Glossary — the map's vocabulary

Load-bearing terms and the structural concepts the modules assume. Volatile specifics (tools, stats) are dated and cited to their sources.

Structural concepts (how the map is built)

Discipline-primary spine — the map's top-level organizing axis: the 16 disciplines (A–P), not the funnel/PESO/lifecycle. Chosen because it's the only axis where a single domain is a shippable slice and boundaries match real teams/tools/budgets.
Type-dependent organizer — the resolved finding that each domain's internal spine differs by domain type: some disciplines run on a decision axis, others on a graded property, an audience lens, or a craft-taxonomy × governance-gate (a few are two-natured). The "one decision axis per domain" rule was deliberately not locked as universal — un-primed tests broke it. (Which organizer governs a given discipline is worked through in that discipline's chapter.)
Human↔agent audience lens — the lead secondary lens applied in every domain: who/what is being marketed to — humans (B2B/B2C/DTC) or agents (B2A/A2A). Promoted to a full domain (L) only where the agent pole earned distinct components; elsewhere a lens.
Continuous threads (4) — concerns that run through every domain, each with a home domain that owns the methodology: brand (B), data/measurement (I), compliance (K), budget/unit-economics (A). Pattern: home owns the method; every domain owns the application. ("AI" is deliberately not a thread — it decomposes into the human↔agent lens + the maturity lens + domains J/L.)
Funnel-as-lens — the customer-journey funnel was demoted in all 16 domains (it's non-linear under AI discovery, loops in lifecycle/community/commerce); kept only as a minor lens where work is genuinely stage-shaped.
AI-maturity (dated lens) — manual → automated → agentic; a fast-decaying lens (lives in the ledger), never a structural spine.
The hedge / leading re-test trigger — the map is discipline-primary now but built to flip toward agent-primary if agents become the dominant buyer; the trigger = autonomous-agent-completed order share (watch 5% / trigger 10%), currently dormant because ~70% of AI traffic is referrerless and unmeasurable.

Audience & agents

B2A (business-to-agent) — marketing to/through the AI assistants that discover, evaluate, and buy on a person's behalf.
A2A (agent-to-agent) — the seller's agents transacting/negotiating with the buyer's agents.
Agent-as-operator vs agent-as-audience — an agent that runs marketing work for the brand (e.g. an AI service agent) vs. an agent that is the recipient of marketing (a buyer's agent). Several domains have both poles.
Agentic commerce — agents completing purchases; rails incl. UCP (Universal Commerce Protocol, Google/Shopify), ACP (Agentic Commerce Protocol, OpenAI/Stripe), AP2 (Agent Payments Protocol, Google→FIDO), Web Bot Auth (agent identity).
Machine-eligibility — being structured/feed-complete enough for an agent to retrieve, understand, and select you (Domain L's supply-side play).
MCP (Model Context Protocol) — the machine interface (MCP servers, brand-as-MCP, an own-site MCP endpoint) through which AI tools and agents call a brand's systems, guidelines, and catalog.
GTIN — a product's standard global identifier used in feeds and catalogs; treated as mandatory in practice for an agent to find and select a product (machine-eligibility).
KYA (Know-Your-Agent) — the emerging duty/practice of verifying an agent's identity before transacting with it; the agent-era analog of KYC.
HITL (human-in-the-loop) — a human review, approval, or escalation step kept inside an otherwise-automated workflow.

Search & discovery (E)

SEO / GEO / AEO / LLMO — link-list optimization (SEO) · being cited in AI answers (GEO/AEO = generative/answer-engine optimization) · optimizing for LLMs (LLMO). The map organizes E by retrieval-mode (link-list → answer-synthesis → agent-selection), not by these labels.
Zero-click — searches resolved without a click to an external site; the dominant 2026 reality that makes GEO "influence without a click."
Retrieval-mode — E's organizer. The axis itself — and how to run it — is worked in the full map.

Measurement (I) & the discipline

Incrementality is the arbiter — the load-bearing rule: no platform-reported ROAS/lift/deflection number is a fact until an independent holdout/geo-test confirms it (platforms over-attribute ~20–60%). Anchored to peer-reviewed economics (Lewis-Rao, Gordon et al., Ghost Ads), not vendor decks.
EMV (earned media value) — banned as a success metric (vendor-favored, no standard definition); use incrementality/MMM/self-reported attribution.
MMM / MTA / DDA — marketing-mix modeling · multi-touch · data-driven attribution; the layered stack (MTA daily / MMM strategic / incrementality arbiter).
Signal loss — the post-cookie/ATT erosion of identity-level data that I organizes around (basis-of-truth × signal-integrity).
MER (marketing efficiency ratio) — blended revenue ÷ total spend; part of the shared unit-economics language I reports in.
CAC / LTV / LTV:CAC / payback — the shared unit-economics language: customer-acquisition cost, lifetime value, their ratio, and payback period; owned by I and read as bands, not exact points.

Paid, partner & the rest

Buying-model × mediation — F's organizer. The axis itself — and how to run it — is worked in the full map.
Rail-control / surface-strategy — L's organizer. The axis itself — and how to run it — is worked in the full map.
Create-vs-capture — N1 (affiliate)'s organizer. The axis itself — and how to run it — is worked in the full map.
PESO — paid/earned/shared/owned; a minor classification lens, not a spine.
The four-way agent-surface rule — for llms.txt/schema/MCP on owned surfaces: C authors · D/P hosts · E specifies citability + owns the efficacy verdict · L defines commerce semantics.

Production

Dual mandate — every module must both ADVISE (decision logic) and RUN (executable workflow). One without the other fails the build-spec.
Locked baseline — the frozen v1.0 taxonomy the published edition is built against, so the map's structure stays stable.