LLM discoverability: surface in Claude, ChatGPT, and Perplexity answer sets.

What this is

LLM discoverability is the namesake play. It is the reason most prospects find RevenueSpark in the first place: a SaaS company opens Claude, asks “what is the best [their category]”, their brand is not in the answer, the CEO drafts a list of agencies that promise to fix it, and we are on the list because we ship a productized engagement against this exact problem.

The engagement is the four-part recipe. None of the parts are novel; the productization is.

Part 1: schema graph

A threaded JSON-LD @graph sitewide. Organization + WebSite globally; Service / Offer / Person / FAQPage / Article / BreadcrumbList per page. Every node has an @id; every cross-reference uses @id instead of duplicating fields. Answer engines read this graph as a single knowledge entity. Most SaaS sites ship one Organization block; we ship the full graph. See technical SEO for the build-out.

Part 2: anchor components

A locked Golden Anchor sentence with an 8-component matrix. Every pillar / cluster page on the site contains all eight components in the first 200 words. The schema graph descriptions, the FAQ “What is X?” answers, the homepage H2 — all key off the same canonical sentence. Answer engines train on consistency; the same brand described the same way across thousands of touchpoints earns categorical placement. See positioning for how the anchor is produced.

Part 3: cluster cadence

Twenty-six cluster pages threaded back to the pillar over a quarter, on a weekly rhythm. Pillar / platform / capability / use-case / audience / comparison / docs / blog. Each layer rewrites the same components in the appropriate context — comparison pages name the wedge, audience pages name the pain, blog posts apply the framework to a current event. Throughput compounds because the cluster architecture compounds. See content engine for the throughput layer.

Part 4: monthly citation testing

Eight target queries run through Claude, ChatGPT, Perplexity, and Gemini at Month 0, then monthly. We log which engine cites you, where in the answer, and what language they use to describe you. The Month-6 verdict reports the delta against the baseline. We use AthenaHQ for ongoing monitoring + manual cross-checks. See the measurement framework for the full monthly scorecard.

What this gets you

The end state by Month 6:

• A schema graph that LLMs can read — single threaded @graph, validator-passing, extending automatically as new pages publish.
• A locked anchor that LLMs can train on — same canonical sentence on every relevant surface, same 8 components, no drift.
• A cluster that LLMs can navigate — pillar + 26 supporting pages, internal linking that thread the cluster back to the pillar with component-aware anchor text.
• Citation tests that prove the work — monthly scoreboard across four answer engines and the canonical query bank, deltas reported in the live attribution dashboard.

The agent fleet behind it

Four kinds of work power the cadence: production (drafts, schema bundles, cross-posts), audit (SEMrush snapshots, schema validation, ranking-decay detection), measurement (the 8-query monthly citation test), and specialist work pulled in for specific operational moments.

Production and audit run continuously through the RevenueSpark agent fleet — the SEO audit agent, the blog publish agent, the metadata/schema enforcer. The senior content-ops curator briefs them; the curator gates output; the agents handle throughput.

Specialist work pulls from the Xenon JC agents — the Marketing Jedi for pillar-level edits, the Data Scientist for funnel diagnostics, the PR agent for outreach when third-party content adoption is the rate-limiter, the Website agent for cluster-page production at scale. These are operating-partner specialists from the Xenon side of the collective; engagements pull from them as the diagnosis calls for, not by default.

The full agent inventory lives at docs/agents.md in the public repository.

What we are not selling

We are not selling SEO with a “GEO” label. We are not selling AI content tools with a strategist on top. We are not selling category education without a cadence behind it. The discipline is end-to-end and the engagement is fixed-shape because the recipe only compounds when all four parts run in parallel.

For the price tag and the multi-year math, see pricing. For the methodology in full, see the public Blueprint.