Problem

Search is moving deeper into retrieval, summarization and query expansion. Users increasingly ask broad questions, compare alternatives and expect an answer shaped from multiple sources. Many websites respond by producing more pages, more keyword variants and more generic AI-written summaries. That is the wrong direction.

The durable problem is not how to trick an answer engine. It is how to make a public corpus worth retrieving, easy to crawl, structurally understandable and editorially distinct. AI search rewards the same foundations that matter to humans: useful content, clear ownership, visible sources, stable pages and meaningful internal connections.

Introduction

AI search does not eliminate SEO. It changes the operational center of gravity. A search system that uses retrieval-augmented generation still needs crawlable web pages, ranked sources and identifiable evidence. If a page is not indexable, if it has no visible value, or if its claims cannot be tied to an entity and source, it is weak material for both search engines and AI systems.

Google’s guidance for generative AI search emphasizes useful, non-commodity content, technical clarity and crawlable pages. Structured data remains useful because it gives explicit clues about what a page describes. But structured data should not become invisible decoration. It should compress the same truth that readers can see.

Electronic Artefacts is already well positioned for this shift because the site is not only a portfolio. It is a knowledge graph with publications, concepts, technologies, programs, projects, sources and relation statements. The SEO opportunity is to make that architecture legible as a coherent body of expertise.

Architecture

Knowledge graph SEO combines editorial content, entity identity, internal linking, structured data, source metadata, canonical URLs, sitemap coverage, search documents and graph relations. The page is the human surface. The entity record is the durable semantic object. JSON-LD, identifier routes and relation exports make the same object machine-readable.

Search is becoming retrieval

Generative search uses retrieval patterns to ground answers in indexed pages. That means the page still matters, but it is often consumed as a source fragment, not only as a ten-blue-links destination.

This changes the kind of page that performs well. A shallow article optimized around one keyword variant may be less useful than a focused page that answers a real problem, defines terms, links to adjacent concepts and includes enough context to be quoted or cited.

For Electronic Artefacts, the strongest pages are not generic explainers. They connect a topic to a system perspective: graph runtimes, archive provenance, local AI, creative tooling, media credentials or cultural infrastructure. That point of view is difficult to reproduce from commodity search results.

Non-commodity expertise

Non-commodity content contains something specific: a position, a workflow, an implementation model, a failure mode, a comparison, a system diagram, a critique or a domain example. It does not merely restate public definitions.

An article about structured data can say that JSON-LD exists. A stronger article explains why structured data must match visible content, how identifier routes avoid confusing a page with the thing described, and how relation validation prevents a knowledge hub from decaying into broken links.

This is the editorial standard Electronic Artefacts should keep. The site should answer questions that a serious reader or AI retrieval system would ask: what is the entity, why does it matter, what is the system boundary, what are the risks, what are the sources, and where should the reader go next?

Structured data

Structured data helps search systems understand the page. Schema.org Article, CreativeWork, Organization, DefinedTerm and related types can describe title, author, publisher, dates, description, subject, citation and image.

The key rule is correspondence. Do not add structured data that the visible page does not support. If the JSON-LD says the article is about provenance, the page should visibly discuss provenance and link to the concept. If the structured data names a publisher, the page should have a clear publishing context.

JSON-LD is useful because it lets static pages carry structured meaning without changing the visible design. But it should be generated from the canonical content model, not hand-written separately for each page.

Entity-first SEO

Keyword-first SEO asks what phrase a page should rank for. Entity-first SEO asks what thing the page describes and how that thing relates to the rest of the corpus.

An entity-first article has a stable ID, canonical slug, subjects, sources, related concepts and relation statements. A query about “AI search structured data” may retrieve an article, but the system can also see connections to JSON-LD, metadata, knowledge graphs and RAG.

This is especially important for a studio with multiple identities: creative technology studio, cultural infrastructure lab, archive system, AI research practice and software design partner. Entity structure lets those identities reinforce each other instead of competing as disconnected landing pages.

Query fan-out

AI search may expand a user’s query into related subqueries. A page that only targets one exact phrase can miss this broader retrieval behavior. A page that covers the problem space with clear sections has more useful entry points.

For example, a user asking whether SEO still matters for AI search may also need structured data, crawlability, helpful content, RAG, citations and entity authority. One strong article can cover those dimensions without becoming a loose list of keywords.

