlebot "doesn't 'click' buttons") — pagination guidance - [ ] 🟡 No duplicate title/content clusters (unique titles and consolidating duplicates are Google-documented; duplicate H1s are an industry/usability check; duplicate content is a canonicalization issue, not a penalty)

International (multilingual / multi-regional sites — e.g. BM + English + Chinese) - [ ] 🟢 hreflang implemented via ONE method (HTML link tags, HTTP headers, or sitemap): every version lists itself + all alternates with fully-qualified URLs, plus x-default. ⚠️ Return links are mandatory — Google: "If two pages don't both point to each other, the tags will be ignored." Valid ISO codes only (no "UK"/"EU") — localized versions. ⚪ Crawler-based hreflang audits (Screaming Frog/Ahrefs) are industry tooling

Architecture & rendering - [ ] 🟡 Important pages reachable through crawlable internal links, ideally linked from the homepage or other prominent pages. No official click-depth number exists — "≤3 clicks" is industry lore - [ ] 🟡 No orphan money pages (links are Google's primary discovery mechanism; no formal "no orphans" rule — sitemap-only URLs can still be indexed) - [ ] 🟢 Internal anchor text is descriptive — Google's doc literally lists "Click here to learn more" and "Read more" as bad examples — links best practices - [ ] 🟢 URL Inspection → rendered HTML contains the actual content and links (JS sites especially) — JavaScript SEO basics - [ ] 🟡 Minimize redirect chains — Google follows up to 10 hops but says long chains "have a negative effect on crawling"; redirect directly to the final URL ("≤1 hop" is our own tightening) - [ ] ⚪ No 404s in internal links (404s are normal web behavior per Google; fixing broken internal links is UX/crawl-efficiency practice) - [ ] 🟢 No soft-404s — Google: "soft 404 pages will continue to be crawled, and waste your budget" — crawl budget management

Performance & experience - [ ] 🟢 Core Web Vitals field data: LCP ≤2.5s, INP ≤200ms, CLS ≤0.1 at p75 — exact official thresholds (INP replaced FID March 2024). p75 is assessed segmented by device — mobile and desktop evaluated separately; prioritize mobile — web.dev/vitals - [ ] 🟢 Mobile parity: same content, same structured data, equivalent titles/metas as desktop (mobile-first indexing) — mobile-first indexing - [ ] 🟢 HTTPS everywhere, no mixed content. Framing: CWV "are used by our ranking systems," but other page-experience aspects "don't directly help your website rank higher" — HTTPS is best practice + canonical preference, not a ranking boost — page experience - [ ] 🟢 No intrusive interstitials (they "make it hard for Google... to understand your content"; exceptions for legal/age gates) — interstitials

Structured data & brand appearance - [ ] 🟢 Organization schema on home/about — Google recommends exactly this placement, for entity understanding/disambiguation (knowledge panels) — Organization. ⚠️ For AI surfaces: "no special schema.org structured data that you need to add" — don't promise AI visibility from markup — AI features - [ ] 🟢 Site name + favicon controlled: WebSite structured data with name/alternateName on the homepage (root URI only; keep og:site_name/title consistent); favicon <link rel="icon">, square, ideally >48×48px, favicon and homepage both crawlable — site names, favicons - [ ] 🟢 Correct type per template: Article, Product, LocalBusiness, BreadcrumbList — plus author markup inside Article + ProfilePage for author bio pages (standalone "Person" is not a Google search feature) — structured data gallery - [ ] 🟢 ⚠️ Do NOT build for deprecated rich results: HowTo removed from all results (2023); FAQ rich results stopped appearing May 7, 2026 (docs removed June 15; previously gov/health-only since Aug 2023). Leftover markup is harmless but earns nothing — strip it from templates; never sell "FAQ schema for rich results" — docs changelog · deep dive: technical-seo/12-structured-data.md §12.5 - [ ] 🟢 Validated in Rich Results Test (dev) + Rich result status reports (deployed); zero spammy/false markup — draws a "Structured data issue" manual action; penalty is rich-result ineligibility — SD policies

Phase 2 — Per-Page On-Page (run for every page you create or optimize)¶

Why this phase exists: this is where a page tells Google — and the searcher scanning results — what it's about. Most items here are display-layer: they win the click, not the ranking.

