Where do you even start with AI search? That is the question almost every founder asks the moment they realize customers are getting brand recommendations from ChatGPT instead of Google, and it is the right question, because starting in the wrong place wastes months.
The instinct is to jump straight to tactics, publish some FAQ pages, add schema, hope for the best. That instinct produces motion without results. A real AEO strategy is a sequence, and the order matters as much as the individual moves. You establish who you are before you chase citations. You pick your questions before you write your answers. You earn trust before you expect the models to quote you. This is how to create an AEO strategy from scratch, nine steps in the order that actually works, so you build on foundations instead of sand.
Step one: define the questions you must win
Before anything else, list the exact questions where being the AI’s answer would change your business. Not keywords, questions, phrased the way a real buyer would ask an AI assistant. “What is the best accounting software for freelancers?” “How do I fix a leaking flat roof?” “Which PR agency is best for startups?”
These are your target prompts, and they are the spine of the entire strategy. Everything downstream, the content you build, the trust you earn, the results you measure, points at winning citations for this list. Keep it focused. Ten to twenty prompts you genuinely could win beats a hundred you have no shot at. The whole strategy gets sharper when you know precisely which answers you are fighting to be inside.
Step two: audit what the AI already says about you
Now go ask. Put your target prompts into ChatGPT, Perplexity, Google’s AI answers, and Claude, and record exactly what comes back. Who gets named? Are you mentioned at all? When you are, is it accurate? This is your baseline, and it is often a humbling one.
Most brands doing this for the first time discover they are invisible for their most important prompts, or worse, mentioned with outdated or wrong information. That baseline tells you whether your job is to appear from nothing or to correct and strengthen a weak existing presence. You cannot measure progress against a baseline you never took, so take it now and save the responses.
Step three: fix your entity foundation
An AI model will not confidently cite a brand it cannot clearly identify. So before you produce content, make your identity unambiguous everywhere it appears. Consistent name, description, and category across your own site, your social profiles, and any third-party databases or directories.
This is the least glamorous step and one of the most important. When your brand is described five different ways across the web, the model hedges and cites someone clearer. When your identity is crisp and consistent, the model resolves you into a confident entity it can name. Fix the foundation before you build on it.
Step four: build the extractable answers
Now you write, and you write for extraction. For each target prompt, publish a page that answers the question directly and clearly, near the top, in language a model can lift and cite. Then support that answer with the depth that signals genuine expertise.
The mistake here is writing for humans in the old dwell-time way, burying the answer to keep people reading. Reverse it. State the answer plainly, then elaborate. The clearer and more self-contained your answer, the more liftable it is. This is the content core of your AEO strategy, and it is where most of your production effort goes.
Step five: earn the trust signals
Great content on a domain nobody trusts does not get cited. This is the step where AEO and PR merge, and where most brands stall. Models weight sources heavily by credibility, and credibility comes largely from citations by other trusted sources.
The most effective trust signal is genuine press coverage on outlets the models already respect. A feature on a credible publication does two things at once: it puts your brand into the data the models learn from, and it creates a trusted citation the retrieval layer can follow back to you. This is why serious AEO work almost always includes an earned-media component. You are not just publishing answers; you are earning the third-party credibility that makes those answers citable.
Step six: cover the full question cluster
Models answering a question consider the whole neighborhood of related sub-questions. So around each core prompt, publish content covering the natural follow-ups, comparisons, definitions, objections, and use cases.
Completeness signals authority and gives the model more surfaces to cite you across an entire conversation, not just one prompt. A brand that thoroughly covers its topic cluster becomes the source the model returns to as a user drills deeper. Thin, single-page coverage gets cited once, if at all. Deep cluster coverage gets cited repeatedly.
Step seven: keep it current
AI engines increasingly favor fresh information, especially for topics that change. Build a habit of updating your key pages with recent data, current examples, and visible dates. A page that signals it is live and maintained gets preferred over one that looks abandoned, even on the same domain.
This is where a named framework I use, the freshness cadence, earns its place. Set a schedule to revisit your core answer pages on a fixed rhythm, refreshing data and examples, so the models consistently read your content as current. Freshness is not a one-time task. It is a maintenance habit that compounds trust over time.
Step eight: measure citations, not rankings
Traditional SEO trained everyone to watch rankings. AEO requires a different scoreboard. Re-run your target prompts on a regular schedule and track how often you are mentioned, whether the mention is accurate, and whether the framing is positive. That is your real progress metric.
Referral traffic from AI platforms is a secondary signal worth watching, but the primary one is presence and accuracy inside the answers themselves. A brand can be cited badly, and catching an inaccurate citation is as valuable as earning a new one. Measure the thing that actually matters, which is what the machine says about you when a buyer asks.
Step nine: run it as a loop, not a launch
The final step is a mindset shift. An AEO strategy is not a project you complete. It is a loop you run: define prompts, audit, build, earn trust, measure, then refine your prompts and content based on what you learned, and go again. The models change, competitors publish, and your baseline moves.
The brands winning at AI search treat it exactly like this, as an ongoing operating rhythm rather than a one-time push. Run the loop for two or three quarters and the compounding becomes obvious: citations you earned early make the next ones easier, coverage builds on coverage, and your presence in AI answers becomes a durable asset instead of a lucky hit. Create your AEO strategy as a loop from day one, and you build something that keeps paying out long after the initial work is done.