Decoding the Google Leak: What It Means for SEO in 2026 by Shaun Anderson at SMX Paris 2026

Decoding the Google Leak: What It Means for SEO in 2026

Here is the full breakdown of the presentation "Decoding the Google Leak: What It Means for SEO in 2026" by Shaun Anderson at SMX Paris.

The presentation provides an evidence-based roadmap analyzing the 14,014 leaked Google ranking attributes and translates them into actionable framework layers based on E-E-A-T (Experience, Expertise, Authoritativeness, Trust).

Anderson told the audience this would be the last year at an SEO conference where everyone didnt have AI agents working 24 hours for them. This was the first SMX keynote focused on Content Effort.

📋 Overview of Presentation Structure.

The presentation spans 60 minutes and is divided into three core technical parts plus a final takeaway section:

Part 1 (35 mins): The New Canon of SEO Truth – Decoding 14,014 attributes.
Part 2 (12 mins): Evidence-Based Roadmap – Contextual SEO & the Agentic Web.
Part 3 (8 mins): Case Study – Testing the framework with multi-agent automation.
Close (5 mins): Key Takeaways – From conjecture to certainty.

🔒 Part 1: The New Canon of SEO Truth (Trust & Authority).

This section unpacks the mechanics behind how Google translates user activity, technical markers, and content layout into ranking scores.

1. Core Paradigm of E-E-A-T.

E-E-A-T is not a single factor; it is a framework.
The Goal: E-E-A-T acts as Google’s codified doctrine.
The System: Q* (Q-Star) is the actual machine system that measures it.
The Score: site_quality is the numeric value assigned to your domain.

2. SegIndexer (Tiered Indexing).
Before a site can rank, it is processed through SegIndexer, which places it into one of three structural tiers. This is referred to as the "algorithmic glass ceiling":

Base Tier: High-quality sites that are eligible to compete in main search results.
Zeppelins Tier: Mid-tier content that receives heavily limited visibility.
Landfills Tier: Low-quality content that is effectively filtered out before ranking calculations ever begin.
Ranking Vector: Position within this system is determined via scaledSelectionTierRank.

3. User Engagement: The CRAPS System & Navboost.
Despite historical public denials regarding click data (e.g., Gary Ilyes in 2019 calling it "crap"), DOJ trial evidence confirmed that Chrome data directly influences rankings.

CRAPS: A System mentioned in the API leaks.
Data Window: Tracks a 13-month window of aggregated user click history.
Tracked Attributes: Specifically evaluates GoodClicks, BadClicks, lastLongestClicks, and unsquashedClicks.
Popularity Signal: chromeInTotal parses Chrome browser data to generate a site Popularity (P*) signal.

4. The Title & Header Mechanics ("Goldmine").
Titles are treated like an election by user data:

goldmineNavboostFactor: Bad titles lead to poor click choices, resulting in systematic demotions.
goldmineBodyFactor: Your title tag must explicitly align with the actual copy in the body.
goldmineUrlMatchFactor: The title must semantically align with the URL structure.
goldmineHeaderIsH1: Gives specialized weight calculations directly to the H1 tag.

5. The Clutter Tax & Demotions.
A poor page experience can harm an entire domain via "smeared signals":

clutterScore: A negative user-experience metric scaled and distributed across the whole domain using isSmearedSignal.
violatesMobileInterstitialPolicy: Acts as a boolean (true/false) "kill switch" for mobile layouts.
adsDensityInterstitialViolationStrength: A severity slider scoring intrusive elements on a scale from 0 to 1000.
pandaDemotion: Generates a site-wide penalty referred to as "algorithmic debt".

6. Trust Checklist Matrix.
Google uses explicit signals to cross-reference site legitimacy:

Spam & Deception: Tracked via scamness, spamrank, and GibberishScore.
Sandbox Environments: Checked via hostAge and documentHistory (aimed at filtering "sandbox fresh spam").
Technical Trust: Audited by flags for badSslCertificate and forwardingdup (requires clean canonical paths).
Local Verification: Looks for physical verification via brickAndMortarStrength.
Content Integrity: Evaluated using ContentChecksum96 and pandaDemotion to catch duplicate or thin writing.

🎨 Part 2: The Evidence-Based Roadmap (Experience & Expertise).

E-E-A-T is not a static checklist that can be "sprinkled" onto web pages; it is entirely contextual.

1. Content Effort Metrics.
Google evaluates the sheer human effort behind content generation to combat low-tier AI writing:

contentEffort: Measures replication difficulty using LLM-based detection models.
OriginalContentScore: Distinguishes whether your site is the primary source or merely a summarizer ("the origin or the echo").
original_media_score: Measures if images are unique photographs or generic stock photography.
BlockBERT: Evaluates semantic depth and conceptual density.

