Episode 8: How to Get Entities and Their Attributes in the Knowledge Graph
Episode 8: How to Get Entities and Their Attributes in the Knowledge Graph
The Knowledge Graph episode that came before this one - Episode 7 - asked how well Google was doing at understanding the world. Episode 8 asked the more practical question: if your brand isn’t in the graph, or isn’t in it accurately, what do you actually do about it?
Jason Barnard had been working on his own Knowledge Graph presence since before the #SEOisAEO series began, using himself as a test subject. His own admission is that some of what he did was opportunistic - leveraging things that were easy to get into the graph rather than following a principled methodology. Episode 8 was the corrective: David Amerland, Sam Underwood, and Andrew Cock-Starkey joined him to lay out the principled approach.
The core insight is that the Knowledge Graph doesn’t build itself from your website. It builds from a network of corroborated signals across sources it trusts. Your website is one input. Wikipedia is a more trusted input. Wikidata is the structured backbone that many AI systems now draw on directly. Your Google Business Profile anchors your local entity. Your About page, structured correctly with schema markup, is the definition you offer the algorithm about who you are and what you do. Press coverage, directory listings, and third-party references are the corroboration that converts a claim into a confirmed fact.
The distinction between an entity and a string matters here. When you type a brand name into Google and it doesn’t trigger a Knowledge Panel, the algorithm is treating that name as an unresolved string - a pattern of characters that might refer to any number of things. When the algorithm has enough corroborated signals to commit to an interpretation, that string becomes a resolved entity. It gets a node in the graph. It gets attributes - properties the algorithm associates with confidence. It gets relationships - connections to other entities that give it context and meaning.
The practical steps the episode covered follow directly from that model: establish the entity on structured platforms (Wikipedia where notable, Wikidata regardless), define it with schema markup on the entity home page, distribute consistent signals across trusted third-party sources, and monitor what the graph has actually learned by checking how Google surfaces the brand in Knowledge Panels and entity-based results.
In 2018, that process was manual and specialist. It’s now the foundation of what Jason formalised as The Kalicube Process™ - and the 10-gate DSCRI-ARGDW pipeline that describes how a brand moves from unknown string to trusted, recommended entity.
The presentation deck used during Episode 8 of the #SEOisAEO series is preserved on SlideShare.
Published by: Semrush. Host: Jason Barnard. Guests: David Amerland, Sam Underwood, Andrew Cock-Starkey. October 23, 2018
