Three Frameworks That Make 10 AI SEO Agents Work
- The argument in three lines
- Why do frameworks matter for a 10-agent team?
- Framework 1 — Three Pillars (measurement)
- Framework 2 — The 7-Layer Operator Playbook (execution)
- Framework 3 — The 3-Legged GEO Stool (strategy)
- How do the three frameworks compose?
- What are the failure modes if you skip a framework?
- The closing question
The argument in three lines
- A team of ten specialized AI SEO agents only works if you can measure it, run it, and direct it.
- Three frameworks cover those three jobs: Three Pillars measures, the 7-Layer Operator Playbook runs, the 3-Legged GEO Stool directs.
- Without all three, the agents work in isolation and the program drifts.
Why do frameworks matter for a 10-agent team?
Most "AI SEO" rollouts fail in the same place. A team buys an AI tool. The tool produces output. The output is real work — but nobody can tell whether the work is moving the right number, nobody knows who owns which loop, and nobody can connect day-to-day activity to a strategy.
The failure mode is not the technology. The failure mode is the absence of three operating disciplines:
- Measurement — what number proves AI search visibility is changing?
- Execution — what is the daily, weekly, monthly cadence of work that compounds?
- Strategy — what are the load-bearing investments, in priority order, that earn citations at all?
Each of the three frameworks below answers one of those questions. Together, they make a ten-agent team coherent. Run any one without the other two and the program does not compound.
Three Pillars (the measurement layer)
The Three Pillars framework, introduced by Ethan Crump at Foundation Inc and now adopted broadly across enterprise GEO programs, identifies the three things you must measure if you want to know whether AI search visibility is improving.[1]
Pillar 1 — Visibility. Is the brand named in AI answers at all? Visibility is the entry threshold: an answer engine has to know the brand exists, in the right category, before it can cite it. Visibility is measured as raw mention counts and as a share-of-voice ratio (or Share-of-Model) against named competitors.
Pillar 2 — Citation. When the brand is mentioned, is it cited as a source — with a link or domain attribution — or is it mentioned as a passing reference? Citations are the durable currency of AI search; they drive referral traffic, knowledge-graph reinforcement, and the second-order signal that downstream models inherit.
Pillar 3 — Sentiment. When the brand appears, what is the surrounding context? Positive (recommended, ranked, named as a leader)? Neutral (listed in a category survey)? Negative (cited as a cautionary example, or contrasted against a recommended competitor)? Sentiment matters because share-of-voice without polarity is misleading — a high mention rate inside negative comparisons does not convert.
How the ten agents wire into Three Pillars
| Pillar | Owning agent | Daily output |
|---|---|---|
| Visibility | GEO Manager | Share-of-Model, prompt coverage, competitor delta |
| Citation | SEO AI Engineer + GEO Outreach Manager | Citation-graph entries, schema-driven entity authority, third-party mentions |
| Sentiment | Content Strategist + SEO Web Analyst | Narrative tone scan, comparative-list polarity, reputational drift signals |
Three Pillars is the dashboard the senior strategist reads to decide where the team's attention goes next. Without it, the agents' work has no scoreboard.
The 7-Layer Operator Playbook (the execution layer)
Three Pillars tells you what to measure. The 7-Layer Operator Playbook tells you what to actually do — every day, every week — to move those three numbers.
This is the 7-Layer Operator Playbook we use across every enterprise GEO program — built from the pattern that recurs in the field. The seven layers are not a sequence; they are concurrent loops, each owned by a specific agent.
Layer 1 — Discovery surfaces. Are the brand's pages discoverable by AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Google-Agent)? Discovery is binary. If the crawlers can't reach the page, none of the other six layers matter. Owning agent: Technical SEO Manager.
Layer 2 — Structured authority. Does each page carry the JSON-LD that lets answer engines bind the content to the right entity? Author schema, Organization schema, Article schema, FAQ schema, HowTo schema — each is a signal that the page is structured for machine consumption. Owning agent: SEO AI Engineer.
Layer 3 — Chunk-shaped content. Is the content written in passages that survive being chunked, embedded, and re-emitted by an answer engine? Short paragraphs, structured lists, named entities front-loaded, definitions in self-contained sentences. Owning agent: Content Engineer.
Layer 4 — Citation pattern. Does each page include a verifiable source pattern — inline citations, a sources section, dates, author attribution — that an answer engine can parse and reproduce? Research published in 2025 across 42,971 AI citations showed that structured, citation-bearing content was cited at a 91.3% higher rate than prose-only equivalents.[2] Owning agent: Content Engineer + SEO AI Engineer.
Layer 5 — Fan-out coverage. When a user prompt generates a fan-out of related sub-queries inside an answer engine, does the brand appear across enough of those sub-queries to be the recommended source? Studies across hundreds of thousands of AI search pages indicate that fan-out coverage, not single-query rank, is now the dominant predictor of cited share-of-voice.[3] Owning agent: GEO Manager + Content Strategist.
Layer 6 — Off-domain citation graph. When third-party publishers cite the brand, does the citation pattern reinforce the entity? Reviews, comparisons, listicles, podcasts, research papers — every off-domain mention either tightens or dilutes the citation graph. Owning agent: GEO Outreach Manager.
Layer 7 — Measurement feedback loop. Does the program ingest Share-of-Model, citation rate, sentiment, and crawl health daily, and route signals back to layers 1 through 6? Without a measurement loop, the layers do not adapt. Owning agent: SEO Web Analyst + SEO Manager.
