Aeo

AEO (Answer Engine Optimization)

What this skill does

A growing share of the buyer journey now happens inside AI assistants — ChatGPT, Claude, Perplexity, Google AI Overviews, Gemini, Copilot, and the rest. Buyers use these systems to research, compare, and shortlist vendors, making generated answers a meaningful discovery surface alongside traditional search.

For PMMs, this is a category shift. The buyer surface you compete on is changing from search-engine results pages to AI-generated answers — where the "page" is one synthesized response with maybe three citations and a recommendation.

This skill helps PMMs do the strategic work of AEO: figuring out where they stand in AI answers today, why, and what to do about it. It does not generate content or scrape AI assistants. It tells you what work to do, how to prioritize it, and how to measure it.

When to invoke

• The user is auditing how their brand appears in AI assistants for the first time.
• The user just learned a competitor is cited more than they are and wants to fix it.
• The user is briefing content/SEO/PR teams and needs an AEO-specific brief.
• The user is choosing between AEO platforms and needs a tool-selection framework.
• The user is setting up measurement for AI visibility and needs to know what metrics matter.
• The user is closing the loop between AEO signals and positioning, messaging, or PR decisions.

Prerequisites

1. Read pmm-context.md first when available. If it is missing, offer to run pmm-context. If unavailable or declined, collect a minimum brief covering product, category, ICP, competitors, buyer questions, geographies, proof points, and claim boundaries. Mark assumptions and continue.
2. Ask the user about their starting point:
   • First-ever audit, or do they have a baseline?
   • Which AI assistants matter most to their ICP? (B2B usually = ChatGPT, Claude, Perplexity, AI Overviews. eCommerce adds Rufus, voice assistants.)
   • Do they have an AEO platform already (Profound, Peec, Otterly, Majentics, Goodie, etc.) or are they doing this manually?

The strategic AEO workflow

                Audit         →  Identify       →  Diagnose       →  Prioritize     →  Brief          →  Measure
                ↓                ↓                  ↓                  ↓                  ↓                  ↓
                What do AIs    Where are we      Why is the gap     What fixes         Hand off to       What changes
                say about us   absent / weak     there?             matter most?       content, SEO,     over time, when
                today?         / wrong?                                                PR, product       to act?

Step 1 — Audit current exposure

Goal: understand what AI assistants currently say about your brand, your category, and your competitors for the queries your ICP actually runs.

Design the query set

This is the most-skipped, most-important step. Bad query set → useless audit. Build the query set from:

• Buyer JTBD queries. What questions does your buyer actually ask when they have the problem you solve? (Pull from customer-research outputs if available.) Examples: "best AEO tool for B2B SaaS", "how do I track ChatGPT citations", "alternatives to [your category leader]".
• Category definition queries. "What is AEO?", "How is AEO different from SEO?", "Tools to monitor AI visibility".
• Comparison queries. "[You] vs [Competitor]", "alternatives to [you]", "best [your category] for [persona]".
• Branded queries. "What does [your brand] do?", "Is [your brand] worth it?", "[Brand] reviews".
• Negative queries. "Problems with [your brand]", "[Brand] complaints", "Why people leave [brand]". These reveal what AI assistants have ingested from negative-leaning sources.

Aim for 20-50 queries in v1, grouped by persona and buying stage. More is not better — synthesis bandwidth is the constraint.

Run the queries

Two paths:

• Manual (free, slow, painful): paste each query into ChatGPT, Claude, Perplexity, AI Overviews, Gemini, Copilot one at a time. Capture the full answer, citations, and order of mention in a spreadsheet. Repeat at least 3x per query — answers vary.
• Tooling (paid, fast): an AEO platform (see TOOL_LANDSCAPE.md) runs the query set across assistants on a schedule, captures responses, extracts citations, scores sentiment, and tracks change over time.

For a one-time audit, manual works. For ongoing monitoring, tooling is essentially mandatory — answers change too often to track by hand.

Capture for each query

• Exact query text and query-set version
• Assistant, model or experience, and access method
• Date/time, geography, language, and logged-in/personalization state
• Run number and whether the conversation was fresh or continued
• Full response or durable capture, including citations
• Was the brand mentioned at all? (Mention rate)
• What position? (1st, 2nd, 3rd, "also-ran", absent)
• Was the brand cited as a source? (Citation)
• What's the sentiment of the mention? (Positive / Neutral / Negative)
• What other brands appeared? (Competitive context)
• What sources did the AI cite? (URLs ingested)

Step 2 — Identify gaps

Cluster the audit results into gap types:

Gap type	What it looks like	Root cause family
Absence	Not mentioned at all for relevant queries	Content gap, citation gap
Misrepresentation	Mentioned but wrong (outdated, wrong category, wrong features)	Stale sources, weak entity signals
Under-positioning	Mentioned 3rd or 4th when you should be 1st	Authority gap, share-of-voice gap
Negative framing	Mentioned but with negative sentiment	Reputation issue, review/PR problem
Wrong audience association	Cited for the wrong persona or use case	Positioning leakage in indexed content
Competitor over-citation	Competitor cited disproportionately	Authority gap, link/citation gap

Map every query result to one or more gap types. Tag with severity (Critical / High / Medium / Low) based on:

• How close the query is to the buying decision
• How much pipeline rides on the persona who runs that query
• How fixable the underlying cause is

Step 3 — Diagnose causes

For each gap, ask: what causes AI assistants to answer this way? Common root causes:

Root cause	Diagnostic question	Typical fix owner
Content gap	Do we have a definitive page on this topic?	Content team
Citation/authority gap	Do trusted third parties (analysts, press, communities) talk about us?	PR, partnerships
Structured data gap	Is our content marked up so AIs can extract it?	SEO/engineering
Stale information	Is the indexed content out of date?	Content, ops
Wrong category framing	Are we described in the wrong category?	PMM (positioning issue)
Weak entity signals	Does the AI even know we exist as a distinct entity?	SEO, brand
Reputation surface	Are reviews/discussions skewing negative?	Customer marketing
Comparison gap	No "you vs competitor" pages exist	Content, PMM

Diagnosis is the PMM's highest-value contribution. AI platforms can show you the what; the PMM (with this skill) figures out the why.

