Read every Google patent and the 2024 API leak. Distilled into 3 laws and 4 prompts you can run tonight.
Three laws. Four prompts. One page. Everything Google said. None of the opinions. Sourced from Google patent US 8,595,225 NavBoost plus the 2024 Google Search API leak (siteFocusScore, originalContentScore, site-wide quality modifier fields).
Most SEO advice is opinion stacked on opinion. This Cheat Sheet is built only from what Google itself has put on the record: two granted patents and the May 2024 Search API leak that exposed roughly 14,000 internal ranking fields. Three laws cover how Google measures your visitors, your content, and your whole site. Each law comes with the fix to run this week and a Claude prompt that does the heavy lifting.
Before Google cares about your content or your links, it checks what real humans do when they land on your site. Two systems run this.
Sources: Google Patent US 8,595,225 (NavBoost). Named Entity recognition, documented across multiple Google patents and the Knowledge Graph methodology.
Tracks what happens after someone clicks your result on Google. Stay 30 seconds or more without bouncing back to search? Good click. Bounce straight back? Bad click. Your rankings are impacted by this massively.
Think of it like a movie trailer. If thousands of people watch your trailer and then walk out of the cinema without buying a ticket, the distributor stops giving your movie the premium screening slots. Your title and meta description are the trailer. Your first 100 words are the opening scene.
Google tracks your business as a named entity. If your name, address, and phone number conflict across different platforms (your site, Google Maps, review platforms, directory listings, social profiles), Google's confidence in you drops. That uncertainty affects every ranking decision it makes.
It is like applying for a loan with three different versions of your name on three different documents. The bank cannot trust any of them, so it slows the whole application down. Google does the same thing with your business identity.
Two mistakes. (1) Title and meta description over-promise something the first 100 words do not deliver. The user bounces. (2) Business name listed differently across the website footer, Google Maps profile, Yelp listing, BBB, and social profiles. Google cannot reconcile the identity.
The fix this weekA local services client had fantastic content, strong links, good engagement signals. But their business name was listed three different ways across their profiles. Google Maps had one version, their website had another, a directory listing had a third. Once we standardised everything to be word-for-word identical, their local rankings started moving within weeks.
Google doesn't just check if your content is good. It checks whether it belongs in its lane and whether it adds anything new. Three systems run this. All three are in Google's own documentation.
Sources: Google Search API leak (May 2024), the siteFocusScore field and the originalContentScore field (integer 0 to 127). Google Information Gain patent (US 11,816,176).
Scores how tightly every page on your site sits to your core topic. A plumbing site that blogs about cryptocurrency has a low siteFocusScore. A site entirely about commercial plumbing in Melbourne scores high.
Think of it like a Venn diagram. The tighter your pages cluster around one topic, the higher your score. The more scattered, the lower.
A literal field in Google's system that scores how original your content is. Zero to 127. Most content sits at a 4. Original research, proprietary data, and real case studies push it toward 90 plus.
If everyone studying for an exam writes the same essay, the grader cannot tell them apart. The one essay that says something nobody else said is the one that stands out.
Measures whether your page adds anything new to what is already indexed on the web for that query. If everything on your page can already be found on the top 5 results, your page adds zero information gain.
It is like a job interview. If every candidate gives the same answer, nobody gets remembered. The one who says something nobody else said gets the callback. Google works the same way.
Most content strategies do this. Find a keyword. Google the keyword. Read the top 5 results. Write something covering the same ground in the same order with slightly different words. originalContentScore probably 4. Information Gain probably zero.
The fix this weekA B2B SaaS client had published over 80 articles. Almost all covered the same ground as their competitors. Best practices, how to set up, what to do. Classic category content. None of it was ranking after 14 months. We switched their strategy. Every piece had to contain something original. A client survey result, an internal data point, an expert opinion that contradicted conventional wisdom. Within 90 days of publishing the first wave, the category rankings started moving.
Google applies a site-wide quality modifier to high-authority domains. It lifts every page on the domain simultaneously. Not just the pages that earned it. Every page. This is why some sites seem to rank no matter what.
Sources: Google Search API leak (May 2024), the site-wide quality modifier field, plus Q* (quality) and siteAuthority scoring documented in the same leak.
