AI SEO foundation — what publishing on AILK does for an AI SEO agency

For agencies whose deliverable is AI search visibility, the platform the client publishes on determines whether the work shows up.

The ai seo cluster runs at 5K monthly searches across six correlated terms — ai for seo, ai and seo, ai seo tools, ai tools for seo, artificial intelligence seo. These are AEO-practitioner agencies researching where their citation work lands. This article explains what the foundation has to do and what AILK ships out of the box.

The SEO deliverable has changed

When ChatGPT recommends a business, the recommendation was assembled from sources. The model read pages. Attributed them to a brand. Quoted what it trusted. The brand that gets recommended is the brand whose pages got cited.

That changes what the agency's deliverable is.

Traditional SEO delivered rankings. You ran a campaign, you measured positions, you reported traffic. The underlying assumption was one entry point: Google's blue links. Position one or two meant visibility. Position three through ten still meant something.

AI search doesn't work that way. There is no rank. There is cited and not cited. A ChatGPT answer to "what's the best foundation for a marketing agency site" either names your client or it doesn't. An AI Overview for a local service query either attributes your client or it attributes a competitor.

The implication: for an agency whose deliverable is AI search visibility, the architecture of the client's site is no longer a technical preference. It's a campaign requirement.

The page is the unit of citation, not the site.

A site doesn't get cited. Pages do. The AI system read a page, found an entity, trusted the structure, and quoted it. The campaign work — content, schema, entity consistency — happens at the page level. If the platform makes page-level schema correct and citation-ready by default, the agency's work lands. If the platform requires the agency to fight the foundation on every page, the work lands less reliably and requires more re-fighting to stay landed.

What the foundation has to do

If you were specifying an AI-search-ready foundation from scratch — knowing what AI search has become — five things would be true.

Be read correctly. The model should know what each page is about and which business it belongs to. An entity-opaque page — one that uses "we" and "our service" instead of the business name, or one that emits no JSON-LD — forces the model to guess. Models that have to guess cite someone else who doesn't make them guess.

Work for non-human visitors. Agents now research on behalf of buyers. A client's prospective customer can ask an AI agent to research options, compare services, and return a recommendation — without visiting a single page directly. The foundation should expose itself to those agents as a first-class audience, not a degraded approximation.

Surface its own gaps. AEO problems are silent. The site doesn't crash. It just doesn't get cited. A foundation that catches structural gaps the way a linter catches type errors — before the site ships, in CI — shifts the cost from "three months after the client notices citations drifted" to "before the site went live." Those are not the same cost.

Belong to the people running on it. For an agency reselling client engagements, the foundation is either a partner or a tax. Apache 2.0 means the agency owns the substrate. If a host gets expensive or a CMS vendor raises rates, the codebase leaves with the agency. The engagement economics don't depend on anyone else's pricing decisions.

Fit inside the engagement. An SMB site engagement runs weeks, not quarters. A foundation that requires a multi-month custom build breaks the engagement economics before the first line of client-facing code is written. The deploy path has to fit the work.

Any foundation can be held against these five. Most fail three.

Where the alternatives break

WordPress + plugins. Forty percent of the web. Every AEO plugin is real — Yoast, Rank Math, Schema Pro. The structural limit: they don't integrate with each other, they break on updates, and the schema work depends on whichever combination is installed this month. An agency running AI SEO campaigns on WordPress re-fights plugin conflicts on every campaign. The foundation was built when Google was the only entry point. The plugins are retrofits. The agency is doing AEO work on infrastructure that was not designed for it.

Closed website builders. Squarespace, Wix, Webflow. Beautiful templates. Weak schema. No agent surface. Vendor lock-in is the business model. The agency resells on someone else's roadmap. If the vendor's schema support is six months behind schema.org, the agency's client sites are six months behind. The agency has no path to fix this.

AI-generated builders. Wix AI, Hostinger AI, Durable. "Describe your business, AI generates a site." The structural problem: the site is generated by AI and then invisible to the AI systems doing discovery. AI-generated, not AI-ready. The inversion is complete.

SaaS starter kits. ShipFast, makerkit, create-t3-app. Ship a SaaS in a weekend. Auth, dashboards, billing. Wrong shape. These are starters for B2B software products. SMB marketing sites need case studies, FAQs, local-business identity, schema-typed content surfaces. The agency builds the content and AEO side from scratch on top.

The cost is borne by the agency's campaign capacity. Every hour of platform fighting is an hour not spent on the citation work the client actually paid for.

What AILK ships as defaults

AI Launch Kit (AILK) is the open-source website foundation for the agent-first internet. Apache 2.0. Deployable in days. The structural prerequisites for AI search visibility ship as defaults, not add-ons.

