Entity SEO for AI Search: Fix Identity to Get Cited

Most of this week’s chatter is hand wringing about the September 2026 spam update, and sure, it matters. Here’s the official word if you’re tracking outages or turbulence: Search Status Dashboard. But if you’re getting skipped or misattributed in AI Overviews, your problem isn’t rankings, it’s identity. That’s why Charles Floate’s chat with James Dooley on Entity SEO is worth your time: it’s about making your brand, people, and products machine resolvable so AI systems can cite you reliably. One solid afternoon of entity hygiene will outlast months of chasing volatility. 🔧 (status.search.google.com)

// My reaction

Relief

Source and more context: Entity SEO for AI Search w/ James Dooley on Charles Floate’s channel — watch it here: https://youtu.be/eaVpWC3rxbQ. Also surfaced by James on LinkedIn if you want the thread: linkedin.com.

What entity SEO actually is (and isn’t)

Entity SEO is making machines certain about three things: who you are, what you offer, and how that relates to known things. Not just JSON LD pasted on a page. It’s alignment across three planes:

• On site: canonical naming, unique IDs, stable URLs, tight internal links.
• Off site: corroborated profiles (Wikidata when eligible, Crunchbase, LinkedIn, G2 or BBB), consistent NAP for GEO, supplier or manufacturer or author connections.
• Markup: Organization, Person, Product, and Article schemas that reuse the same preferred names and SameAs links.

If those disagree, models hedge. Hedging means you don’t get cited.

How AI search uses entities

Large models compress documents into vectors and tie them to nodes and relationships. Google’s AI Overviews use retrieval augmented generation that pulls from the live web index and the UI includes links to sources; your job is to make selection easy and unambiguous (Google developers on grounding/citations, Search help page). (ai.google.dev)

So make selection easy:

• Unambiguous: stable identities and clear relations.
• Consolidated: fewer, stronger canonical profiles beat 20 throwaway socials.
• Trusted by proxy: links and mentions from already known, editorially reviewed entities.

Disambiguation beats volume. One correct SameAs to a high quality profile can do more than 50 weak mentions. 🧭

The minimal viable entity (MVE) checklist

Do this once per brand; repeat for top authors and flagship products.

• Decide and use your canonical name and shortname site wide (title, H1, logo alt, legal footer).
• Organization schema with @id yoursite.com/#identity, legalName, and 5–10 high signal SameAs (LinkedIn, YouTube, Crunchbase, G2 or BBB, Wikidata if eligible). No junk directories.
• Person schema per author with stable @id and SameAs to LinkedIn, personal site, and one notable directory if applicable.
• Product or Service pages with unique @id, GTIN or SKU if a real product, or clear serviceType for services; connect to Organization via brand or manufacturer.
• Article schema referencing the author’s @id and the Organization publisher @id.
• About and Contact pages that explicitly name execs or authors and reference the same IDs.
• Internal link hubs: one topic hub per core intent; hub links to supporting articles; supporting articles link back with the exact head term where natural.

Off site corroboration that actually matters

Keep it short and high signal so you can maintain it.

• Wikidata item if notable. If not, Crunchbase or vetted industry registries with editorial review.
• Company LinkedIn, YouTube, and one reviewer platform that fits your model (G2, Capterra, BBB, Trustpilot).
• Manufacturer or supplier or partner pages that list you correctly and link to the canonical product or brand page.
• Author bios on third party publications that match on site names and link to your canonical bio.
• GEO reality: clean NAP on the primary listings you actually control; don’t obsess over 300 aggregator sites if you won’t maintain them.

Schema implementation ranked by impact

Two rules I don’t break:

• One Organization @id to rule them all. Reuse that @id across all schemas so nodes merge.
• SameAs only to editorially controlled or authoritative pages.

If you’re new to AI Overviews and LLM behavior, I mapped the operational side here: AI Search Visibility: Get Your Brand Recommended by LLMs (/blog/ai-search-visibility-llms) and the retrieval basics in AI Search Information Retrieval Primer (/blog/ai-search-ir-primer).

Internal linking as entity training data

• Build one hub per intent or topic. Each hub should define the concept, answer the core FAQs, and link to supporting content.
• Use precise anchors. Not “click here.” If the node is “industrial dehumidifier specs,” say that.
• Triangulate: brand ↔ product ↔ author. Link these where it helps the reader so models learn who said what about which thing. 📐
• Follow Google’s link best practices: make links crawlable, descriptive, and consistent.

Measurement: verify you exist where it counts

• Run branded, author, and product queries and check AI Overviews. Are you cited and linked? If not, ask why a machine would choose you over a stronger, clearer node.
• Ask leading LLMs neutral questions (“Who offers X in Y?”). Log whether you’re mentioned and which URL they cite.
• Track a fixed prompt set monthly. Log AIO citations, misattributions, and zero citation cases. Expect low referral clicks from AIO, one field study found about ~1% of AI Overview visits click a cited link (study link) so measure presence and brand mentions, not just traffic. 🔎 (arxiv.org)
• Audit schema coverage and SameAs consistency monthly. Tools: AI Visibility Checker (/tools/ai-visibility) and Schema Opportunity Analyzer (/tools/schema-analyzer).
• For context on desktop prevalence and tactics, see Google AI Overviews: Desktop Prevalence and SEO Moves (/blog/google-ai-overviews-desktop-seo).

Also, remember: you can’t fully turn AIO “off” via a permanent account setting; Google treats generative AI responses as part of Search and offers a Web filter to show text only links for that query (Google’s AIO help and FAQ page confirming the Web filter behavior: FAQ). AIO appears by default when Google deems it helpful and in supported regions. (support.google.com)

Common pitfalls I keep seeing

• Spray and pray SameAs: bloated lists to low quality directories.
• Multiple author names for the same person, initials, nicknames, company aliases. Pick one canonical form.
• Orphan product or service pages with no hub context.
• Schema that fights the page, JSON LD says one thing, H1 or body say another.
• Chasing links before fixing identity. That’s building a roof before framing the house.
• Treating AIO like blue links. You need concise, quotable passages and off site corroboration, or you’ll get summarized without being cited. 🚫

What I’d do this week

• Ship or fix Organization schema with a stable @id and a tight SameAs list, 5–10 high signal profiles.
• Publish a canonical brand About page and canonical bios for two authors; reference their @id in Article schema.
• Stand up one topic hub and connect five existing posts to it with crisp anchors.
• Update LinkedIn, YouTube, and one industry profile to match names and URLs exactly; add links back to your canonical pages.
• Document the entity map, Org, People, Products, key Hubs, in your SEO playbook; schedule a quarterly review. If you want help, here’s how I approach it: /services/seo-strategy 🧰

Entity work is compounding. Once machines know who you are and what you’re tied to, every new page and mention accrues more value—and AI answers start pulling you in by default.

Final takeaway

Dooley’s talk is a good reminder: you don’t need 100 new posts, you need unambiguous entities, clean relationships, and proof outside your site. Fix identity, then content, then links. That’s how you earn AI Overview and LLM citations and protect demand in GEO and classic Search. Keep this tight, then scale. That’s modern SEO or GEO or AI practice, make your brand citeable and extractable, or watch models send your traffic elsewhere.