Seven layers. Skip one and the engine skips you.
Most AI visibility problems are not media problems. They are stack problems. One layer is broken — and the engine skips the entity entirely.
This is the AI Communications Stack: seven layers that determine whether a brand, founder, institution, or idea appears inside AI-generated answers. Foundation at the bottom. Action in the middle. Operations on top. Each layer sits on the layer below it the way a building's third floor depends on the second.
I am publishing it here as the second canonical piece in the AI Communications series — because if the discipline is going to mean something, it needs a shared architecture. Most of what is being sold as "AI visibility" right now is one or two layers being sold as a stack.
"AI Communications is a mix of journalism, psychology, and engineering."
The seven layers
Layer 1 — Identity Layer 2 — Authority Layer 3 — Earned Media Layer 4 — Generative Engine Optimization (GEO) Layer 5 — Amplification Layer 6 — Measurement Layer 7 — Defense
The headline metric across all seven layers is Citation Share — how often an entity appears inside relevant AI-generated answers compared to competitors. Each layer moves it. None of them moves it alone.
Layer 1 — Identity
The canonical answer to what is this entity. Who it is. What it does. Why it matters. Stated cleanly, consistently, repeatedly, across every owned surface.
This sounds elementary. It is the most-skipped layer in the stack.
When an AI engine assembles an answer, it is trying to produce something coherent. If the entity says one thing on its website, another in its founder's LinkedIn bio, a third in an old press release, and a fourth in a Wikipedia entry written by an intern in 2018, the engine averages. The average is usually wrong.
The work at Layer 1: a single canonical positioning statement, a single canonical biography, a single canonical product description, deployed consistently across the entity's owned properties. Every higher layer compounds on this one. Every higher layer fails when this one is incoherent.
What gets it wrong: assuming the engines will figure out a coherent identity from inconsistent inputs. They won't. They will assemble whatever average is statistically most likely — and the average is rarely flattering.
Layer 2 — Authority
Structured proof. The third-party signals that tell an engine an entity is real, established, credentialed, and worth retrieving. Wikipedia. Wikidata. Schema markup. Glossary inclusion. Primary research. Author bylines in publications the engines trust.
In plain language: AI systems trust entities that are consistently documented, structured, and independently verified across the web.
This is the most technical layer for communications professionals to absorb. It is also the highest-leverage. A single well-structured Wikipedia entry, accurately sourced, can do more for an entity's Citation Share than a year of media tour appearances.
The work at Layer 2: Wikipedia and Wikidata entries that are accurate, sourced, and current. Schema markup on every important page. Primary research — surveys, reports, original data — published under the entity's name. Author bylines in publications the engines trust.
What gets it wrong: treating schema and Wikipedia as IT projects rather than communications work. They are communications work. The communications team that does not own them will lose to the team that does.
Layer 3 — Earned Media
The substrate. The publications the engines actually pull from when answering a question in the entity's category.
For thirty years, this layer was a guess. A 2015 media planner said tier-one was The Wall Street Journal and The New York Times and you believed it because there was no way to check. Now there is. AI engines are observably trained on, and pull from, publications — and which ones can be measured query by query.
The list is not what most teams think. The trade publication being ignored is doing more citation work than the consumer book on the wall. The interview that ran in 2019 may still be the dominant source in an engine's answer in 2026.
The work at Layer 3: pitch into the publications the engines actually pull from in the entity's category. Not the publications the media plan from three years ago said were tier-one. Different list. Different pitch. Different success metric.
What gets it wrong: confusing impressions with citations. A piece in a publication the engines don't pull from is good for ego. It is not good for Citation Share.
Layer 4 — Generative Engine Optimization (GEO)
The content engineering. Writing and structuring content so AI engines can read it, understand it, and quote it cleanly.
GEO is what SEO was fifteen years ago: a technical practice that becomes a strategic one once leadership understands the metric. Most of what is being marketed as "AI Communications" today is actually just GEO. GEO is a layer in the stack — not the whole stack — but it is a critical one.
The work at Layer 4: question-shaped headlines. Clear definitions of who and what the entity is. Internal links that connect related concepts. Content that reads well to humans and parses cleanly for machines.
What gets it wrong: writing for humans only, or writing for machines only. The piece that lands in both is the piece that gets quoted.