A typed schema registry, native to the foundation. Forty-five page types. Per-type Zod validation. JSON-LD generated from frontmatter. When a content author declares type: BlogPosting in a page file, the schema builder emits the correct BlogPosting JSON-LD — headline, datePublished, author, publisher — automatically. No form. No plugin. No snippet to paste. Ship a page, ship the structure.

Peer surfaces over one substrate. apps/web for humans. apps/mcp for agents. apps/api for programmatic callers. One repo. One-to-one parity, CI-enforced. Whatever a person can do on the client's site, an agent can do through a typed contract. This is the architecture, not a bolt-on integration added after launch.

ailk audit CLI, in the OSS. Runs pre-launch. Runs in CI. Catches missing schema, ambiguous entity markup, and pages that don't emit JSON-LD — before the site ships. Enterprise AEO platforms charge $295/month to measure the same problems after they've already cost citations. ailk audit catches them before launch. Free, in the OSS.

Apache 2.0. Complete. Not source-available. Not "free until you make money." Paid tiers add capability. They don't gate the foundation. The agency owns the substrate. The downside is bounded: if AILK never finds traction, the agency keeps the codebase and keeps the client.

Agent-assisted scaffolding. create-ailk generates the project. The .claude/ skill bundle handles initial deploy and ongoing iteration — add a service page, restructure a content surface, respond to a citation gap. On WordPress, every new page is a project. On AILK, it's a Claude Code session. The marginal cost of a page approaches zero.

How the citation work compounds

On WordPress plus plugins, AEO work fights the platform. Every new campaign re-fights plugin conflicts, update-related schema regressions, and weak entity markup. The work does not compound. It maintains.

On AILK, every page ships already structured for citation. Entity markup is consistent across the site because the schema registry enforces it at the type level. The ailk audit CLI catches gaps in the development cycle, not the client-review cycle. New pages pick up the right schema without manual intervention.

The citation work lands. It stays landed. The next campaign builds on what the last one established rather than re-doing it.

The window. The next 12–18 months are when AI search behavior cements. Pages cited today shape pages cited tomorrow — model retrieval leans toward sources already in the answer set. Foundations adopted now compound. Foundations adopted in 2027 chase ground that's already been taken.

Frequently asked

What is AI SEO?

AI SEO is the practice of optimizing a site's pages to be cited by AI-powered search systems — ChatGPT, Perplexity, Google AI Overviews, and similar surfaces that assemble recommendations from sources rather than ranking links. The deliverable is citation share, not position. The structural requirements — entity clarity, schema markup, agent-accessible surfaces — differ from traditional search ranking signals.

What is AEO and how does it relate to AI SEO?

AEO stands for Answer Engine Optimization. It is the practice of structuring pages so that AI systems can extract, attribute, and cite the answers they contain. AI SEO is the broader discipline; AEO is a specific technical layer within it. An agency doing AI SEO work for a client is doing AEO work on that client's pages — making sure the schema is correct, the entity markup is consistent, and the content leads with the answer rather than burying it.

Does AILK replace the agency's SEO and AEO workflow?

No. AILK is the foundation — the substrate the client's site publishes on. It ships the structural prerequisites: schema registry, agent surface, AEO audit CLI. The agency's campaign work — content strategy, keyword targeting, entity authority building, link acquisition — runs on top. AILK makes that work land more reliably because the foundation is not fighting it.

How does AILK handle entity markup across the site?

The schema registry's SiteIdentityPack carries stable site-wide metadata — organization name, URL, logo — that every page builder uses to populate publisher, url, and inLanguage. Entity consistency across pages is a product of the architecture, not a manual per-page step. The ailk audit CLI checks entity markup as one of its four launch rules — flagging any page where the organization entity is inconsistent or absent.

What does the ailk audit CLI check?

Four rules at launch: schema coverage (does a JSON-LD block exist on this page?), JSON-LD validity (does the block parse correctly and conform to schema.org?), entity markup consistency (is the organization entity consistent across the site?), and FAQ/HowTo incentive (are pages that would benefit from FAQ schema missing it?). The CLI outputs a score and a per-rule breakdown. --format json pipes output to other tooling. --baseline exits with status 1 if the score regresses from a saved baseline — CI integration in one flag.

Can an agency adopt AILK for existing clients or only new builds?

The Launch Service — a fixed-scope two-week deployment — is the standard path for new builds. For existing clients, the question is whether the current site's foundation can carry the AI SEO work the agency needs to deliver. Most cannot — not because they can't install schema plugins, but because the underlying architecture was not built for agent-readable surfaces and per-type schema at the foundation level. An honest evaluation is the right first step. The ailk audit CLI can audit any AILK site — comparing a client's current schema coverage against an AILK baseline is a place to start the conversation.

The foundation is what makes the citation work compound

Talk to us about what AILK looks like for your agency stack. Or run the audit CLI against your current site to see what the score says.