⚪ Page targets ONE intent; not duplicating an existing page's target ("keyword cannibalization" is an industry concept — check overlap via GSC query/page data; site: search is only a rough signal)
🟢 Title: descriptive, concise, honest, no keyword stuffing — Google's actual guidance. ⚪ "Front-loaded" and "~50–60 chars" are industry display heuristics — Google states no character limit; truncation is by device width — title links
🟡 H1 clearly states the page topic. H1 count is not a Google factor — heading count/order "doesn't matter" for Search; single H1 is an accessibility convention. ⚪ Question-phrased H2/H3s: industry tactic, fine where natural
🟡 Direct, clear answer to the core query up front — good for readers, consistent with people-first guidance. ⚪ The "first 100–200 words" and "40–70 word block" numbers are third-party. Google says you can't mark or opt a page in as a featured snippet ("You can't" — its systems decide what to elevate), and there are no special optimizations for AI Overviews
🟢 URL descriptive and hyphenated (both explicitly recommended); keep it simple — Google recommends a "simple URL structure" and trimming unnecessary parameters; consistent casing (lowercase by convention) — URL structure
🟢 Meta description written to earn the click — Google's framing: "like a pitch" that can improve "the quality and quantity of your search traffic"; no length limit; Google may substitute page text — snippets
🟡 Page linked from relevant pages with descriptive anchors, and contextual links OUT to related pages (outbound linking "can help establish trustworthiness"). ⚪ "3+" is our working target — Google: "There's no magical ideal number of links"
🟢 Images: modern formats (WebP/AVIF — Lighthouse recommends, Google Images indexes), descriptive alt text ("the most important attribute"), lazy-load below the fold only — never the LCP image, and content must load on viewport entry since Google Search doesn't scroll — lazy-loading
🟢 Schema appropriate to page type, matching visible content, validated — SD intro

Conditional modules (apply when relevant)¶

Ecommerce - [ ] 🟢 Product pages carry the right Product markup class: merchant listing markup (price, availability, shipping, returns) where users can buy on-page; product snippet markup where they can't. Pair with a Merchant Center feed — Google: "Providing both structured data on web pages and a Merchant Center feed maximizes your eligibility" — Product structured data. ⚪ Automated feed sync to avoid disapprovals is industry practice

Video - [ ] 🟢 Video embedded via <video>/<iframe>/<embed> so Google can find it; key videos get a dedicated indexed watch page where "the video is the main content"; every video has "a valid thumbnail that's available at a stable URL" (⚠️ missing thumbnail = not indexed as video); VideoObject + video sitemap — video best practices

Content sites → Google Discover - [ ] 🟢 Discover-eligible by default (indexed + content policies; no special markup), but enable large previews: max-image-preview:large + compelling images ≥1200px wide; monitor the Discover performance report in GSC. ⚠️ Google's own warning: Discover traffic is "less predictable or dependable" than search — never forecast on it — Google Discover

Phase 3 — Content Quality Gate (Google's people-first self-assessment — be brutal)¶

Why this phase exists: this is the quality-systems rulebook — the one core updates enforce. There's no penalty to avoid here, only the question of whether a searcher who lands on this page would need to go back and search again. Score honestly; these questions ARE Google's core-update self-assessment (the core updates doc points to exactly these).

(The old "HCS" name is retired — the helpful content system was absorbed into core ranking systems in March 2024; the questions live in Creating helpful content.)

Phase 4 — Local SEO (any business with location or service area)¶

Why this phase exists: local results run on their own documented triad — relevance, distance, prominence — and on Google Business Profile, which has its own policies and its own suspension regime. For a local business this phase usually moves the needle more than everything else combined.

