The 7-Step AEO Execution Plan

Core principles

• 01LLM answers = LLM + RAG summarizing many citations, so you must be mentioned as many times as possible, not just rank #1
• 02The 'head' is different: being cited most often beats holding the top blue link
• 03The 'tail' is larger than in search — people ask ~25-word, highly specific follow-up questions vs ~6-word Google queries
• 04Split work into on-site (answer every follow-up question) and off-site (show up in as many citations as possible)
• 05Never trust best practices — validate every tactic with your own reproducible experiment

How to run it

1. 1

   Identify the questions you want to rank for

   Take your search data, your money terms, and your competitors' paid-search terms, then paste those keywords into ChatGPT and ask it to turn them into questions. Also mine questions from sales calls, customer support, and Reddit to capture the specific long tail that never appears in search data.

   Pro tip Give the raw keywords to ChatGPT and say 'make these into questions' — it does a pretty good job.

   Watch out There is no truth-set of question volume like Google's ads API gives for keywords, so you must infer demand from proxy search volume and mined real questions.

2. 2

   Track them in an answer tracker

   Put your questions into an AEO/answer tracker to establish a baseline share of voice — how often you show up and your average rank across surfaces. Ask each question multiple times and in variant phrasings because answers are a weighted random sample and vary per run and per surface.

   Pro tip Pick the cheapest tracker that does what you need — there are ~60 and they're roughly all the same, like keyword trackers.

   Watch out A single query run is not enough; answers have real variance, so you need multiple runs and question variants to measure true frequency.

3. 3

   Map who is showing up as citations

   Identify which citations appear for your target questions and break them into groups: your own site, video (YouTube/Vimeo), UGC (Reddit/Quora), tier-1 affiliates (Dotdash Meredith), tier-2 affiliates, and blogs. Each group needs its own strategy.

   Pro tip The citation mix differs by business type — Tech Radar dominates B2B, Glamour/Cosmopolitan for commerce, Eater/Yelp/TripAdvisor for marketplaces.

4. 4

   Build your own landing pages that answer every follow-up

   Figure out which page type is showing up most (listicle, category page, article, tool page) and build your own version. To get it cited, answer all the follow-up questions and subtopics — the more questions you answer that others don't, the more likely you appear.

   Pro tip Group your keywords into themes/topics and make each page answer that whole cluster of questions, not one keyword.

   Watch out If you fail to answer a follow-up question that a competitor answers, you'll likely be excluded from that answer.

5. 5

   Win the off-site citations, group by group

   Run a distinct play per citation group: pay affiliates (Forbes) to mention you — expensive but easy and controllable; make YouTube/Vimeo videos — easy because no community gatekeeps them; and post authentic Reddit comments. Prioritize non-glamorous high-LTV B2B topics on video where almost no one is publishing.

   Pro tip Video is a huge untapped opportunity for boring high-value B2B terms like 'AI-powered payment processing API' — nobody makes those videos, so you win by default.

   Watch out Reddit is community-gatekept — you can't guarantee your comment survives, so don't rely on it alone.

6. 6

   Run a controlled experiment to see what actually worked

   Take ~200 questions: leave 100 untouched as a control group and intervene on the other 100 (comment on Reddit, publish video, pay an affiliate). Track a couple weeks before and after, compare the test group against the control, and keep only what moved the test group when the control didn't. Reproduce the win multiple times before trusting it.

   Pro tip A control group is essential because answers vary on their own and LLM traffic is rising generally, which would otherwise look like a false win.

   Watch out Most SEO/AEO 'best practices' are wrong because nobody ran analysis — a claim gets repeated until it's accepted. Reproducibility across multiple runs is what separates real effects from coincidence.

7. 7

   Assign the right team

   Point your existing SEO team/agency at the on-site AEO work, then add a community/generalist marketer for the off-site work — because most SEO people are not good at making YouTube videos or running Reddit strategy.

   Watch out Don't assume one person covers both; the off-site citation work is the hard-to-hire-for gap.

In the wild

Common mistakes

Is it for you?

From the transcript