Step 4 — Prioritize fixes

Not every gap deserves a fix. Use a 2x2 to triage:

              High pipeline impact
                    │
       Fix now      │      Plan
       (quick wins) │      (strategic)
                    │
   ─────────────────┼─────────────────
   Low fix effort   │   High fix effort
                    │
       Backlog      │      Defer
       (when slack) │      (or skip)
                    │
              Low pipeline impact

Map each gap to the matrix. Generally:

• Critical-severity Absence and Misrepresentation gaps go in "Fix now" — these are losing you deals today.
• Reputation surface issues are usually "Plan" — they require sustained effort.
• Wrong category framing is "Plan" — it's a positioning project (invoke positioning skill).
• Comparison gaps are often "Fix now" — they're high-impact and tractable.

Step 5 — Brief execution teams

The PMM's job ends when execution teams have a clear, prioritized brief. AEO work fans out across:

• Content team: definitive pages, FAQ pages, comparison pages, original research, data studies
• SEO/engineering team: structured data, schema markup, entity signals, crawlability for AI bots
• PR team: earned media in publications AI assistants trust, analyst briefings, executive thought leadership
• Customer marketing: review collection on the platforms that get cited (G2, TrustRadius, etc.)
• Product/engineering: in-product changes that change what's true about the product, not just what's said

See AUDIT_TEMPLATE.md for the brief format that each team should receive.

Step 6 — Measure over time

AEO is not a one-time exercise. Answers change weekly as models update, sources get re-indexed, competitors publish. Set up measurement against the standard AEO metrics (full definitions in METRICS_REFERENCE.md):

• Mention rate — % of queries where you're mentioned
• Share of voice — your mentions vs. competitor mentions
• Citation rate — % of queries where you're cited as a source
• Citation share — your citations vs. competitor citations
• Sentiment score — positive/negative skew of mentions
• Average position — where in the answer you appear (1st mentioned, 2nd, etc.)
• First-mention rate — % of queries where you're the first brand named
• Answer count — total volume of answers analyzed

Set a baseline. Re-run the same query-set version under comparable conditions on a stated cadence. Separate real visibility change from methodology drift when models, interfaces, geography, personalization, or query wording change. Define alert thresholds only after observing normal baseline variance.

Step 7 — Close the loop with positioning

AEO data is one of the strongest signals for whether positioning is working. Specific feedback loops:

• If AI assistants put you in the wrong category → positioning problem, invoke positioning skill
• If your unique attributes never appear in AI summaries of you → messaging/content gap, content fix
• If buyers' AI queries don't match your persona model → persona drift, invoke personas skill
• If a new competitor is suddenly co-cited with you → competitive shift, invoke adaptive-messaging skill

AEO without these loops is a vanity dashboard. With them, it's a positioning radar.

Outputs

A. AEO audit report

• Query set (with persona/stage tags)
• Per-query results (mention, position, sentiment, citations, competitors)
• Gap inventory (by type and severity)
• Diagnostic notes (root cause per gap)

B. Prioritized fix list

2x2 matrix with each gap mapped. Top 5-10 "fix now" items with owners and SLAs.

C. Team briefs

Per-team briefs (content, SEO/eng, PR, customer marketing) using the template in AUDIT_TEMPLATE.md. Each brief has specific gaps, specific outputs requested, specific deadlines.

D. Measurement dashboard spec

What metrics to track, what cadence, what alert thresholds, who reviews. If using an AEO platform, mapping of metrics to platform dashboards.

E. Repositioning triggers (optional)

Signals from AEO data that should trigger broader positioning, messaging, or persona work. Connects to the other skills in this repo.

Quality bar

• Evidence per gap. Every gap claim has supporting query results — the actual AI output that demonstrates the issue. Screenshots or transcripts, not summaries.
• Root cause before recommendation. Don't recommend "write more content" without diagnosing why content is the right fix. Sometimes it's PR, sometimes it's product, sometimes it's positioning.
• Quantify pipeline impact. "We're absent from 12 queries" is useless. "We're absent from 12 queries that 40% of our ICP runs in the consideration stage" is decision-grade.
• Tool-neutral framing. Describe the work, name tool categories. Don't anchor the workflow to any single platform.
• Refresh cadence stated. Every audit ends with a "when do we re-audit" answer.

Anti-patterns to refuse

• Running an audit without either shared context or a task-specific brand, buyer, category, and competitor brief
• Generating fix recommendations without root-cause diagnosis
• Treating AEO as a content-team-only problem (it's cross-functional)
• Reporting raw mention counts without competitive context (share of voice matters more than absolute counts)
• One-time audits with no measurement loop (AEO is a discipline, not a project)
• Recommending tools the user didn't ask about (this skill is workflow, not procurement)