Most SEO workflows still start with a keyword map. Pull search volume, group by intent, assign pages. That’s not wrong — but it’s modelling a system that stopped being the full picture around 2015, when Google acquired Metaweb and started taking the Knowledge Graph seriously.
What Google is actually building when it crawls your site isn’t a keyword-to-URL index. It’s an entity graph: a structured representation of real-world things, how they relate to each other, and which sources it trusts to describe them. Your keyword map tells you what words you want to rank for. The entity graph determines whether Google thinks your site is a credible node in the network of meaning around those concepts at all.
That’s a different problem. Different fix.
What an Entity Actually Is (In Google’s Model)
Google’s definition is specific: an entity is a thing or concept that is singular, unique, well-defined, and distinguishable. A person, a place, an organization, a product, a concept. Not a keyword phrase. Not a topic cluster. An identifiable thing that can have attributes, relationships, and authoritative sources.
The Knowledge Graph — the structured database Google has been building since the Metaweb acquisition — stores entities and their relationships. When someone searches “Python,” Google determines: programming language or snake? That disambiguation happens at the entity level, not the keyword level. The same process runs on every query, even when it’s less obvious.
Your site gets evaluated, in part, by which entities it’s associated with and how confidently Google can make those associations. A site Google can clearly resolve as “entity X, with attributes Y and Z, related to entities A and B” is in a fundamentally stronger position than a site with solid keyword coverage but ambiguous entity identity.
Where Keyword Maps Break Down
Concrete example. You run a site covering B2B SaaS pricing strategy. Your keyword map probably includes:
- “SaaS pricing models”
- “usage-based pricing”
- “freemium vs paid”
- “SaaS pricing page examples”
You’ve built pages for all of them. Traffic is mediocre. The problem might not be the pages — it might be that Google hasn’t resolved your site as a credible entity in the SaaS pricing space. You’re not in the graph as an authoritative node on that concept. You’re a site with relevant text.
The keyword map optimizes text-to-query matching. The entity graph optimizes trust and relationship between sources and concepts. Both matter, but fixing keyword coverage when your entity problem is unresolved is like adjusting your antenna when the receiver is broken.
How Google Builds Its Entity Model of Your Site
This is the part most SEO guides skip — but it’s where the real leverage sits.
Google infers entity associations from several signals working together:
Named entity recognition in your content
Google’s NLP systems — testable directly via the Natural Language API — identify entities mentioned in your content and tag them by type: person, organization, location, event, consumer good. Each gets a salience score: roughly, how central is this entity to the document?
If your content consistently mentions a set of entities with high salience, Google builds an association between your site and those entities. The implication: your core concept appearing tangentially across hundreds of pages is worse than it being consistently central and clearly contextualized on fewer, better-framed ones.
Structured data as explicit entity declarations
Schema markup gets discussed mostly as a rich-snippet tool. It’s also an entity declaration mechanism. Marking up your organization with schema.org/Organization and including sameAs pointing to your Wikidata entry, LinkedIn, and Crunchbase is you explicitly telling Google which real-world entity your site represents — and where to find corroborating signals.
We’ve covered this in more depth in the Schema post, but the entity-identity function of structured data is chronically underused relative to the rich-result function. Most sites use schema to chase stars in the SERP. Fewer use it to anchor their entity identity in Google’s graph.
Co-citation and co-occurrence patterns
When trusted sources mention your brand in the same context as certain entities — your company name appearing alongside “programmatic SEO” across multiple industry publications — Google treats that as a signal your entity is related to those concepts. This is why building topical presence off-site (Reddit threads, LinkedIn posts, trade publications, podcast mentions) matters for entity graph positioning, not just referral traffic.
For new sites this is a hard constraint. Even technically solid content on a new domain can be invisible in AI Overviews because the entity hasn’t accumulated enough off-site co-citation to be treated as a trustworthy source on the topic. Google reaches for Reddit or LinkedIn instead — not because those platforms wrote better content, but because the entity graph trusts them more for that concept. The content quality argument doesn’t resolve an entity trust problem.
Internal link anchor text as entity relationship signals
Your internal link anchor text is a map of how you declare relationships between concepts. Generic anchor text (“click here,” “learn more”) sends Google a flat, undifferentiated signal. Specific, consistent anchor text — the same entity described with consistent terminology across your site — reinforces entity-relationship associations that contribute to your topical graph.
This isn’t keyword-stuffing internal links. It’s conceptual consistency: if you’re building authority on “automated contract review,” every internal reference to that concept should use terminology that clusters around that entity, not drift into loosely related synonyms that dilute the signal.
Practical Audit: Where Does Your Site Sit in the Entity Graph?
A few concrete checks worth running:
Run your key pages through Google’s Natural Language API. Check which entities are extracted and at what salience. If your core concept isn’t appearing as a high-salience entity on your most important pages, your content framing is probably the problem — not your keyword targeting.
Check Knowledge Panel status. No Knowledge Panel for your organization, product, or core topic is a proxy for graph ambiguity. It doesn’t directly cause ranking problems, but it’s diagnostic. Getting into Wikidata with accurate, citation-backed information and connecting it via sameAs in your schema is the most direct path to resolving this.
Audit your sameAs coverage. Does your Organization schema point to your Wikidata ID? LinkedIn? Crunchbase? Wikipedia where applicable? Each is a corroborating source that helps Google anchor your entity identity. Missing these is leaving a free signal on the table.
Search for your brand co-cited with your core topic. Google your brand alongside the concept you want to own. If results are thin, sparse, or dominated by sources describing you differently than you’d want — your off-site entity signal is weak, and content on your own domain alone won’t fix it.
One Trade-off Worth Being Honest About
Entity SEO work is slower to produce measurable results than page-level keyword optimization. Under pressure to show ranking movement in 30 days, structured entity work is not the lever to pull first. It’s compounding infrastructure.
But if you’ve been doing technically correct SEO for 12+ months and can’t work out why certain pages won’t move despite good content and decent links — entity ambiguity belongs on the short list of actual causes. Google may simply not trust your site as an authoritative source on that concept yet, and no amount of keyword optimization resolves that.
Start with the NLP API check on your five most important pages. What entities is Google actually extracting? That answer is almost always more diagnostic than another pass at your keyword map — and it usually points at a framing problem you didn’t know you had.