Foundation: 🟢 Google's official local ranking factors are relevance, distance, prominence (the page's intro sentence now words the third as "popularity"; the detailed sections are still Relevance / Distance / Prominence) — improve local ranking

🟢 GBP claimed + verified ("more likely to show up in search results"), complete and accurate info, categories exact and specific ("The categories you select affect your local ranking") — categories help
🟡 Primary category set precisely — Google confirms categories affect ranking; "primary category = biggest lever" is an industry survey finding (Whitespark), not Google's statement
🟡 Real photos uploaded (officially encouraged; no stated frequency or ranking effect); Posts used (documented engagement feature; no stated ranking effect). ⚠️ GBP Q&A was retired (Q&A API discontinued Nov 3, 2025; public Q&A replaced by AI-generated answers) — put FAQ content on your own site and GBP business description, where Google's AI answers can source it
🟢 Review program: ask all customers post-service (Google supports asking via review link/QR — but prohibits selectively soliciting positive reviewers and review gating); reply to reviews ("More reviews and positive ratings can help your business's local ranking" — official); zero purchased/incentivized reviews ever — review policy
🟡 Business info accurate and consistent with your real-world identity on site + GBP. ⚪ Identical NAP across Facebook/directories is industry citation practice — Google states no cross-directory consistency factor
🟡 One genuinely unique page per location/service-area with LocalBusiness schema (display eligibility, not ranking). ⚠️ City-swapped boilerplate is doorway abuse per Google's own example; mass-generating such pages can also qualify as scaled content abuse. ⚪ Embedded map: usability practice
🟢 Business name in GBP = real-world name, no keyword stuffing — "could result in the suspension of your Business Profile" ("Vicinity" is the industry nickname for the Nov 2021 local update, not Google's name) — name guidelines
🟡 Local links build prominence — Google confirms "how many websites link to your business" feeds prominence. ⚠️ Paid sponsorship links must carry rel="sponsored"/nofollow or they're link spam

Phase 5 — Links & Brand (ongoing program, not a task)¶

Why this phase exists: links remain a discovery mechanism and a prominence/trust input, but the link rulebook is almost entirely the spam-policies one — the wins are earned, the shortcuts are all named violations.

🟡 Earn links via genuinely link-worthy content — Google: other sites linking to you "happens naturally over time" (SEO Starter Guide). ⚠️ Partnership/supplier/association links from commercial relationships must be qualified rel="sponsored"/nofollow — requiring a link by contract "without allowing a third-party content owner the choice of qualifying the outbound link" is listed link spam — spam policies
🟡 Never create/commission links with optimized exact-match anchors ("links with optimized anchor text in articles, guest posts, or press releases" = listed link spam). ⚪ Brand-vs-exact anchor ratios are industry heuristics — Google documents no ratio
🟢 Disavow ONLY if both conditions hold: "a considerable number of spammy, artificial, or low-quality links" AND "the links have caused a manual action, or likely will" — Google's exact criteria; misuse "can potentially harm your site's performance" — disavow doc. Google devalues most junk automatically ("most sites will not need to use this tool"; SpamBrain "neutralize[s] the impact of unnatural links" — Dec 2022 link spam update). ⚪ Quarterly review cadence: our habit
🟢 Brand demand tracked with GSC's Branded queries filter (launched Nov 2025, all eligible sites since Mar 2026; top-level properties with sufficient volume; includes Insights brand-recognition card) — announcement. ⚪ "Rising brand = update resilience + AI visibility" is industry observation — Google says the filter "has no effect on how Google Search ranking works"
⚪ Presence where LLMs learn (profiles, databases, genuine community participation, Wikipedia/Wikidata only if merited) — aimed at third-party LLMs. ⚠️ Google warns: "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem," and forum comments with optimized links are listed link spam
🟢 Qualify outbound links: rel="sponsored" for paid/advertising links, rel="ugc" for user-generated content, rel="nofollow" when other values don't apply and you'd rather Google not associate with or crawl the target — qualify outbound links

Phase 6 — AI-Era Visibility (GEO/AEO layer)¶

Why this phase exists: AI Overviews/AI Mode changed how often searchers click, not how Google indexes. The baseline is plain SEO; what's genuinely new is measurement and the premium on content an AI summary can't replace.

