Episode 7: The Knowledge Graph - How Well is Google Doing in ‘Understanding the World’?
By October 2018, I’d spent over a year arguing that SEO was becoming something else entirely, an optimisation discipline aimed not at ranking pages but at feeding answers to machines. The #SEOisAEO series on SEMrush was my attempt to build the evidence base for that argument in public, with some of the smartest people in search joining me every week.
Episode 7 was the pivot point. The first six episodes had tackled the infrastructure question: how do you structure a web page, a schema markup, a piece of content so that an Answer Engine can actually process it? Episode 7 asked the harder question: does Google actually understand the world well enough to be the engine we’re all building toward?
The short answer, then and now, is: better than most people realise, and nowhere near as well as Google wants you to think.
The Knowledge Graph is the hub of every Answer Engine, and Google’s is the largest
The Knowledge Graph isn’t a database in the way most marketers think of databases. It’s a network of entities, people, organisations, places, concepts, connected by typed relationships. George Washington is a US President. A US President is a Person. A Person has a nationality. Each of those connections is a semantic triple: Subject, Predicate, Object. Stack enough triples together, with enough corroboration from enough sources, and the machine doesn’t just store facts - it understands context.
That distinction matters for Answer Engine Optimisation. A search engine matches strings. An answer engine resolves entities, and to resolve an entity it needs a Knowledge Graph deep enough to draw confident conclusions from. Every answer Google provides through its AI systems, every Knowledge Panel it surfaces, every featured snippet it selects, draws from that graph. The Knowledge Graph is the hub, not a feature.
Kristine Schachinger, Bill Sebald, and Paul Woodhouse joined me for this episode, and the conversation kept returning to the same uncomfortable reality: Google’s Knowledge Graph is vast, but it’s patchy. Confident on the entities it knows well, major brands, public figures, well-documented organisations, and genuinely uncertain about the rest. For a mid-size business, a specialist professional, or a brand that hasn’t actively fed the graph, the engine’s understanding is somewhere between thin and wrong.
Google understands what it has been told clearly, consistently, and repeatedly
This is the insight that, for me, defines the entire AEO discipline. Google doesn’t understand the world through some magical intelligence. It understands the world through the quality of the information it has been given. The Knowledge Graph reflects the web, which means it reflects the web’s noise, contradictions, gaps, and ambiguities unless something actively corrects them.
That correction is the job of the brand. Not the SEO. Not the technical team. The brand, which owns the facts about itself and has the only legitimate right to define them.
In 2018, I framed this as the entity home problem: every entity needs a single authoritative page that acts as the machine’s point of reconciliation when sources conflict. Structured data - schema markup - is the language you use to write that page in a way the machine can process without guessing. And third-party corroboration is the mechanism that converts a claim into a confirmed fact.
The Knowledge Graph’s answer to the question of how well Google is understanding the world is really a question about how well the world has made itself understandable. The engine is better than its critics admit and worse than its outputs suggest, and the brands that recognise that asymmetry are the ones that use it deliberately.
What this looked like in 2018, and what it looks like now
The conversation in Episode 7 was happening at a moment when the Knowledge Graph was still a relatively specialist concern, well understood by entity SEO practitioners and largely invisible to the broader marketing community. The mainstream was still arguing about keyword density and link velocity.
I’d been arguing since 2017, first in a whitepaper co-authored with Chee Lo at Trustpilot, then in a piece for Search Engine Watch in February 2018, then through fifteen weeks of weekly webinars, that this was exactly backwards. The race to optimise for answers was already underway. Google was already evaluating entities, not pages. The Knowledge Graph was already the mechanism that determined whether a brand got recommended or ignored.
The industry took another five years to broadly agree. ChatGPT arrived in late 2022, AI Overviews rolled out across Google in 2024, and suddenly every brand was asking how to appear in AI answers. The answer, as the Authoritas Weighted Citability Score study confirmed in late 2025, is the same as the answer in 2018: build a Knowledge Graph presence that’s clear, corroborated, and consistent enough that the engine commits to it.
The presentation deck used during Episode 7 of the #SEOisAEO series is preserved on SlideShare. The Knowledge Graph - How Well is Google Doing in ‘Understanding the World’?
Episode 7 of the SEMrush #SEOisAEO series was one of the earliest public examinations of why that matters and how it works. The full 15-episode series is the context. This episode is the hub.
Published by: Semrush. Host: Jason Barnard. Guests: Kristine Schachinger, Bill Sebald, Paul Woodhouse. October 16, 2018
