Everything else we get asked.

Both, but mostly a platform. The plugin is tiny — a thin WordPress component that accepts a signed webhook and writes your verified JSON-LD to the page. The platform is where everything interesting happens: the 3-pass AI pipeline, the MCP server, the visibility tracker, and the Wikidata sync all run in our cloud. If it helps, think of the plugin as a receipt printer and the platform as the whole restaurant.

Yoast and RankMath are SEO plugins — their job is to help you rank better on Google search. They generate flat, isolated schema blocks intended to earn a Rich Result snippet. Schema Monkee is data infrastructure — our job is to build a deeply connected entity graph that ChatGPT, Claude, Gemini, Perplexity, and the next wave of AI agents can query. When Schema Monkee’s schema is present, we automatically suppress Yoast/RankMath’s JSON-LD to prevent conflict — they keep doing meta titles, sitemaps, and focus keywords, which they’re good at.

Just one small WordPress plugin. It’s sub-10KB, adds one REST endpoint and one wp_head filter, has no cron jobs, no AJAX loops, and no scanners. All the heavy lifting (Gemini, Wikidata, Playwright, Vision) runs in our cloud. Your server sees one signed webhook every time your schema is regenerated. More on the integration model here.

All of them. We’ve tested on WP Engine, Kinsta, Pantheon, Cloudways, SiteGround, and bare-metal LEMP stacks. The plugin is page-cache-safe and object-cache-safe. Works behind Cloudflare. No PHP extensions required beyond what core WordPress needs.

We replace it, but cleanly. When our graph is present, we automatically suppress JSON-LD output from Yoast, RankMath, and a handful of other schema-emitting plugins to prevent duplicate Organization blocks fighting for the root @id. If you deactivate Schema Monkee, the other plugins take over again — nothing is permanently rewritten, nothing is orphaned in your database.

Because single-pass generation is how hallucinations ship to production. Pass 1 (Generate) reads your site and builds a draft graph — but the model is guessing at your phone number, legal name, and NAP data. Pass 2 (Enforce) overrides every field with your verified Golden Source values. Pass 3 (Perfect) then optimizes architecture, relationships, and authority links on top of already-verified data, which is why we let it run without restrictions. Full walkthrough here.

Your verified, authoritative facts — exact legal business name, parsed and normalized address, E.164-formatted phone number, official email. You provide these during onboarding (or we extract them from your existing wp-config and LocalBusiness records and you confirm). Pass 2 uses this set as the override layer: anything Pass 1 hallucinated gets replaced with your Golden Source values before anything ships.

Up to 100,000 characters per generation cycle — typically your homepage, services, locations, about, team, and FAQ. For enterprise sites or multi-location franchises that exceed 100K characters of relevant content, we run the pipeline per-entity-cluster (e.g., per location) rather than per-site, and Pass 2 maintains sibling awareness across all of them.

The whole point of the three-pass architecture is that it can’t — at least not in the ways single-pass tools fail. Pass 2’s Golden Source override means we can’t publish a hallucinated phone number (it gets replaced before ship). Pass 3 only runs on data Pass 2 has verified. If something does go wrong, the webhook delivery is idempotent — we can roll back to the previous known-good graph instantly.

By default, weekly — enough to catch content changes without wasting compute. You can trigger a manual re-run from the dashboard anytime you publish a major change (new location, rebrand, leadership change). On higher tiers, we can also run nightly with drift detection.

Model Context Protocol — an open standard (introduced by Anthropic in late 2024, now adopted across major agent platforms) that lets AI clients query live data sources structurally, the same way they call their own tools. Schema Monkee turns your WordPress site into a native MCP provider so agents query you directly instead of scraping your HTML. Full primer here.

Claude Desktop supports MCP natively today. Cursor supports it. A growing list of third-party clients (including every IDE integration and agent framework built on Anthropic’s SDK) can query any MCP endpoint you point them at. Beyond that, Google’s Gemini, OpenAI’s GPT, and the other major labs are all moving toward similar or compatible standards. Publishing a proper MCP surface now is the rare SEO investment that compounds rather than erodes.