Google's official baseline (all 🟢, from AI features, the May 2026 AI optimization guide, and the June 2026 controls announcement): - Eligibility for AI Overviews/AI Mode = indexed + snippet-eligible. Nothing else. "No additional requirements... nor other special optimizations necessary" - No special markup, AI files, or schema needed; llms.txt-style files "neither harm nor help" - Snippet controls apply: nosnippet, data-nosnippet, max-snippet, noindex limit what AI features show; the new Search Console "Search generative AI" control (Settings → Search generative AI) adds a dedicated opt-out — rolling out gradually (UK subset first, global to follow) — help doc - Google-Extended does NOT govern AI Overviews/AI Mode (Gemini training/grounding only)

🟡 Every key question answered in clear, well-structured, self-contained prose. ⚪ The "40–70 word passage" figure is third-party GEO lore — Google: "There's no ideal page length"
🟡 Track AI-feature visibility: GSC's Search Generative AI performance reports (rolling out from June 2026 to a subset of sites: AIO + AI Mode impressions by page/country/device/date, plus a companion Discover report — no query breakdown, no clicks), plus third-party tools for per-query citation share — announcement, report doc
🟢 AI-assistant referrals segmented in GA4 via the built-in "AI Assistants" channel (May 2026; medium ai-assistant — ChatGPT, Copilot, Gemini, Grok, etc.). ⚠️ AIO/AI Mode clicks are NOT in this channel — they stay in google/organic — GA4 channels
🟡 AI crawler access matches strategy: for Google AI features, visibility = Googlebot indexing + snippet eligibility; for ChatGPT/Perplexity/etc., check each vendor's crawler docs — blocked crawler = invisible in that engine
🟡 Content weighted toward original research, first-hand tests, unique perspectives — Google-endorsed ("people are... more likely to click into an in-depth review, an original post, a unique perspective"). ⚪ Bottom-funnel/local/tools resilience is reasonable industry inference
⚪ CTR-compression figures (30–60%, Seer/Ahrefs/Pew) are third-party. ⚠️ Google's counter-position (Aug 2025): aggregate clicks "relatively stable," click quality up, disputes "dramatic decline" reports — while publishing no data. Hold both: plan for per-query CTR risk on informational queries; never cite the third-party figures as Google-confirmed

Phase 7 — Monitoring & Update Response (monthly + on every traffic move)¶

Why this phase exists: diagnosis before treatment. The single most expensive SEO mistake is "fixing" the wrong cause of a drop — the sequence below forces you to identify the layer (tracking → update → display → ranking → technical) before touching anything.

Event-driven playbooks¶

Site move / domain change / replatform — the highest-stakes event in SEO; botched migrations lose years of equity overnight: - [ ] 🟢 Page-level 301/308 redirect map old→new (no chains); keep redirects "at least 1 year" (Google's stated minimum; the Change of Address tool's signal-forwarding window is only 180 days — keep the redirects the full year regardless); verify BOTH old and new sites in GSC including all subdomain and www/non-www variants (per the June 2026 doc update, submit CoA for each); submit Change of Address for domain moves (not needed for HTTP→HTTPS); update canonicals + sitemaps to new URLs; monitor Page indexing on both properties — site moves · deep dive: technical-seo/16-redirects-migrations.md. ⚪ Migrating in sections and freezing other changes during the move: industry risk management

Hacked site (Security Issues flagged, or SERPs show gibberish/Japanese keywords/cloaked spam): - [ ] 🟢 Follow Google's recovery sequence: quarantine the site → assess damage via Search Console → clean and patch → request review to clear the flag. ⚠️ "Hacks are often invisible to users" — inspect rendered HTML via URL Inspection as Googlebot, not just your browser — help for hacked sites. ⚪ Full credential rotation, malware scan, CMS/plugin updates before review: standard incident practice

⛔ Never-Do — Google spam & platform policies (manual-action / spam-update risk)¶

Verified against the spam policies page. Stated consequence: sites "may rank lower in results or not appear in results at all" — via spam updates (algorithmic) or manual actions (human-flagged, visible in GSC, reconsideration path available). "Deindexed" is the ceiling, not the default.