2. Graffiti vs. Authority.

Self-declared claims on your own site are treated by Google as "graffiti".
True authority occurs when external, authoritative domains back those claims with links.
Google uses isAuthor and authorObfuscatedGaiaStr to build a machine-readable experience profile for individual authors across the web.

3. Freshness and Quality Trajectories.

lastSignificantUpdate: Google differentiates major topic revisions from minor textual edits.
VersionedFloatSignal: Tracks a site's overall quality trajectory over time to see if the content baseline is improving or worsening.
mobileCwv & desktopCwv: Core Web Vitals are calculated and stored directly on a per-URL basis.
productReviewPUhqPage: Specifically rewards Ultra High Quality product review pages.

4. The Context Stack & Topical Authority.
Topical authority is mathematically quantified through The Context Stack:

Query Class: Know / Do / Website / Visit.
Risk Profile: YMYL (Your Money Your Life) / Soft-YMYL / Non-YMYL.
Expectation Baseline: What a reasonable user expects.
Entity Role: Brand / Publisher / Retailer / Individual.
Proportional E-E-A-T: Minimum → Competitive → Dominant.

site2vecEmbeddingEncoded: Encodes your entire site as a multi-dimensional vector in semantic space.
siteFocusScore / siteRadius: Measures how far individual pages drift from your core niche topic. Every off-topic page increases the radius and dilutes your score.
Signal Coherence (The ABCs of Topicality): Requires Anchors (sourceOnsiteAnchor), Body (goldmineBodyFactor), and Clicks (goldmineNavboostFactor) to align flawlessly.
Stricter Evaluations: YMYL niches are filtered using chard (initial content predictor) along with specialized ymylHealthScore and ymylNewsScore components.

🤖 Part 3: Case Study (Multi-Agent Automation).

This section demonstrates how a complex multi-agent AI system ("The Marketing Cyborg Technique") was built over 4 weeks by 1 developer to automate SEO audits at an unprecedented scale.

1. 6-Layer Grounding Architecture.
To prevent LLM hallucination during automated audits, the case study highlights a 6-layer structure:

Grounding Directive: Agents can only reference explicitly crawled URLs.
Crawled HTML: Injects raw HTML rather than summaries.
Performance Data: Pulls real Google Search Console (GSC) + Google Analytics 4 (GA4) metrics.
Live Search: Looks up and cites active competitive intelligence.
Expert Knowledge: Connects directly to domain-specific knowledge bases.
Deep Research: Implements upgraded reasoning systems backed by search grounding.

2. The Multi-Agent Org Chart.
The case study maps out an automated enterprise team structure powered by specialized digital agents across multiple levels of expertise.

🔄 The Ranking Pipeline.
The final takeaway contrasts the speed of Google's dual evaluation pipelines:

Relevance Systems (Fast): Uses Mustang for initial scoring, T* (Topicality via ABC signals), and P* (Popularity via Chrome data). These systems discover and index content changes within hours.
Quality Systems (Slow): Controlled by Q* (trustworthiness ceilings), Navboost (13-month click reranking loops), and Twiddlers (the final algorithmic filtering layer). These systems evaluate structural entity quality and can catch manipulative tactics over months.

📍 Venue, Dates & Scale
The 15th annual SMX Paris 2026 took place as a live conference on Monday, 9 March and Tuesday, 10 March 2026. https://www.hobo-web.co.uk/smx-paris-2026/

The Venue: Hosted entirely at the Espace Saint Martin in Paris.
The Crowd: Gathered over 1,000 attendees consisting of SEO managers, digital marketing directors, acquisition specialists, and agency executives.
The Experts: Featured more than 50 international and French speakers delivering over 20 focused sessions. [1], [2]

🔀 Event Format & Core Tracks
The event ran two parallel tracks across its two-day agenda to split technical execution from advertising strategy: [1], https://www.hobo-web.co.uk/smx-paris-2026

The SEO Track: Focused heavily on organic visibility in the AI era, Generative Engine Optimization (GEO), navigating massive engine leaks, entity structure, and algorithmic data mapping. [1]
The Paid Media (SEA) & Social Track: Covered advanced Google Ads automated tactics, lead generation, Meta Ads, multi-channel search (Pinterest, TikTok), and conversion rate ROI. [1]

🏢 Corporate & Industry Representation
The attendee and speaker directory represented a diverse split between major digital platforms, media groups, and consumer brands: [1]