It’s a bookmark, not a service. llms.txt is a static file — agents fetch it once, cache it, and move on. A native MCP server is queried live: agents can ask “what are your retail locations?” or “what does your warranty cover?” and get a structured answer from your current graph, not a frozen snapshot. We serve llms.txt too (it’s automatic), but it’s the floor, not the ceiling.

Five: Google AI Overviews, ChatGPT (both Search mode and default conversational answers), Claude, Perplexity, and Gemini. For each, we capture bounding-box position, visual share, citation type, competitor co-mentions, and sentiment. Full details and sample JSON output here.

Yes — that’s the primary workflow. You give us the queries your customers actually ask, and we track those specifically. We also auto-generate a baseline set (usually 20-50 queries) by analyzing your site and its natural intent clusters, so you see immediate coverage on day one.

No measurable impact. The plugin adds one REST endpoint that wakes up only to validate an HMAC signature and write one option row, plus one wp_head filter that emits the JSON-LD you already wanted served. Your TTFB doesn’t change. Query Monitor stays clean. We specifically do not run cron jobs, scanners, or AJAX loops — the whole pipeline is in our cloud.

All fine. Our output is inline JSON-LD, which is static once written — it’s cached alongside your page HTML with no additional purge steps. When Pass 3 publishes a new graph, we call your standard cache-purge hook (or a configured cache plugin’s API) to invalidate affected pages, same as if you’d edited the post in WordPress.

Deactivate the plugin — site is byte-identical to before. Delete it — we also remove the one option row we created during setup. No leftover tables, no orphaned transients, no phantom cron entries. The only thing that persists is your account on the platform side, which you can close from the dashboard or by email.

Yes. In headless setups we skip the wp_head filter and expose the stored JSON-LD through a REST endpoint your frontend can fetch during SSR/SSG. If you’re on Next.js, Astro, or SvelteKit with WordPress as a headless CMS, you pull the graph at build time and inline it in your template — same functional result.

No. Your content is sent to Gemini 3 Flash Preview for inference only — Google’s enterprise terms for this model explicitly prohibit training-data use of API inputs. Nothing you give us is pooled into a dataset for future model improvements.

The generated graph, your Golden Source record, and your visibility history live in our primary region (US-East by default). Enterprise customers can request EU residency. Nothing about your business is stored or processed outside those regions.

Every delivery is signed with an HMAC-SHA256 signature using a per-site shared secret. The plugin rejects any request with an invalid or missing signature, or a timestamp more than 5 minutes out of sync (to prevent replay). The plugin also has no outbound internet access — it can only receive, never call home.

Yes. We’re the data processor, you’re the controller. A standard DPA is available on request. We don’t process personal data beyond what your public website already publishes, plus the email you provided during signup. EU data residency and right-to-erasure are both supported.

The free audit is free forever. Beyond that, Schema Monkee is a paid product — running Gemini, Playwright, and Vision at citation-grade quality isn’t something we can do at zero cost. We do offer a 14-day free trial on every paid tier, no card required.

A one-time run of Pass 1 against your homepage (to show you what a single-pass AI invents), a check of your MCP surface status, a scan of five major AI surfaces for your primary query, and a report on your current Wikidata footprint — all delivered as a white-label PDF within 10 minutes. Start here.

Yes — volume discounts kick in at 10 client sites or 10 franchise locations. Agencies also get the white-label PDF branding, GoHighLevel ATM proxy integration, and a client-management dashboard. Details on the Solutions page.

Yes. Every PDF can carry your logo, color palette, domain, and contact block instead of ours. Clients never see the Schema Monkee brand unless you want them to. Configuration takes about two minutes per brand in the agency dashboard — one config applies across all audits you run.

We ship as an Agentic Tools Mobile (ATM) proxy inside GoHighLevel. Your account managers fire audits, review reports, and trigger re-generation from the same workspace they already use for the rest of their client work — no extra tab, no extra login, no new training. It’s a standard GHL marketplace install.

Each location gets its own @graph with the correct parentOrganization reference pointing back to the franchisor’s @id. Pass 2 maintains cross-domain sibling awareness — the Dallas site knows about Denver and Detroit via sibling @id references, without duplicating parent data. A franchise acquisition becomes one record change in our franchise_network database, not a forty-site schema migration.