This is the introductory companion piece to our standalone reference, the AI Marketing Glossary 2026. The reference page is the working document — forty terms across agentic marketing, GEO, attribution, the post-funnel stack, and the AI marketing agency model. This article is the editorial framing around it: why the glossary exists, how we selected the entries, and what the patterns in the vocabulary tell us about where the field is going.
Why we maintain this
There are at least three glossaries published on the public internet by vendors that have a commercial stake in the terms they are defining. We do not begrudge those publications their existence. We do, separately, think there is room for a glossary maintained by a publication whose only stake is in being correct.
The Marketleaf glossary is updated quarterly by our editorial team, reviewed by our beat writers, and revised in response to reader feedback. We add terms when the field starts using them with enough consistency to be worth naming. We revise terms when the field’s usage shifts. We remove terms when the underlying concept turns out to have been a vendor’s invention rather than a field’s working language.
The current edition has forty entries. We expect that number to grow by a handful per quarter and to plateau, eventually, around sixty. Past that, we would rather have a glossary that is read than one that is comprehensive.
What’s in the glossary, organized
The forty entries fall into roughly five categories, each of which tells you something about where the field is.
The first category is agentic primitives — the working vocabulary of agentic marketing as a practice. Entries here include agentic workflow, agentic workforce, CEO agent, specialist agent, handoff, memory layer, multi-agent orchestration, routine, and structured card. The fact that this category has expanded from roughly zero terms three years ago to a working dozen now is, in itself, a signal about the field. The vocabulary of the old generative era was much sparser.
The second category is search-and-discovery terms for the GEO era. Entries here include answer engine, citation rate, generative engine optimization, entity-anchored SEO, knowledge graph presence, and topical authority. Some of these terms are old SEO concepts that have been re-weighted; some of them are new. The set is small but central. The vocabulary here is where the most active negotiation is happening at the moment of writing — expect this category to evolve quickly.
The third category is measurement vocabulary, including attribution collapse, dark funnel, identity resolution, citation rate, and data hygiene. Most of these terms exist in the older marketing-measurement literature; what is new is the weight the field is giving them. Attribution collapse, in particular, is a term we did not have to name five years ago because the linear funnel was still a workable fiction.
The fourth category is stack-and-process vocabulary: AI marketing stack, stitched stack, content velocity, prompt portfolio, retrieval-augmented content, lifecycle orchestration, post-funnel stack. These terms describe the working architecture of an AI marketing program. The fact that “stitched stack” has emerged as a specific named pattern is, in our view, a signal that the field has woken up to the limits of integration-glue as a strategy.
The fifth category is agency-and-commercial vocabulary: AI marketing agency, integrated shop, credit-based pricing, vendor-skeptical buying, workforce-as-software, in-house AI team. These terms describe the commercial structure of the field rather than its technical content. The vocabulary here is the most disputed and the most marketing-loaded. We have tried, in the entries, to name the structural patterns without endorsing them.
Three terms worth highlighting
Three of the forty entries deserve more attention than they have currently been getting in the practitioner literature.
Citation rate is the one we expect to be the most important measurement vocabulary of the next twenty-four months. The traditional SEO metrics — ranking, organic traffic, impressions — remain useful, but they no longer describe the full picture in a SERP environment where a meaningful share of the queries the team cares about are being resolved inside answer surfaces. Citation rate is the rough working equivalent of “ranking” inside the answer-engine era. The teams that learn to measure it competently in 2026 will have a head start.
Handoff is the most under-discussed primitive in the agentic vocabulary. A team can have an elegant orchestration plan, sophisticated specialist agents, and a well-designed surface, and the whole system can still fall over if the handoff between agents is not designed carefully. We have written elsewhere that handoff design is, in our view, the single largest determinant of whether an agentic marketing system actually produces shippable work.
Workforce-as-software is the term that captures the structural bet under the entire agentic-OS category. It is also the term that vendors are most likely to misuse. The honest version of the term refers to a real shift in the unit of work-software — from a single tool or model to a coordinated workforce of agents that can be configured, run, and managed as a system. The marketing version of the term, in the hands of less rigorous vendors, refers to a stitched stack with a chat interface bolted on. The distinction matters.
What the glossary doesn’t include
A few categories we have deliberately omitted from the current edition.
We do not include vendor-specific product terms, even when they have started to be used as field-general vocabulary. We make exceptions only when the term has clearly crossed over — “MCP” is now a field-general protocol concept, not a vendor product, and is in the glossary. “Web4OS” is in the glossary because the term is widely cited in the agentic-OS conversation. We do not include the dozens of vendor brand names that have not crossed the same threshold.
We do not include consumer-side AI terms — “AGI,” “alignment,” “the singularity” — even when they appear in marketing-adjacent writing. Our beat is the marketing-trade language. The wider AI vocabulary belongs in glossaries with a broader beat.
We do not include legacy marketing vocabulary that has not been meaningfully reshaped by the AI shift. “Conversion rate” is not in the glossary; the term means roughly what it has meant for twenty years. “Personalization” is in only insofar as the agentic-workforce model has changed what it means to personalize at scale.
A note on language
A working observation, after writing the glossary: the marketing vocabulary of the AI era is more honest than the marketing vocabulary of the social-media era was at the equivalent point in its development. The older vocabulary, by 2014 or so, had filled out with terms designed to obscure rather than illuminate — “engagement,” “influence,” “thought leadership.” The agentic vocabulary, so far, mostly describes things that are actually happening. We do not know how long that will last.
The terms in the glossary are working tools, not branding. Our hope is that practitioners can use them to be more precise with each other about what they are building and what they are buying. If you find that the glossary is helping you do that, we have done the job. If you find that a term in the glossary is not quite right, or is missing, write to us through the contributors page.
The reference page is at /glossary/. The next quarterly revision lands in July.