Positioning — GEO and the Agentic Action Layer

Source: ../source/JSGlobtech_GEO_Agency_Playbook.pdf §1 (Executive Summary & Market Positioning), extracted 2026-07-30. Repo claims cross-checked against the tree on the same date.

The thesis

Search behaviour is moving from traditional search engines to LLMs — ChatGPT, Gemini, Perplexity. A business that is invisible to those systems loses demand it never sees.

The distinction the playbook draws:

	Focus	Optimises for
SEO	Human-facing links and rankings	A person clicking a blue link
GEO	Machine-readable facts an LLM can ingest	An AI engine citing, recommending, and factualising the business

GEO is therefore not content rewriting. It is publishing the business as structured, machine-consumable fact so a generative engine can represent it accurately without hallucinating.

The differentiator: Agentic Action Layer (MCP)

The playbook's stated "core unfair advantage":

Legacy marketing agencies only offer content rewrites. JSGlobtech installs the Agentic Action Layer using the Model Context Protocol (MCP).

The intended consequence is that an AI engine does not merely recommend the client — it can act on the client's behalf inside the chat interface: capture leads, book consultations, and process reservations.

This splits the offer into two layers that are worth keeping conceptually separate, because they are priced and delivered separately:

1. Visibility layer (GEO) — schema, entity consistency, llms.txt, content restructuring. Makes the engine say the right thing.
2. Action layer (MCP / agentic) — remote MCP server plus OpenAPI actions with real tools bound to real business systems. Makes the engine do something.

The action layer is the defensible part. Any agency can rewrite copy; wiring a tool-exposing server into a client's CRM is engineering work, and JSGlobtech is already a development shop.

Reality check against this repository

The action layer is genuinely implemented on JSGlobtech's own site, which matters because the demo depends on it (../sales/demo-framework.md):

• Remote MCP endpoint: app/api/mcp/[transport]/route.ts, stateless Streamable HTTP, transports sse and mcp.
• One registered tool: request_dev_consultation in lib/mcp-server.ts.
• OpenAPI mirror for ChatGPT Actions: public/docs/openapi.json → POST /api/leads/submit.

Two claims in §1 are not implemented anywhere in this repo and should be treated as capability JSGlobtech can build rather than capability it ships today:

• "book consultations" — the tool captures a lead; there is no calendar or scheduling integration.
• "process reservations" — no booking, inventory, or reservation system exists. This maps to the Tier 3 "Live Inventory/Property sync" line, which is aspirational.

See ../integrations/mcp-server.md for the verified surface.

Market position

Attribute	Value
Base	Dubai, UAE (Dubai Marina per lib/content.ts and public/llms.txt)
Stated target market	UAE & global markets
Practical outreach focus	UAE / Dubai high-CAC service businesses (see ./target-markets.md)
Delivery identity	Software engineering agency, not a marketing agency

Being Dubai-based is used as a proximity and credibility signal for UAE accounts, while delivery itself is remote-capable and therefore globally sellable. The site reinforces the local signal deliberately: components/GeoSchema.tsx publishes Dubai Marina address, GeoCoordinates, areaServed for Dubai and the UAE, and DHA/DED compliance alignment as a PropertyValue.

Tension with published positioning

The playbook positions JSGlobtech as a GEO agency selling productised GEO/MCP retainers to clinics and rental companies.

The live site positions JSGlobtech as a development agency whose three services are Flutter apps, React/Next.js platforms, and AI-native MCP integrations — with GEO/MCP described in public/llms.txt as "Scoped per engagement (often bundled with app/web builds)".

Both can be true, but they imply different lead sources, pricing, and delivery motions. This is the same unresolved question tracked in ../sales/pricing-tiers.md and ../decisions/0005-productized-tiers-vs-custom-scoping.md — status Proposed, needs an owner decision.

Talking points that survive scrutiny

Use these when the audience is technical enough to push back:

• GEO artifacts are verifiable, not promises: /.well-known/llms.txt and JSON-LD are fetchable by the client on the spot.
• The MCP endpoint is live and public, so "AI can act, not just mention us" is demonstrable rather than asserted.
• Lead delivery is a single webhook hop, so it can be pointed at whatever the client already uses.

Avoid asserting the "over 40%" statistic without a citation — see ./kpis.md.