⛔ Scaled content abuse — "many pages generated for the primary purpose of manipulating search rankings and not helping users." Method explicitly irrelevant: "no matter whether content is produced through automation, human efforts, or some combination" (Google, March 2024)
⛔ Site reputation abuse — third-party content published mainly to exploit the host's established ranking signals (Google's coupon/casino/payday-loan examples). No first-party involvement or oversight exempts it (Nov 2024 clarification); a dedicated manual action exists and is issued
⛔ Expired domain abuse — expired domains repurposed "primarily to manipulate search rankings by hosting content that provides little to no value" (reusing an old domain for a genuine site is fine)
⛔ Link spam — buying/selling links for ranking, "excessive link exchanges," widely distributed footer/template links, optimized-anchor links in guest posts/press releases, contractually required unqualified links. PBNs/"link networks" are industry terms Google never uses — but they fall squarely under the definition (links created "primarily for the purpose of manipulating search rankings"), and SpamBrain detects "sites used for the purpose of passing outgoing links"
⛔ Doorway abuse — including Google's own example: city/region-swapped template pages funneling users to one destination
⛔ Cloaking, hidden text/links, sneaky redirects — all named policies with corresponding manual actions
⛔ Spammy/misleading structured data — violates the structured data policies; draws a "Structured data issue" manual action (its own policy set, not the spam-policies page)
⛔ Buying or incentivizing reviews — prohibited under GBP/Maps fake-engagement policy. Documented consequences: review removal, temporary blocks on new reviews, unpublished ratings, public warning banner, up to account-level suspension. (Legal exposure — e.g. the US FTC fake-review rule — is real but separate from Google policy)

⛔ Never-Do — quality/E-E-A-T guidance (algorithmic ranking harm, not spam policies)¶

No manual action exists for these — the harm is algorithmic (core-update/quality systems) and reputational:

⛔ Fake bylines, fabricated credentials, invented "experts" — undermines the E-E-A-T/trust signals Google's systems and quality raters assess (raters don't directly demote sites; readers and competitors find fakes)
⛔ Fake freshness — date changes without substantive changes; flagged in Google's self-assessment as a search-engine-first tactic (question added Sept 2023)
⛔ YMYL content without demonstrable expertise — Google gives "even more weight" to strong E-E-A-T on health/financial/safety topics; weak-expertise YMYL content is algorithmically disadvantaged

Official Google features shipped 2024–2026 (all verified real)¶

Feature	Status	What it does
Branded queries filter (GSC)	Nov 2025; all eligible sites Mar 2026	Auto-splits branded vs non-branded; Insights brand-recognition card
Search Generative AI performance reports (GSC)	June 2026, rolling out from a subset of sites	AIO + AI Mode (plus gen-AI in Discover) impressions by page/country/device/date; no queries/clicks
GA4 "AI Assistants" channel	May 2026	Auto-segments ChatGPT/Copilot/Gemini/Grok referrals (`medium=ai-assistant`); excludes AIO/AI Mode
Search generative AI control (GSC toggle)	Announced June 2026; UK subset first, global to follow	Dedicated opt-out from AI Overviews/AI Mode/gen-AI in Discover (Settings → Search generative AI)
AI optimization guide	May 2026	Google's official GEO doc: no special optimizations; llms.txt "neither harm nor help"
Faceted navigation guidance	Dec 2024	Official handling for filters/parameters
⚠️ GBP Q&A retired	API off Nov 3, 2025; public Q&A → AI answers	Remove Q&A tasks from local workflows; FAQ content goes on-site/GBP description