Search, Tech & Software Platforms: Teams from Microsoft, Yoast, Semrush, Channable, and Uberall.
Enterprise, Retail & Transport: Representatives from Air France, BUT, Salomon, Bricoman, Legrand, Accor, and Aroma-Zone.
Media & Publishing: Digital acquisition leads from France Télévisions, BDM, Le Figaro, Radio France, and Les Echos Solutions.
Insurance, Health & Finance: Strategic professionals from Harmonie Mutuelle, Foyer Assurances, MAIF, MAAF Assurances, Matmut, Sofinco, and Meilleurtaux. [1], [2]

🏆 Key Event Components

The SEMY Awards: Held at the conclusion of the first day, the 9th annual SEMY Awards Paris celebrated the finest tactical search marketing campaigns across Europe. The primary competitive battles narrowed down to top-tier execution in SEO, SEA, and Innovation.
Independent Curriculum: The programming maintained strict vendor-neutrality. Sales pitches and sponsor-driven content were filtered out of the presentation roster to guarantee practical, data-first strategies.

🔬 Core Algorithmic Realities Disclosed
The overarching narrative across the sessions defined a shifting relationship with traditional search tracking:

The Reality of Click Data: Unpacking historical API data validated that search engines rely heavily on multi-month windows of aggregated user interaction data via browser loops (such as prolonged site stays or immediate bounces) to adjust visibility dynamically.
The "Cost of Replication": Linguistic and search platform experts detailed how advanced modern retrieval systems mathematically calculate the effort it takes to create copy. Low-tier, easily replicated text is algorithmically deprioritised in favour of unique research, first-hand data, and original media.
Topical Radius & Embedding Constraints: True authority is increasingly determined by site focus vectors. Straying too far from a brand's verified semantic core dilutes its core thematic tier rank, proving that modern strategy requires narrowing expertise rather than broadening thin keyword coverage.

Shaun Anderson AKA Hobo - https://www.hobo-web.co.uk/shaun-anderson/

This is Agency Website Quality Evaluator - https://thisisagency.ai/the-website-quality-rater-emulator

