Most of the Patents Your Agency Quotes Aren't Google's

Somewhere in SEO’s collective knowledge, a handful of patent numbers get passed around as proof of how Google works. They appear in conference talks, in agency proposals, in blog posts explaining why you should write content a particular way.

We went to read them properly. Most of them aren’t Google’s.

The one that started it

There’s a patent quoted constantly in semantic SEO advice, usually described as Google’s, and used to justify writing content in clean subject-verb-object statements.

It belongs to a British telemedicine company. It’s a medical natural-language patent, and the worked example inside it decomposes a sentence about optic neuritis for a diagnostic knowledge base. It has nothing to do with ranking web pages. The word “ranking” doesn’t describe any part of the claimed invention.

The company that owned it collapsed in 2023. The patent now belongs to a healthcare business that acquired the assets.

Why this keeps happening

The explanation is almost funny.

All these patents are hosted on patents.google.com. Which is Google’s patent search engine, a free tool that indexes patent offices worldwide. Being on that domain says precisely nothing about who owns the patent.

Eleven that get miscredited

We checked ownership on every patent we came across. These are commonly attributed to Google and belong to someone else.

• The semantic triple patent, a healthcare company, formerly a British telemedicine firm
• The Hilltop patent, HP, later Valtrus. Google never owned it
• A patent cited as Google’s Quality Score patent, Microsoft
• A click-through-rate prediction patent, Yahoo
• Two fact-extraction patents, Microsoft and a company called Finch Computing
• A widely cited “write clean sentences” patent, the University of Washington
• A ranking patent quoted in several SEO posts, IBM, and unrelated to search
• An AI-search grounding patent, British Telecom
• An AI citation patent, Amazon
• An intent-ranking patent, Citibank

Three more things that don’t exist

There is no E-E-A-T patent. It’s guidance for Google’s human quality raters, a description of what good looks like, not a mechanism.

There is no MUM paper. Google’s announcement links to a different model’s paper and a carbon-cost study.

“Query fan-out” appears in zero patents. It’s Google’s product vocabulary. The filings say “synthetic queries” and “query variants.”

How to check in thirty seconds

Next time anyone quotes a patent number at you:

1. Search the number on patents.google.com
2. Look at Assignee. That’s the owner, not the URL you’re reading it on
3. Check whether the claim actually says what they said it says. A patent’s description can discuss anything, but only the claims are the invention
4. Check whether it’s expired. Plenty of famous ones are

That last one catches more people than you’d think. A lapsed patent still tells you how a system was built. It just isn’t a live claim, and describing it as one is misleading.

Does any of it matter?

The advice isn’t always wrong. Writing clear factual sentences is still sensible, whoever owns the patent.

But the reason matters. If your agency’s justification collapses the moment a client checks a patent number, everything else they’ve told you becomes questionable at the same moment.

And the correct citations exist. There’s a Google patent, filed by one of the founders, that genuinely does describe extracting facts from web text.

It just isn’t the one everyone quotes.