Headings matter because they expose the shape of the answer. Sections such as “Structured data”, “Entity-first SEO” and “Implementation” are not only helpful to readers. They are retrieval handles.

Content clusters

A knowledge hub should not create endless near-duplicate articles. It should build clusters with distinct roles. A concept page defines. A technology page documents a protocol. A publication explains an argument or workflow. A collection groups a research wave.

For AI search, clusters help machines and readers understand topical authority. One article about knowledge graph SEO can link to Linked Data, JSON-LD, RAG, entity identity and knowledge graphs. Those pages then link outward to related publications and project contexts.

This is stronger than a blog archive sorted only by date. It gives every article a place in a semantic network.

Crawlable technical base

The technical baseline remains simple: pages must be crawlable, indexable, canonical, fast enough, internally linked and available as server-rendered or statically generated HTML. Client-side enhancements should not be the only place where important content appears.

Electronic Artefacts already follows this direction by generating HTML, JSON-LD, sitemaps, search documents, identifier routes and graph exports from typed content. That turns SEO from a manual afterthought into a build output.

The remaining work is editorial depth. Search systems can reach the pages. The pages then need to deserve retrieval.

Metrics

AI search measurement is less tidy than classic ranking checks. Useful signals include indexed pages, crawl errors, search queries, click-through rate, impressions, pages with structured data, internal search behavior, cited pages from external AI experiences where visible, and conversions from deep knowledge pages.

Qualitative review matters. If a page appears in search but attracts the wrong audience, the content may be too broad. If AI systems cite a page without capturing its nuance, the page may need stronger definitions and clearer headings.

Metrics should be interpreted by cluster. A niche article may create value by supporting authority, internal linking and sales conversations even when direct traffic is modest.

Electronic Artefacts applications

Electronic Artefacts should use AI search as a reason to strengthen its knowledge architecture, not to flatten it into trend posts. The strongest SEO content will explain difficult intersections: AI agents and consent, provenance and synthetic media, local AI and privacy, graph runtimes and editorial tooling, metadata and cultural memory.

The site can also act as proof of work. Every article can demonstrate the same discipline it advocates: stable IDs, visible sources, structured metadata, relationship panels and canonical pages.

This creates a defensible brand position. Electronic Artefacts does not merely write about knowledge graphs. It publishes through one.

Implementation

Start with a focused editorial cluster around AI search and knowledge architecture. For each article, define the primary entity, search intent, related concepts, sources, internal links, claims and expected reader action.

Generate JSON-LD from the same record that renders the page. Validate canonical URLs, sitemap inclusion, article schema, headings, citation metadata and relation statements. Avoid separate SEO-only content fields that drift away from the visible article.

Review each page against a simple question: would this still be useful if search traffic disappeared? If yes, it is likely good material for both humans and AI retrieval.

Evidence

Google Search Central describes generative AI search as rooted in core Search ranking and quality systems, including retrieval patterns over indexed pages. Its structured data guidance explains that structured data gives explicit clues about page meaning, while its AI search guidance emphasizes useful, non-commodity content and crawlable technical foundations.

Schema.org Article provides a shared vocabulary for describing article metadata. Electronic Artefacts already uses generated structured data, identifier routes and relation files to expose a machine-readable knowledge corpus.

Limitations

No markup guarantees inclusion in search results or AI answers. AI search interfaces may cite, summarize or omit sources in ways publishers cannot fully control. Structured data can improve clarity, but it cannot compensate for weak content, missing authority or poor crawlability.

Knowledge graph SEO also requires maintenance. Broken relations, stale sources and duplicate entities weaken the corpus over time.

Read Linked Data, Knowledge Graph, Metadata, Entity Identity and Retrieval-Augmented Generation.

See JSON-LD and RDF.

Glossary

AI search: search experiences that use generative AI, retrieval and summarization to answer user queries.

Knowledge graph SEO: SEO based on entities, relations, sources and structured data rather than isolated keyword pages.

Entity identity: a stable identifier for a concept, publication, project, organization or other knowledge object.

Structured data: machine-readable metadata embedded in a page or linked from it.

Non-commodity content: content with distinctive expertise, experience, evidence or point of view.

References

  • Google Search Central. Optimizing your website for generative AI features on Google Search.
  • Google Search Central. Introduction to structured data markup in Google Search.
  • Schema.org. Article.