Applies a domain-level lift to every page on the site. Compounds from every quality decision (legitimate backlinks, original research, brand mentions, structured data, engagement signals) stacked over time into a signal that moves everything at once.
Think of it like two buildings. Building A has a multiplier applied because of where it sits. Even its average floors sit higher than Building B's best floors. The location lifted everything together. Domain authority is the location. Every page benefits.
Until you understand this law, you will not be able to explain why a competitor with a worse website still outranks you. Thinner pages, fewer links, half the effort. Sitting above you on Google. The modifier is doing the heavy lifting for them.
The most common mistakeOptimising single pages in isolation. Treating SEO as a per-page exercise instead of a domain-wide compounding signal. Investing in cheap link packages that build no domain authority. Ignoring digital PR because it does not have a direct page-level CTA.
The fix this weekJobAdder is a global recruitment SaaS. Saturated markets across Australia and the UK. Cost-per-lead surging. We built domain-level authority through original research using their own platform data: State of the Market reports. Then 58 media placements across News.com.au, HR Daily, and others. 38 high-authority backlinks pointing deep into product pages, not just the homepage. Result: 361 percent organic traffic increase in Australia. Ranking for 'ATS Software' went from position 66 to number one. TOFU conversions went from 7 to 174.
Copy any prompt below, paste it into Claude (or ChatGPT or Gemini), and run it on your own site tonight. Each one operationalises one of the laws.
You are a local SEO consistency auditor. I will paste the business name, address, and phone number as they appear on my website footer, my Google Business Profile, and 3 of my directory listings. Your job: (1) identify every variant of the business name, address, or phone format, (2) flag every discrepancy that would prevent Google from confirming this is the same business entity, (3) recommend the single canonical version that should appear everywhere, (4) give me a copy-paste-ready standardised NAP block I can replace across all platforms. Be ruthless. Even a missing 'Ltd.' or 'Pty Ltd' or a difference in suburb capitalisation counts as a discrepancy. Paste your NAPs below. Format each one as 'Source: [website / GBP / Yelp / etc]' then the NAP on the next line.
You are an SEO content auditor checking title-to-page alignment. I will paste my page title, meta description, and the first 200 words of my page content. Your job: (1) state in one sentence what my title promises the user, (2) state what my first 100 words actually deliver, (3) score the alignment 0 to 10, (4) identify the specific gap that would cause a real user to bounce back to search, (5) rewrite my first 100 words to land the promised value inside the fold. Be brutal. If the rewrite needs to drop a paragraph, drop it. Paste your title, meta, and first 200 words below.
You are an SEO content strategist auditing my site for siteFocusScore (topical focus) and Information Gain. I will paste a list of my last 10 to 20 published page URLs with one-line summaries OR paste the actual page bodies. Your job: (1) for each piece, score topical focus against my core niche (high / medium / low / off-topic) and explain why, (2) flag the pieces to prune or redirect (high topical drift, no traffic), (3) for each remaining piece, identify whether it adds Information Gain (original data, proprietary insight, contrarian view) or just rehashes the top 5 SERP results, (4) give me a 30-day content prioritisation: prune list, refresh list, original content backlog. Be honest about pieces that are not pulling their weight. Paste your business niche (one sentence), then your page list below.
You are a digital PR and domain authority strategist. Tell me your business name, primary service or product, target market, and 2 to 3 internal data points you could turn into original research (transaction data, survey data, anonymised client outcomes, internal benchmarks). Your job: (1) generate 5 original research or data study ideas using my internal data, (2) identify the 3 highest-leverage digital PR angles for my niche, (3) describe the linkable asset structure that will earn citations across multiple publications, (4) give me a 90-day execution roadmap with target publications named. Prioritise compounding signals over one-off placements. Paste your business details below.
Fastest: copy a prompt above. Paste it into Claude, ChatGPT, or Gemini, add your own details where the prompt asks, and read the output. Start with the law that matches your biggest problem.
Printable: download the PDF. The 3 Laws on page 1, the 4 prompts on page 2. Stick it next to your monitor or hand it to whoever runs your marketing.
Best: install the Claude Skill. Download the .md file below, drop it into ~/.claude/skills/ (Mac) or %USERPROFILE%\.claude\skills\ (Windows), restart Claude Desktop, and ask Claude to "audit my site against the 3 ranking laws". All four prompts become named modes.
