SEO vs GEO: stop choosing sides—and add knowledge graph publishing to the stack

A useful correction is circulating in the SEO world: generative engine optimization (GEO) and answer-engine optimization (AEO) are not replacements for SEO—they presuppose it. In a recent essay, Ann Smarty argues that SEO expertise remains foundational for GEO, for reasons that boil down to discoverability, journey coverage, and the way large models and their tools actually gather context.

We agree—and we would add a third layer that is still under-discussed in that debate: knowledge graph publishing. Not as a gimmick, and not instead of keywords or site architecture, but as structured entity infrastructure that makes brands easier to find, disambiguate, and place in the right context when systems reason over graphs and retrieval—not only over a single page of copy.

This post connects Smarty’s framing to why knowledge graph publishing belongs in the same stack, and what value it adds that content and links alone do not guarantee.

What “stop choosing sides” really means

Smarty’s piece pushes back on a narrow story: “SEO is dead; optimize for prompts.” Her counter is straightforward:

• Organic visibility still feeds AI-era discovery, because models and assistants often reach the web (and brands) through mechanisms that look a lot like search and crawl.
• Keyword and intent work still matters, because we do not have reliable, complete prompt logs; query and content-gap thinking remains the best proxy for how people research and buy.
• Content, information architecture, and technical accessibility (including sites that degrade gracefully without heavy client-side dependency) still shape whether a business is understandable—to users and to machines.
• Authority and mentions remain plausible indirect levers even when the direct backlink → LLM effect is unproven: they affect rankings, associations, and trust signals in the broader ecosystem.

That is a coherent picture: GEO changes what you optimize for in reporting (citations, inclusion in synthesized answers, brand presence in AI surfaces), but it does not erase the machinery that gets a URL or a brand name in front of a system in the first place.

Our read is aligned. The risk for agencies is not “doing GEO”; it is treating GEO as a reason to starve top-of-funnel SEO, thin out site structure, or skip entity-level clarity. That trades short-term narrative for long-term fragility.

Where GEO genuinely diverges from classic SEO

Smarty also acknowledges that success metrics differ: reference rates, answer inclusion, and AI visibility are not the same as position and click-through rate alone. That is why monitoring and experimentation matter—as we discuss in the context of evaluation metrics and benchmarks and AI brand visibility tooling.

So the productive synthesis is not “SEO or GEO.” It is SEO for discoverability and journeys, GEO for how often and how accurately you appear in generative answers, and—our addition—knowledge graphs for durable, typed identity and relationships.

Knowledge graph publishing: what it adds to the same stack

knowledge graph publishing means placing or enriching a canonical entity (your client’s organization, practice, or brand) in a graph such as Wikidata: typed links to industry, geography, official site, and related entities, with references where the community expects them.

Why that matters alongside Smarty’s SEO points:

1) It is the structured counterpart to “keyword intent”

Keywords tell you how people phrase problems. A well-modeled entity expresses who the business is, where it operates, and what it is an instance of in machine-readable form. That reduces ambiguity (“which Springfield?”) and aligns with how graph-backed retrieval slices the world—see hub nodes and local business AI visibility.

2) It complements content, it does not replace it

Your pages still need to help humans and crawlers. Graph publishing answers a parallel question: Is there a stable, queryable node for this brand in the public knowledge layer many tools and studies treat as infrastructure? For a deeper take on Wikidata’s role, see why Wikidata is premier knowledge graph infrastructure for GEO.

3) It reinforces information architecture at entity scale

Internal linking shapes how a site is understood. Graph statements shape how a business is attached to cities, countries, sectors, and peers. Both matter; they operate at different scopes.

4) It sits in the “mentions and associations” story

Smarty notes that brand mentions and co-occurrence with known competitors may shape how trustworthy or familiar a business appears in broader signals. A knowledge graph item is not a substitute for PR, but it is a clean, explicit encoding of associations (location, industry, official identity) that downstream systems can reuse without scraping ambiguity from ten conflicting directories.

A practical stack for agencies

A defensible 2026 playbook looks like this:

1. Keep SEO fundamentals: intent-led content, solid IA, crawlability, and earned visibility where strategy warrants it.
2. Add GEO measurement: track how brands show up in target AI surfaces; set expectations that differ from SERP-only reporting.
3. Publish and maintain graph entities: Wikidata publishing for business with correct types, locations, and references—not one-off spammy stubs.

GEMflush exists to make step three operationally feasible at scale (API-backed publishing, quality discipline, and monitoring) so agencies do not have to choose between “classic SEO” and “AI visibility”—they can offer one continuous system: discoverable sites, measurable generative presence, and durable entity data in the graph layer assistants and researchers actually query. For methodology and proof loops, see our methodology; for agency positioning, AI visibility for SEO agencies.

Takeaway

Smarty’s essay is a useful antidote to zero-sum thinking. SEO and GEO are not enemies. Neither replaces the other. Knowledge graph publishing is not listed in every SEO checklist yet—but it is the natural third column: entities and relationships that persist across pages, campaigns, and model releases.

Stop choosing sides. Build one stack: journeys and crawlability, generative metrics, and published graph reality.