The Google Patents Behind Sharkly (And Why They Matter)

When most people talk about SEO, they talk about best practices — things that seem to work, tips that have been passed around for years, strategies that someone once tested and wrote a blog post about. It's mostly educated guesswork.

Sharkly was built differently. Every recommendation our platform makes — every keyword we flag, every content structure we suggest, every priority we assign — is grounded in primary source research. Specifically: nine verified Google patents that describe the actual mechanics of how Google's search ranking system works.

We're not going to give you the full playbook here. But we want you to understand what's under the hood — and why it means Sharkly's recommendations aren't opinions. They're engineering.

Why patents? Doesn't Google lie about how their algorithm works?

Google is famously cagey about its algorithm. Public statements are vague, often contradicted by independent testing, and sometimes outright misleading. But patents are different.

A patent is a legal document. It describes a system that has been built and filed with the United States Patent and Trademark Office. Google engineers put their names on these filings. The systems they describe are real, assigned to Google Inc. or Google LLC, and in many cases have been independently confirmed.

Most critically: the 2023 U.S. Department of Justice antitrust trial against Google produced sworn testimony from Google's own engineers confirming that specific patented mechanisms — particularly their behavioural ranking system known internally as Navboost — are active, primary ranking signals. This is not inference. It is court-confirmed.

The 2024 Google Search API leak (2,500 internal documents made public) independently corroborated the same mechanisms. When a patent, sworn testimony, and a leaked internal document all describe the same system, that system is real.

The six patent groups that power Sharkly

Our research team has identified nine patents across six functional groups. Each group describes a different layer of Google's ranking system, and each informs a different part of how Sharkly builds your strategy.

1. The Panda Quality System

Three patents work together to form Google's site-wide quality classification system. The core mechanism — documented in US8682892B1, filed by the engineer the system was named after — establishes a group modification factor: a multiplier applied to every page score across your entire domain based on the ratio of genuine inbound links to branded reference queries.

What this means in plain English: if your site has a lot of low-quality pages, Google doesn't just penalise those pages. It penalises every page on your domain simultaneously. Site quality is a domain-level signal, not a page-level one.

A continuation patent (US10055467B1) adds a behavioural layer to this system — repeat visits, deliberate clicks, and average session duration all feed into the same site-wide multiplier. Google is measuring whether real people actually want to be on your site.

2. Navboost — The Behavioural Ranking Signal

US8595225B1 is the patent Google's own VP of Search, Pandu Nayak, described under oath as their "most important" ranking signal. Navboost assigns per-topic popularity scores derived from user navigational patterns — it tracks how people behave on your pages for specific queries and uses that behaviour to adjust your ranking for those queries.

This is why click-through rate, dwell time, and pogo-sticking aren't just UX concerns — they are documented, patent-grounded ranking inputs. And critically, Navboost is query-specific, not domain-global. A page can rank well for one query and poorly for another based entirely on how users behave when they arrive from each.

3. Historical Trust Accumulation

US7346839B2 describes how Google builds a trust score for every document over time. The inputs include: the date Google first discovered a link to your domain, how quickly you've accumulated links, how stable your anchor text profile is, and how often your content is updated. This patent is the formal basis of what SEOs call the "Sandbox" — the observation that new sites struggle to rank for competitive terms regardless of their content quality.

Trust is not instant. It accumulates. The earlier you start building it, the more compounding advantage you gain.

4. Passage Scoring

Two patents — US9940367B1 and US9959315B1 — describe how Google scores individual passages within a page independently of the page as a whole. A page doesn't need to be entirely about a topic to rank for a specific query. Individual sections are extracted and scored against queries on their own merits.

The second patent establishes that your H2 headings don't just signal keywords — they define the context within which each passage below them is interpreted. The heading hierarchy from title to H1 to H2 to paragraph is evaluated as a coherent path. The way you structure your headings directly affects which passages Google can extract and rank.

5. Information Gain Scoring

US20190155948A1 describes a machine learning model that scores how much novel information a document provides relative to everything Google has already seen on the same topic. Content that closely mirrors what's already on the SERP receives a systematically low information gain score. Content with original data, unique perspectives, or genuine expert insight scores higher.

This is almost certainly the mechanism behind Google's Helpful Content updates. Writing content that is better than the competition is not enough — it needs to be genuinely different. Sharkly's content briefs are built around this requirement.

6. The Reasonable Surfer Model

US8117209B1 replaced the classic PageRank formula in 2010. The original model assumed every link on a page passed equal equity. The Reasonable Surfer model weights link equity by the probability that a real person would actually click that link — determined by placement (body text vs. footer vs. navigation), anchor text quality, visual prominence, and topical relevance of the source page.

A footer link from a high-authority domain can pass less equity than a body-text link from a mid-authority site. Where a link is placed is a first-class ranking variable with a documented formula.

What this means for Sharkly's recommendations

Every time Sharkly tells you to target a specific keyword, structure an article a certain way, build your content in clusters, or prioritise one topic over another — that recommendation traces back to one or more of these six patent groups.

When we tell you to write your H2s as clear, answerable questions, that's US9959315B1. When we tell you which keywords your site can realistically compete for based on your current authority, that's US7346839B2 and US8682892B1. When we build your content in clusters around a money page, that's US8117209B1 and US8595225B1 working together.

We're not going to publish our full algorithm. But we will tell you this: the gap between "best practice SEO" and patent-grounded SEO is significant. Most of what passes for SEO advice online has never been tested against primary sources. Sharkly has.

That's the difference.