Entity SEO

Entity SEO means helping Google and AI-powered search understand your entities and the relationships between them. It isn't only ranking for your name.

Your personal name is an entity. So is your brand name. So is your company name, if it differs from the brand. Each product or service you sell is one, and so is an idea or a slogan. If search engines can't tell how those connect (who founded what, which company makes which product, who is behind which idea), you haven't started entity SEO yet.

The test: draw every entity as a node and a line between each related pair. Can you explain why you drew each line? When you can, you understand your entities. Content and schema then exist to reinforce the reason each line is there.

Entity SEO isn't a separate campaign. It sits around everything else you do: every profile, article, post, title, description, image and video either confirms the lines on the map or misses the chance to. Rank for the exact name is one result you measure (see references/measurement.md). It isn't the goal itself.

Guardrails (state these early and never break them)

• No guarantees. You can't promise a Knowledge Panel or position #1. Results vary by user, location and device, and take days to weeks after indexing.
• Only true relationships. Every line on the map must survive a fact-check. No invented partnerships, awards, "as seen in", products or frameworks.
• Legitimate tactics only. Never scrape Google SERPs, bypass CAPTCHAs, buy links, spam directories or create fake profiles.
• Only real, owned profiles go in sameAs and off-site links.
• Consistency. The site, résumé/PDF, schema and every profile must describe the same entities and the same relationships. Contradictions confuse the graph.
• Don't stuff it. Reinforce relationships where they belong, the way a person would naturally write them. Skip the places where they don't fit.
• Secrets are secrets. API keys and service-account keys: chmod 600, never committed, never printed.

How it runs: actions around one saved map

Entity SEO can't be done in one call. Some steps need the owner's judgment (which relationships are true), some happen off the site (only the owner can edit their LinkedIn), and results only show weeks later. So the skill is split into actions. They all read and update one file, entities.yaml, kept in the project (template: templates/entities.yaml). That shared file is what keeps the schema, profile copy and articles describing the same relationships.

With no argument (/entity-seo), pick the next step from the project's state: no entities.yaml → map; map exists but no audit, or it's older than the last site change → audit; open on-site gaps → site; open off-site gaps → reinforce; site and profiles done → content, then measure on the cadence in references/measurement.md. Say which action you picked and why, then run it. The owner can always name an action explicitly (/entity-seo audit).

Actions in detail

map

Read references/entity-map.md and follow it:

1. Inventory entities from what already exists: the site (about page, product pages, footer, existing JSON-LD), résumé/PDF, LinkedIn, GitHub, company page.
2. Draw the relationship table: from, relationship, to, why (one plain sentence), schema property.
3. Give each entity a home: one canonical URL that is its @id.
4. Write the statement library: one-liners that name an entity and a related entity.
5. Show the map to the owner and get it confirmed before any later action uses it. Don't guess relationships into schema.

Individual vs organization: an individual's map centres on the Person (employer, past employers, education, own products, topics). An organization's map centres on the Organization (brand, founders, products/services, parent or subsidiaries, locations). Both apply the same method.

audit

For every relationship row, record four checks (see entity-map.md §6): stated in visible text, encoded in JSON-LD with @id links, confirmed off-site by an independent source, answered correctly when asked as a question in Google and 1–2 AI assistants. Also run the classic diagnosis:

• incognito name search: who outranks the site, and why
• site:{DOMAIN} indexing check, plus stale indexed PDFs
• word count on key pages (flag under 500 words, or under 1500 for cornerstone pages)

Rank gaps by leverage. Broken person ↔ company ↔ product chains come first.

site

Read references/on-site.md (the multi-entity graph, entity pages, titles, canonicals, sitemap/robots, image SEO, content quality) and references/geo-aeo.md (E-E-A-T, AI-citable content, query intent, answer formatting, and the status table of which answer markup still earns anything). For local businesses, read references/local-voice-seo.md too. Key moves:

• Every entity gets a node with its own @id; relationships reference other nodes by @id, never anonymous inline objects.
• Articles declare about/mentions for the entities they cover, as well as author.
• Visible text states each relationship the schema claims. Schema never claims what the page doesn't say. Deliverable: a deployable patch. Verify with a build and by inspecting the rendered JSON-LD (every @id reference resolves to a node in the graph).

reinforce

Read references/reinforcement.md (per-surface checklist) and references/off-site.md (profiles and authority). Produce a checklist per surface with the exact copy from the statement library, and name which relationships each surface should confirm. Include the other end of each relationship too: the company page listing the founder, the app-store listing naming the developer, the product README linking the maker. Draft what can be drafted, such as READMEs and republish canonicals. The owner applies the rest.

content

Publish substantive, first-hand pages that demonstrate relationships: case studies of work done at an employer, build logs for a product, the story behind a method. Each page names its entities, links to their home pages and declares about/mentions. Every entity without a real page gets one here. No generic AI summaries: real decisions, numbers and outcomes.

measure

Read references/measurement.md: GSC setup and the daily rank check (scripts/gsc_rank.py), the Knowledge Graph check (scripts/kg_check.py), and relationship questions ("who founded X", "who makes Y", "what is Z") asked in Google and AI assistants, logged against the map. A wrong or missing answer points at the line that needs more support.

report (optional)

Read references/reporting.md. Produce a baseline after the first audit, then re-score after each iterate cycle to show the change.

Iterate

Follow the weekly/monthly/quarterly cadence in references/measurement.md. When something changes (new product, role change, rebrand), update entities.yaml first, then re-run audit and let the gaps drive the next actions. If on-site work is done and results are stuck, the lever is more off-site confirmation and more content that demonstrates the relationships. More on-site tweaks won't help.

A worked example (saadsalman.org) is in references/worked-example.md. Tool recommendations (free first, paid only once volume justifies it) are in references/tools.md.