Why seven layers
Five-year-old SEO ran on roughly three layers: technical health, content quality, link authority. The four additions — chunk shape, citation pattern, fan-out coverage, structured authority — are direct consequences of the answer-engine era. They are not optional. A program running on the old three-layer stack will produce content that humans read but answer engines do not cite.
The 3-Legged GEO Stool (the strategy layer)
Three Pillars measures. The 7-Layer Playbook executes. The 3-Legged GEO Stool tells you where to invest first — and what to defend when budgets get cut.
The framework: durable AI search visibility rests on three legs of equal weight, and the program collapses when any one is missing.
Leg 1 — Brand. Earned mentions, name recognition inside the category, narrative ownership of a specific position. Without a brand the answer engines do not know which entity to bind a citation to.
Leg 2 — Technical SEO. Crawlability, render quality, agent-readability, schema completeness, Core Web Vitals. Without technical health the pages do not survive ingestion by the crawlers that feed the answer engines.
Leg 3 — Content. Editorial breadth, narrative depth, structured chunks, citation pattern, persona coverage. Without content there is nothing for the answer engines to ingest in the first place.
The framework is called a stool, not a pyramid, deliberately. The three legs do not stack — they are co-equal. A brand-heavy program with no technical hygiene gets crawled inconsistently. A technical-heavy program with no content has no surface to cite. A content-heavy program with no brand gets mentioned but not bound to the entity.
The 3-Legged Stool and the ten agents
Each leg is owned by a subset of the agent team:
| Leg | Owning agents | Output |
|---|---|---|
| Brand | GEO Outreach Manager, GEO Manager, Content Strategist | Earned mentions, narrative position, third-party citation graph |
| Technical SEO | Technical SEO Manager, SEO Software Engineer, SEO AI Engineer | Crawl health, render integrity, schema, agent-readability |
| Content | Content Strategist, Content Engineer, SEO Web Analyst | Editorial calendar, chunk-shaped pages, performance feedback |
The Senior Strategist (the Forward-Deployed Enterprise Strategist on top of the team) rebalances investment across the three legs every quarter. When one leg weakens, the strategist redirects the team toward it before the program loses citation share.
Is your site agent-ready?
Score your domain against all three frameworks — measurement, execution, strategy. Get a 0–100 score and a stack-ranked priority list of what to fix first.
Take the audit →How do the three frameworks compose?
The three frameworks answer three different questions, and they compose:
- The 3-Legged Stool decides where to invest (which leg is undercapitalized this quarter).
- The 7-Layer Playbook decides what to do (which of the seven concurrent loops is producing or stalling).
- Three Pillars decides whether it worked (visibility, citation, sentiment moved in the right direction).
A senior strategist running a ten-agent team reads the Three Pillars dashboard every Monday, decides which of the 7 Layers to lean into for the week, and revisits the 3-Legged Stool every quarter to rebalance.
That is the operating system of a ten-agent AI SEO program. Not the agents alone — the agents plus the three frameworks that make them coherent.
What are the failure modes if you skip a framework?
A team buying generic "AI SEO" tooling without these three frameworks ends up in one of three predictable failure states:
- Measurement without execution — a dashboard that updates daily and a content calendar that produces nothing the dashboard moves.
- Execution without strategy — agents producing high volumes of content that the 3-Legged Stool would have de-prioritized in favor of an underfed leg.
- Strategy without measurement — a quarterly plan that cannot be validated because no one is reading Share-of-Model, citation rate, or sentiment in real time.
All three failure states are recoverable, but each one costs a quarter. Programs that ship with all three frameworks wired into the agent team from day one skip that quarter.
The closing question
If you are running enterprise SEO in 2026 and your team can answer all three of these questions, you are ahead of most of the market:
- Which Pillar moved this week, and by how much?
- Which of the 7 Layers produced that movement?
- Which leg of the Stool is underfunded next quarter?
If your team can answer none of them, the gap is not the tooling. The gap is the operating system. Tooling without an operating system is what every "AI SEO" pilot stalls inside.
Want to know if your current setup can run all three frameworks? Take the free Agent-Ready audit — 15 questions, 3 minutes, full report by email.
Vijay Vasudevan
Vijay Vasudevan is Founder of Indexable AI, the agentic SEO platform building AI search visibility for B2B SaaS. He previously led SEO at Uber and was the first SEO hire at Uber Eats. He has generated $1B+ in organic pipeline across Uber, Zendesk, RingCentral, and Williams-Sonoma Group.
Further reading
- Ethan Crump, Foundation Inc — GEO Metrics: How to Measure Visibility, Trust, and Brand Presence in AI Search
- Daniel Shashko — How Google Picks Which Sentences to Cite in AI Mode (Reverse-Engineering 42,971 Citations) (HackMD, March 2026). Structured content cited at 91.3% match rate vs 39.3% for prose.
- Fan-Out Effect research, 16,851 queries across 353,799 pages, 2025 (publicly available industry research, attribution withheld per editorial policy). Fan-out coverage is the dominant predictor of cited share-of-voice in answer engines.
- Indexable AI — The 3-Legged GEO Stool (execution framework)
Want a free AI visibility audit of your domain?
Drop your work email and website. We'll audit your domain against the three legs of AI search visibility — Brand, Technical SEO, Content — and ship a stack-ranked findings document within 24 hours. We do the work first. Talk later.
Built by professional SEOs in San Francisco · Your email goes to Vijay personally · No spam, no sharing