https://www.linkedin.com/in/shaun-anderson-hobo/

https://x.com/Hobo_Web

Transcript

1. DECODING THE GOOGLE LEAK What It Means for SEO in 2026 Shaun Anderson • SMX Paris • March 9th 2026
2. PART 1 35 min The New Canon of SEO Truth - decoding 14,014 attributes PART 2 12 min Evidence-based roadmap - Contextual SEO & the Agentic Web PART 3 8 min Case study - testing the framework with multi-agent automation CLOSE 5 min Key takeaways - from conjecture to certainty
3. PART 1: THE NEW CANON OF SEO TRUTH Decoding the 14,014 leaked attributes
5. SECTION 1: TRUST The Gate
6. TRUST E-E-A-T Is Not a Ranking Factor. It Is a Ranking Reality. Layer What It Is E-E-A-T The goal — Google’s doctrine codified. Q* - Q-Star The system — the machine that measures it. site_quality The score — the number your site gets. “E-E-A-T is the goal. Q* is the system. Site quality is the score.”
7. TRUST SegIndexer - You’re Tiered Before You Rank Tier What It Means Base High-quality - eligible to compete. Zeppelins Mid-tier - limited visibility. Landfills Low-quality - effectively filtered out before ranking begins. scaledSelectionTierRank determines your position. “A new, low-authority blog might be filtered out before its content is ever deeply evaluated. This is the algorithmic glass ceiling.”
8. TRUST Dwell Time & CTR “Dwell time, CTR, whatever Fishkin’s new theory is, those are generally made up CRAP. Search is much more simple than people think.” Gary Ilyes, Google - 2019 Reddit Thread
9. TRUST The CRAPS System & Navboost “Google doesn’t rank pages. Users rank pages. Google just counts.” • 13-month window of aggregated click data • GoodClicks, BadClicks, lastLongestClicks, unsquashedClicks • CRAPS = Click and Results Prediction System • chromeInTotal: Chrome data → Popularity (P*) signal • DOJ trial confirmed: Chrome browser data is a direct ranking input
10. TRUST Goldmine — The Title Weapon • goldmineNavboostFactor: bad title → poor clicks → demotion • goldmineBodyFactor: title must align with content • goldmineUrlMatchFactor: title must align with URL • goldmineHeaderIsH1: special weight for the H1 tag “Your title isn’t chosen. It’s elected. The voters are your anchors, your content, and your click data.”
11. TRUST The Clutter Tax “One bad pop-up can, theoretically tank your entire domain.” • clutterScore: scaled, smeared across entire domain (isSmearedSignal) • violatesMobileInterstitialPolicy: boolean kill switch • adsDensityInterstitialViolationStrength: 0–1000 severity • pandaDemotion: site-wide penalty = ‘algorithmic debt’
12. TRUST Trust Checklist Task Signal One-Liner Spam & Deception scamness, spamrank, GibberishScore Are you clean? Sandbox hostAge, documentHistory Leak: ‘sandbox fresh spam’ Technical Trust badSslCertificate, forwardingdup HTTPS + clean canonicals Local Verification brickAndMortarStrength Your building is a signal Content Integrity ContentChecksum96, pandaDemotion No duplicates, no thin content
13. SECTION 2: EXPERIENCE
14. EXPERIENCE Content Effort - The LLM-Driven Engine Signal What It Measures Why It Matters contentEffort Replication difficulty (LLM-based) Can a competitor reproduce this trivially? OriginalContentScore Source vs. summariser Are you the origin or the echo? original_media_score Original vs. stock images Did you actually photograph this? BlockBERT Semantic depth Does your content have conceptual density? “Google can now measure — with a score — whether you actually tried. This is the most important shift for content strategy in 2026.”
15. EXPERIENCE Graffiti vs Authority “What you SAY you are on your own site is graffiti to Google.” • isAuthor, authorObfuscatedGaiaStr: entity-level author tracking. • Google inherently distrusts user input - ‘no better than graffiti’. • Claims are worthless until repeated on authoritative domains. • By tracking authors, Google builds a machine-readable experience profile. “The real E-E-A-T signal is when someone ELSE says you’re an expert - and links to you.”
16. EXPERIENCE Freshness, CWV & Media Signals • lastSignificantUpdate: major revision ≠ minor edit (Google knows the difference) • VersionedFloatSignal: your quality trajectory over time • mobileCwv, desktopCwv: stored per-URL, used directly • productReviewPUhqPage: Ultra High Quality product reviews rewarded “Google tracks your quality trajectory — are you getting better or worse?”
17. SECTION 3: EXPERTISE The Depth
18. EXPERTISE The Site Focus Signal - Topical Authority Quantified • site2vecEmbeddingEncoded: your entire site as a vector in semantic space • siteFocusScore: how niche you are — quantified • siteRadius: how far you drift from your core topic “Google takes every page, averages the embeddings, and measures the radius. The tighter your radius, the higher your focus score.” Strategy: go deeper, not wider. Every off-topic page dilutes your focus.
19. EXPERTISE Signal Coherence - The ABC of Topicality Signal Source Role A — Anchors sourceOnsiteAnchor, sourceOffdomainAnchor What others say you’re about B — Body goldmineBodyFactor What your content says C — Clicks goldmineNavboostFactor What users vote for “When A, B, and C align — that’s signal coherence.”
20. EXPERTISE QBST, Entities & YMYL • QBST: Query-Based Salient Terms - concepts Google expects on a good page • webrefEntities: Knowledge Graph annotations • asteroidBeltIntents: granular intent classification • onsiteProminence: simulated user traffic flow through internal links • chard: initial YMYL content predictor • ymylHealthScore, ymylNewsScore: specialised stricter evaluation
21. SECTION 4: AUTHORITATIVENESS The Reputation