The 3 Laws with sources and this-week fixes on page 1. The 4 copy-paste prompts on page 2. Print it, share it, pin it.
Download the PDFAn installable .md skill file that turns Claude into a 3 Laws auditor with all four prompts as named modes. Same format as every Hawk Academy Claude SEO skill.
Download the Claude SkillEvery claim on this page traces to one of these. No opinions smuggled in.
| Reference | Type | Detail |
|---|---|---|
| Google Patent US 8,595,225 (NavBoost) | Patent | The NavBoost system. Tracks click behaviour on Google search results and uses it as a ranking signal. The 30-second dwell threshold, the good-click vs bad-click distinction. Primary source for Law 1 System A. |
| Google Search API leak (May 2024) | Leaked internal documentation | Roughly 14,000 internal API fields exposed via a GitHub Cloud Client Library publication. Includes siteFocusScore, originalContentScore (0 to 127), Q* (quality) and siteAuthority scoring, site-wide quality modifier. Primary source for Law 2 Systems A, B and Law 3 System A. |
| Google Information Gain Patent (US 11,816,176) | Patent | Describes how a page is scored on whether it adds new information to the corpus already indexed for a given query. Primary source for Law 2 System C. |
| Google Knowledge Graph and Named Entity documentation | Google official | Published methodology around how Google identifies and tracks businesses as named entities. Backs Law 1 System B. |
| StudioHawk JobAdder case study | Internal client work | JobAdder global recruitment SaaS. State of the Market reports built using JobAdder's own platform data, 58 media placements across News.com.au, HR Daily, and others, 38 high-authority backlinks. Result: 361 percent organic traffic increase in Australia, ATS Software ranking from 66 to 1, TOFU conversions 7 to 174. |
| StudioHawk B2B SaaS Information Gain case | Internal client work | Anonymised. B2B SaaS, 80 articles published over 14 months, zero meaningful rankings. Switched strategy to require Information Gain on every piece. Rankings moved within 90 days. |
| StudioHawk siteFocusScore prune case | Internal client work | Anonymised. Client pruned 47 off-topic pages. Traffic recovered 34 percent within 60 days. |
| StudioHawk NAP consistency case | Internal client work | Anonymised local services client. NAP listed three different ways across profiles. Standardised, local rankings moved within weeks. |
| Term | Plain English definition |
|---|---|
| NavBoost | A Google ranking system that tracks what people actually do on your search result. If they click and stay, that is a good click. If they bounce back to search within seconds, that is a bad click. Your rankings move accordingly. |
| Named Entity | How Google identifies your business as a distinct, real-world thing in its system. If your name, address, and phone are consistent across the web, Google has high confidence in your entity. If they conflict, confidence drops and so do your rankings. |
| NAP | Name, Address, Phone. The three pieces of business contact information Google uses to verify a business is the same entity across platforms. |
| siteFocusScore | A field in Google's internal API (revealed in the May 2024 leak) that scores how tightly every page on your site sits around your core topic. Tighter focus, higher score. |
| originalContentScore | An integer field 0 to 127 in Google's internal API (May 2024 leak) that scores how original your content is. Original research sits high. Rehashed content sits low. |
| Information Gain | From Google patent US 11,816,176. Measures whether your page adds new information that is not already indexed for the query. If everything on your page can be found in the top 5 existing results, your information gain is zero. |
| Site-wide quality modifier | From the May 2024 API leak. A domain-level lift applied to high-authority sites. Every page on the domain gets the lift, not just the ones that directly earned it. This is why established brands outrank smaller sites with objectively better individual pages. |
| Q* and siteAuthority | Two of the quality scoring signals revealed in the 2024 API leak. Both contribute to the site-wide quality modifier. |
| Digital PR | Earned media coverage in publications, secured through original research, expert commentary, or newsworthy data. Builds domain authority through high-quality backlinks. |
| Linkable asset | A piece of content (original research, state of the market report, calculator, proprietary data study) that other publications cite because they cannot easily find the equivalent data elsewhere. |
| Cheat Sheet | This page. The page you are reading. Free download, no email required. |
Want the full library? There are 25 more free Claude SEO skills where this came from.