Layer 5 — Amplification
Influencer content. Creator content. Podcasts. YouTube. Video, audio, and image content increasingly pulled by AI engines — especially in consumer categories like beauty, fashion, food, travel, and wellness.
This is where AI Communications stops looking like classic PR and starts looking like something newer. The 2026 source map for a beauty brand includes TikTok creators the brand has never heard of. The 2026 source map for a B2B SaaS company includes podcast episodes nobody on the comms team listened to.
The work at Layer 5: identify the creators and platforms the engines actually pull from in the entity's category. Engage them strategically. Measure the citations that follow.
What gets it wrong: treating influencer work as a separate marketing function disconnected from communications. The engines do not separate them. Neither should the team.
Layer 6 — Measurement
Citation Share. AI-visibility audits. Source mapping. Attribution research.
This is the layer that makes everything else accountable. Without measurement, every other layer is a faith-based investment. With measurement, every other layer becomes a business decision.
Citation Share is the headline metric — the percentage of relevant AI answers in a category in which the entity is named, quoted, or linked. Source mapping is the supporting analysis — which sources the engines pull from for which queries. Together they tell a comms team where to invest and where to stop.
The work at Layer 6: a recurring audit, run weekly or biweekly, across the five major engines, across the queries that matter for the entity's category. Score Citation Share. Map the sources. Identify which layers below are working and which are not.
What gets it wrong: measuring vanity numbers — total mentions, total citations, total surface area — instead of category-specific Citation Share against named competitors. The right number is comparative, not absolute.
Layer 7 — Defense
Crisis response inside AI answers.
When the engine returns a bad answer — outdated facts, surfaced controversy, competitor framing — the response cannot be a press release alone. It has to reach into the sources the engine is pulling from and change them. New earned media. New structured authority. New primary research. Layered carefully so the engine updates its answer.
This is the newest layer and the most underdeveloped. Most communications teams still respond to AI-answer problems with the playbook for responding to journalist problems. The playbooks are not the same.
The work at Layer 7: a defined incident-response process. Monitoring across all five engines for changes in the entity's answer. A pre-built ability to deploy authority-rebuilding content at speed when an answer goes wrong.
What gets it wrong: assuming a bad AI answer will fix itself if you ignore it. It will not. It will compound. Build the infrastructure before the crisis — not during it.
The principle of the stack
Every layer depends on every layer below it.
Earned media (Layer 3) cannot move Citation Share if the entity has no coherent Identity (Layer 1) for the press coverage to attach to.
GEO (Layer 4) cannot make content findable if the entity has no Authority (Layer 2) for the engine to trust.
Defense (Layer 7) cannot fix a bad answer if Measurement (Layer 6) is not detecting that the answer has gone bad.
This is why "investing in GEO" without addressing Identity and Authority almost always fails. It is why "doing AI PR" without Measurement is faith. It is why most of what is sold as a single-layer solution does not produce Citation Share gains.
The stack is the picture you can show leadership. The budget framework. The hiring framework. The audit framework. The single most useful internal document an AI Communications practice produces.
How to use the stack
Audit by layer. Where is the entity strong? Where is it weak? Most entities are strong at two or three layers and broken at four or five. The audit names them.
Budget by layer. Reallocate from tactic-based budgets — PR retainer, paid social, SEO agency — to layer-based investments. The layer producing the worst result gets the next dollar.
Hire by layer. The job descriptions change. A team that owns the full stack needs an Identity strategist, an Authority lead, a retrieval-tested media specialist, a GEO writer, an Amplification lead, a Measurement analyst, and a Defense protocol. Some are existing roles repositioned. Some are new.
What changes next
The stack will change. The engines will evolve. New layers may emerge. The map will need updating.
The point of the stack is not that the seven layers are eternal. The point is that the discipline finally has an architecture — and the architecture can be audited, budgeted, and built against.
Anyone working in AI Communications without a stack is improvising. Anyone selling a single layer as the whole stack is selling SEO with a new label. Anyone using the stack is doing the discipline.
Ronn Torossian is the founder and chairman of 5W AI Communications, the AI Communications Firm. He is the publisher of Everything-PR and the author of two best-selling marketing books, including For Immediate Release. The canonical framework is at everything-pr.com/ai-communications.
