A marketing executive with a Wharton MBA and decades of Silicon Valley experience sat for a live quiz on sixteen AI marketing terms and scored 79 out of 99. That B-minus should comfort and alarm you in equal measure. If a practitioner this deep in the tools misses answers, the average marketing team is guessing at half its own vocabulary, and budget conversations are happening in a language nobody fully speaks.

The quiz ran on a recent episode of Misadventures in Marketing, the AMA San Francisco podcast, with Steve Haney, a fractional go-to-market executive in Silicon Valley, playing quizmaster to Peter Farago, a startup marketing executive in the Bay Area. Their exchange doubles as the most useful working glossary of AI marketing terms a B2B leader will read this quarter, so here it is, organized by what actually changed.

The AI marketing terms replacing search your dashboards still measure

Four terms cover the same tectonic movement. Generative engine optimization, or GEO, is the successor to SEO: structuring content so LLMs like Gemini, Perplexity, and ChatGPT cite and surface your brand rather than ranking you on a page of blue links. Answer engine optimization, AEO, is its near-twin, tuned for concise direct answers an assistant can read aloud. AI overviews, the synthesized summaries now sitting on top of Google results, have spawned a new metric, with marketers tracking AI share of voice instead of keyword rankings alone. LLM visibility rolls it all up: how often your brand appears in the training data and live outputs of the major models.

Peter’s practical notes were sharper than the definitions. Tables, FAQ structure, and a presence on Reddit all appear to improve your odds, since the models vacuum up enormous amounts of their material there. None of this shows up in a classic rank tracker, which means most dashboards are measuring a version of search that is quietly shrinking.

Agents need adult supervision

Agentic AI, the term of the year, means systems that perform tasks on a user’s behalf, authenticating into other tools, pulling data, and executing multi-step work. The essential companion term is HITL, human in the loop, the predetermined checkpoint where a person reviews output before it ships. Peter’s warning about autonomous SDR tools is worth pinning above every automation roadmap: “An agent doesn’t know what it knows or doesn’t know. And it doesn’t know to ask itself, hey, is this good judgment that I should send this? If you don’t think about that, the agent will just keep going and no one will have thought to tell it to stop.”

Prompt slop is the term for the failure mode. Lazy, generic prompting produces low-quality generic output, and as Steve summarized from his research, slop is the 2026 version of spam. Anyone receiving templated AI prospecting emails already knows the feeling, and deletes accordingly.

Signals are finally replacing segments

The most strategically interesting cluster covers targeting. Signal-based selling means acting on behavioral triggers, a pricing-page visit, a relevant job posting, a reported outage, rather than blasting static lead lists. Predictive intent scoring uses models to rank accounts on their propensity to buy now, rather than on firmographic fit alone. Both matter because of the 95-5 rule Peter cited: at any moment, roughly 95 percent of your addressable buyers are not in market, and only 5 percent are actively looking.

His larger point turns this from jargon into strategy. Classic segmentation variables, demographics for consumers and firmographics for B2B, have always been crude, grouping buyers the way astrology groups personalities. Intent signals offer the granularity the old STP model, segmenting, targeting, positioning, always promised and rarely delivered. Add the dark funnel, all the buyer activity you cannot see, from forwarded white papers to peer recommendations in private Slack channels, and the case for signal thinking gets stronger. Most of the decision happens where your attribution cannot follow.

You will end up managing token budgets

A handful of plumbing terms are crossing from engineering into marketing operations, and leaders who can define them will make better vendor decisions. RAG, retrieval-augmented generation, is how AI products ground answers in real source material. Fine-tuning trains a model on your brand voice, technical docs, and past winning emails. Explainable AI, or XAI, matters in regulated industries where teams must justify why a model targeted a specific group. The context window is the model’s working memory, and blowing past it is why long chats drift into confident nonsense, including invented customer quotes.

Then there are tokens, the currency of the whole system, roughly three-quarters of a word each. Since enterprise AI bills by the token, marketing leaders now manage token budgets the way they manage ad spend. That sentence would have sounded absurd two years ago. It will sound perfectly obvious within one more.

Peter’s closing caveat deserves the last word, because it separates fluency from competence: knowing the terms is easy, and doing these well requires systems, trial and error, and a lot of time. Vocabulary is the entry fee, not the win. The marketing leaders pulling ahead this year aren’t the ones who can define signal-based selling at a board meeting. They’re the ones already running it, measuring it, and quietly rewriting their org charts around what it reveals.

Listen to the full podcast on Spotify and Apple Podcasts.

Check out summaries from other episodes:

• What AI can’t replace on a marketing team and what should you outsource to AI
• Founder-led marketing: who’s actually driving the brand
• The Artist Is the Planet: Marketing When Distribution Is Free

Misadventures in Marketing is a weekly podcast by the AMA San Francisco chapter. Veteran Silicon Valley marketing execs Peter Farago and Steve Haney explore the messy, rewarding, and occasionally absurd world of high-tech marketing  –  especially in early-stage startups. Each episode covers real-world challenges, trends, and lessons from the front lines.