Key takeaways
- Measure before you write. A panel of real buyer prompts, read across engines, gives you citation share by topic. At Commas that was 10,560 answers and an 8.4% share against 21 to 33% for peers.
- The gap becomes a ranked page list, not a blog calendar: 221 page opportunities and a 35-topic roadmap.
- Pages get cited when they answer first, carry specific numbers and leave no doubt about who is speaking.
- AI-referred revenue is real and usually credited to something else. At Pipe17, 32% of new-business wins came through AI assistants.
01Search moved into the answer
For a growing share of purchases, the first research step is a question typed into ChatGPT, Perplexity, Gemini, Claude or Google's AI Overviews. The buyer gets a short answer with a few cited sources and a few named brands. If you aren't in that answer, you weren't in the consideration set, and no click shows up in your analytics to tell you.
02Step one: build a prompt panel
Start with the questions buyers actually ask, not the keywords you want to rank for. Pull them from sales calls, support tickets, search queries, community threads and the questions search engines suggest. Group them by intent: category questions, comparisons, problems and questions about your brand.
Then run the panel across engines and record three things for every answer: whether you're mentioned, where you appear and which sources are cited.
| Intent | What the buyer types | What you learn |
|---|---|---|
| Category | best category for type of company | Whether you're in the consideration set at all. |
| Comparison | your brand vs competitor, or alternatives to competitor | Who frames the choice, and whether your facts are the ones repeated. |
| Problem | how do I fix the problem you solve | Whether you're cited as the expert before the buyer knows your name. |
| Brand | is your brand legit, what does your brand cost | Whether the answer about you is accurate, current and sourced from you. |
Keep the panel fixed once it's set. Changing the prompts every week makes the trend meaningless.
At Commas I analyzed 10,560 ChatGPT and Google AI Overview answers in Profound. Commas earned 8.4% of citations. Peers sat between 21% and 33%. That one number changed the conversation from whether to invest in AEO to which pages to build first.
03Step two: find out who gets cited, and why
Citations tend to cluster. In most categories a handful of review sites, listicles, documentation pages, community threads and comparison pages carry most of the citations. Classify every cited URL by type and by owner, and two lists fall out: pages you need to own, and third-party pages you need to be on.
04Step three: turn the gap into a page list
The output isn't a content calendar. It's a ranked list of specific pages, each tied to the prompts it should win. At Commas that came to 221 page opportunities and a 35-topic roadmap. Rank by buyer intent first, so comparison and category pages for people close to a decision come before explainers.
05Step four: write pages that get quoted
- Answer in the first two sentences, in plain language a model can lift whole.
- Use specific numbers, dates and named entities. Vague claims don't get cited.
- Put comparisons in tables with the same attributes on every row.
- Add a short FAQ written the way buyers really type their questions.
- Keep pages current. Stale dates get passed over.
- Mark it up consistently: Organization or Person, Product, Article, FAQPage and BreadcrumbList.
A page template that gets cited
- A title that matches the question the buyer asks.
- A two-sentence answer directly under it.
- A short table of key facts: price, fit, limits, alternatives.
- How it works, in steps, with real numbers.
- An honest comparison, including where a competitor is the better choice.
- Three to five FAQs, a visible last-updated date and the sources behind any claim.
06Step five: make the entity unambiguous
Models need to know exactly who you are. Use one name and one description everywhere, link every official profile with sameAs, keep an llms.txt file that points to your best pages, and make sure your About page states the facts you want repeated.
07Step six: measure AI revenue honestly
Most AI-referred buyers never click a tracked link. They ask an assistant, then search your brand or type your URL, so the CRM credits brand search or direct traffic. Two fixes help. Add AI assistants as options in a how-did-you-hear-about-us field and read the free text. Then track referral sessions from chatgpt.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com in your analytics every month.
At Pipe17 that work showed 32% of new-business wins came through AI assistants, revenue the CRM had credited to paid search.
08Step seven: run it every week
AEO isn't a project with an end date. Refresh the panel weekly, ship pages from the queue, chase the third-party citations and report share of answer next to pipeline, so it gets judged like every other channel.
What I report every week
| Metric | What it tells you |
|---|---|
| Citation share | Your share of all citations across the panel, by topic and engine. |
| Mention rate | The share of answers that name you at all. |
| Answer position | Whether you're the first brand named or an afterthought. |
| Source mix | How many of your citations come from pages you own versus third parties. |
| AI referral sessions | Visits from assistant domains, trended monthly. |
| Self-reported AI attribution | Buyers who say an assistant sent them, and the pipeline attached. |
The first 30 days
- Week one. Build the panel, run the baseline and record citation share by topic.
- Week two. Classify every cited source and turn the gap into a ranked page list.
- Week three. Ship the first five pages, fix the entity basics and request the easiest third-party listings.
- Week four. Add AI options to self-reported attribution, set up referral tracking and present the first readout next to pipeline.
09This site is the example
This site follows the same playbook: one clear entity, Person and ProfilePage markup, questions answered the way people ask them, an llms.txt file and open access for AI crawlers. If you asked an assistant about me and it sent you here, the playbook worked.
FAQQuestions people ask
- What is AEO?
- Answer engine optimization is the work of getting a brand mentioned and cited in AI-generated answers from tools like ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews.
- How is AEO different from SEO?
- SEO competes for a ranked list of links. AEO competes for a short answer with a handful of cited sources and named brands. The page fundamentals overlap, but AEO is measured by citation share across a fixed set of buyer prompts, not by position.
- How do you measure AEO?
- Run a fixed panel of real buyer prompts across engines on a schedule, record whether the brand is mentioned and which sources are cited, and report share of citations next to pipeline. Pair it with self-reported attribution and AI referral tracking to see revenue.