Glossary — the jargon, decoded¶

Crawling / indexing / ranking — fetching your page / storing it in Google's database / scoring it for a query. Three separate stages; each can fail independently.
Googlebot — Google's web crawler. What it can't fetch or render doesn't exist to Google.
Canonical — the one URL Google picks as the "real" version when duplicates exist. rel=canonical is your hint about which to pick.
noindex — a page-level instruction to stay out of search results. Only works if the page is crawlable (not robots.txt-blocked).
Crawl budget — how much Google will crawl on your site; only a real constraint for large sites (roughly 10k+ frequently-changing pages).
SERP — search engine results page. Impressions = times you appeared; CTR = clicks ÷ impressions; average position = your typical rank when shown.
CWV (Core Web Vitals) — Google's three field-measured speed/stability metrics: LCP (loading), INP (responsiveness), CLS (visual stability), each judged at the 75th percentile of real users.
E-E-A-T — Experience, Expertise, Authoritativeness, Trust. Not a score or ranking factor — a rubric from the Quality Rater Guidelines describing what Google's systems try to reward, with Trust as the most important member.
YMYL — "Your Money or Your Life": topics affecting health, financial stability, safety, or society. Held to higher E-E-A-T standards.
Quality Rater Guidelines — Google's manual for ~16k human raters who evaluate search results. Ratings benchmark the algorithms; they do not directly move any site's rankings.
Core update — a periodic recalibration of Google's quality systems. Losing in one isn't a penalty; it's being outscored under the new calibration.
Manual action — a human-issued sanction for a spam-policy violation. Visible in Search Console; fixable via reconsideration request. If it's not in GSC, you don't have one.
Rich results — enhanced SERP displays (stars, prices, breadcrumbs) powered by structured data (schema.org markup). Display-layer only.
AIO / AI Mode — AI Overviews (AI summary atop results) and AI Mode (conversational search tab). Both cite/link source pages; eligibility = normal indexing + snippet eligibility.
GEO / AEO — "generative engine optimization" / "answer engine optimization": industry names for optimizing visibility in AI answers. Google's official position: regular SEO is the whole game.
GBP / NAP — Google Business Profile; Name-Address-Phone consistency.
hreflang — annotations telling Google which language/region version of a page to serve. Fails silently without return links.
Doorway pages / cloaking / PBN — spam: near-duplicate pages funneling to one target / showing Google different content than users / networks of sites built to pass links.
Disavow — telling Google to ignore specific backlinks. Almost never needed; can hurt if misused.

The name-drop decoder (updates you'll hear about)¶

Name	Year	Official?	What it was
Panda	2011	🟢 in ranking systems guide (retired)	Content quality; absorbed into core
Penguin	2012	🟢 in ranking systems guide (retired)	Link spam; absorbed into core
Hummingbird	2013	🟢 named by Google (retired)	"A major improvement to our overall ranking systems" — the conversational/semantic framing is press history, not Google documentation
Pigeon	2014	⚪ industry-coined (Search Engine Land)	Local search update; zero official Google documentation exists
RankBrain / BERT / MUM	2015/2019/2021	🟢 active in ranking systems guide	AI systems for understanding queries/language
"Vicinity"	Nov 2021	⚪ industry-coined (Sterling Sky)	Google confirmed only a "November 2021 local search update" — rebalanced local factors, hit keyword-stuffed business names
HCS (helpful content system)	2022–2024	🟢 named by Google (retired)	Site-wide people-first classifier; absorbed into core March 2024 — "no longer one signal or system"

Full claim-by-claim evidence: seo-verification-report.md. Superseded draft archived in archive/.

Deep technical reference: technical-seo/ — 21 chapters of verified mechanics behind Phases 1–2 and the event playbooks (chapter map: technical-seo/00-index.md; symptom→chapter dispatch: technical-seo/21-diagnosis.md §21.5).