22. AUTHORITATIVENESS Authoritativeness Is Not a Feeling. It’s a Floating-Point Number. • siteAuthority: confirmed in the leak — it’s real • authorityPromotion boost vs unauthoritativeScore penalty • predictedDefaultNsr: a VersionedFloatSignal — your trajectory over time • PageRank-NearestSeeds: modern PageRank evolution • queriesForWhichOfficial: the queries you “own” • isLargeChain, siteSiblings: entity scale & brand signals “Algorithmic momentum: a site with a long, consistent history of quality becomes harder to dislodge.”
23. E-E-A-T Decoded E-E-A-T Primary Machine-Readable Proxies Strategic Imperative Experience contentEffort, originalContentScore, isAuthor, lastSignificantUpdate High-effort, attributable content with original media Expertise siteFocusScore, siteRadius, site2vec, QBST, EntityAnnotations Topically focused sites, clearly defined entities Authoritativeness siteAuthority, PageRank, authorityPromotion, queriesForWhichOfficial Links from trusted sources, build author entities Trust GoodClicks/BadClicks, clutterScore, scamness, pandaDemotion Clean UX, verify identity, optimise for satisfaction Table 3.1 from E-E-A-T Decoded — hobo-web.co.uk
24. The Ranking Pipeline Relevance Systems (Fast) Quality Systems (Slow) Mustang → initial scoring Q* → trustworthiness ceiling T* (Topicality): ABC signals NavBoost → 13-month click reranking P* (Popularity): Chrome data Twiddlers → editorial final layer Can find you in hours Can catch you in months “White hat SEO is not a philosophy. It’s a survival strategy.”
25. SEO Audits in 2026 “No human being can audit all of this. Not manually. Not at scale.” So what’s the roadmap?
26. PART 2: THE EVIDENCE-BASED ROADMAP FOR 2026 Contextual SEO & the Agentic Web
27. E-E-A-T Is Not a Checklist. It’s Contextual. “You can’t sprinkle some experiences on your web pages. That doesn’t make any sense.” — John Mueller, Google
28. The Context Stack Know / Do / Website / Visit 1. Query Class YMYL / Soft-YMYL / Non-YMYL 2. Risk Profile 3. Expectation Baseline What a reasonable user expects Brand / Publisher / Retailer / Individual 4. Entity Role 5. Proportional E-E-A-T Minimum → Competitive → Dominant Quality is judged relative to context, not ideals.
29. Do You Need E-E-A-T? It Depends. Scenario Context What Google Cares About Cat Gifs (Non-YMYL) Fun, no risk Everyday expertise sufficient Medical Advice (YMYL) Health, high risk Qualified authors, citations E-Commerce (YMYL) Taking credit cards scamness, refund policies “Are you offering “expert” advice? Then, yes!”
30. Google Quality Raters Guide
31. Healthy Entity
32. The Hypocrisy Penalty Traps ‘Expert-reviewed’ on About Us but 500 AI posts/day, with zero editing? scamness + unauthoritativeScore “Make sure your website policies reflect your business reality.”
33. Spam
34. Disconnected Entity
35. Correction Algorithms
36. PART 3: CASE STUDY Testing the framework with multi-agent automation
37. The Agentic Web Is Here Event What It Means Chrome auto-browse (Gemini 3) Browsers navigate without you Agentic AI: $4.5B → $76.8B by 2034 The next platform shift MCP + A2A Protocol Agent-to-agent standards shipping “We are not approaching the Agentic Web. We are IN the Agentic Web.”
38. CASE STUDY Testing the Framework 4 weeks • 1 developer • The Marketing Cyborg Technique • Multi-agent orchestration (cf. LangChain, CrewAI, AutoGPT) • Agents accumulate domain expertise over time • DAG pipeline: audit → compliance → legal → human • 6-layer grounding to prevent hallucination
39. CASE STUDY Reducing hallucination in multi-agent systems 6-Layer Grounding Architecture Layer What It Does 1. Grounding Directive Agents can only reference crawled URLs 2. Crawled HTML Actual HTML injected — not summaries 3. Performance Data Real GSC + GA4 data 4. Live Search Cited competitive intelligence 5. Expert Knowledge Domain-specific knowledge bases 6. Deep Research Upgraded reasoning + search grounding An approach any team building agentic systems should consider.
41. CASE STUDY Running A Full Quality Audit Autonomously Dial Leak Signal What’s Happening Content Effort contentEffort LLM replication difficulty Site Focus siteFocusScore Topical authority measured Legitimacy scamness, spamrank Spam classifiers EEAT isAuthor, siteAuthority Authorship, brand signals Q* Composite ALL The master quality score Contextual SEO in practice — YMYL agent applies stricter evaluation.
43. THE NEW CANON OF SEO TRUTH
44. The Virtuous Cycle of Quality 1. Build Expertise → high siteFocusScore, deep original content 2. Attracts links → builds siteAuthority (Authoritativeness) 3. Authority earns visibility → users click 4. Good clicks → Trust signals to NavBoost 5. Trust reinforces Authority → cycle compounds A weakness in one area suppresses the others. This is a holistic system, not a checklist.
45. The Evidence-Based Roadmap Action Why Audit your Content Effort Can your content be trivially replicated? Measure your Site Focus How tight is your topical radius? Build temporal trust Quality trajectory matters. Be patient. Apply Contextual SEO Proportional E-E-A-T based on context Prepare for the Agentic Web Agents are coming. Expertise is the moat.
46. The Signals That Matter Most Signal What It Means For You contentEffort Invest in content that’s hard to replicate siteFocusScore Go deeper, not wider siteAuthority Consistent quality compounds over time Q* Your site has a trust score before anyone searches Contextual SEO Proportional standards, not universal checklists Move from conjecture to certainty.
47. White hat SEO is not a philosophy. It’s a survival strategy in 2026. “AI increases supply. Trust controls demand. E-E-A-T is how real websites survive abundance.”
48. Merci beaucoup, Paris @Hobo_Web