AEO buyer journey is the path a prospect takes through AI answers, from a broad question to a specific vendor choice, and the work of mapping which prompts appear at each step. Instead of treating every query as a ranking target, you sort the questions buyers ask a model by funnel stage, then check whether your brand is mentioned, cited, or recommended for each one. That tells you where visibility actually affects pipeline.
Most AEO content stops at explaining why this matters. This piece is about the working method. You'll build a prompt inventory from real buyer questions, tag each prompt by stage, decide what counts as success at each stage, and connect those prompts to pages you already own. It also covers where the method breaks. Prompt sampling is noisy, and the same model can give different answers on different runs, so treat every number here as a directional signal, not a precise measurement.
Why the Classic Funnel Breaks When Buyers Ask a Model
The classic funnel assumes a buyer clicks through a series of pages: a blog post, a comparison page, a pricing page. Each page earns a visit, and you measure progression by traffic and conversion. An AI answer compresses that sequence. A buyer asks one question, reads a synthesized response, and may never open a website at all. Some of the funnel happens inside the answer.
That changes the unit of measurement. In SEO, a page either ranks or it doesn't. In AEO, the same brand can appear as a named option, as a linked source, or not at all, and each outcome has a different value depending on where the buyer is in their decision. If you want the broader distinction, AEO vs. SEO: what actually changes walks through the mechanics. The short version for this article: you can't map the journey by page anymore, so you have to map it by prompt.
Building a Prompt Inventory From Real Buyer Questions
Your prompt inventory is the list of questions a buyer might put to a model. Start with sources that reflect real language, not invented keywords. Good inputs include sales call transcripts, support tickets, G2 and Reddit threads in your category, the questions your reps answer repeatedly, and your site search logs. Add the search queries from Search Console that have a question or comparison shape.
This is the core of LLM prompt research. Write each prompt the way a buyer would type it into a chat window, which is usually longer and more specific than a search query. "best CRM" becomes "we're a 40-person B2B team on HubSpot, what should we use if we need better forecasting?" Keep the inventory to 50 to 150 prompts for a first pass. Anything larger becomes hard to review and re-run.
For prompt-level citation tracking, log each prompt as a row with its exact wording, the model and date you ran it, and a few columns for results. Run each prompt several times across ChatGPT, Claude, and Perplexity. Single runs mislead. A brand that appears in three of five runs is a different situation from one that appears once, and your tracking should record that spread rather than a yes or no.
Tagging Prompts Across Awareness, Consideration, and Decision
Once you have the inventory, tag each prompt with one of three stages. The AEO funnel stages map closely to the buyer's mental state, not to your sales pipeline labels.
Awareness prompts are problem-shaped. "Why do our sales forecasts keep missing?" or "how do teams track content performance across channels?" The buyer doesn't know the category name yet. These are the prompts where a model's explanation does most of the work.
Consideration prompts name a category or a set of options: "alternatives to [competitor]," "best tools for B2B attribution," "compare X and Y for a 50-person team." This is where shortlists form. Commercial intent prompts in ChatGPT and similar tools tend to return a list, so inclusion in that list is the main event.
Decision prompts are about fit and risk: "how much does [category] cost at this scale," "how long does implementation take," "does it integrate with our warehouse." The buyer is nearly choosing, and the answer often hinges on specifics that only the vendor's own documentation or pricing page can supply.
A practical tip: tag a prompt by the buyer's intent in the moment, not by the keyword's volume. Some high-volume awareness terms are actually consideration prompts when phrased in full. Read the prompt aloud and ask what the person is trying to decide.
Scoring Each Stage: Mention, Citation, and Recommendation
Each stage needs its own definition of a win, because the same outcome means different things. Use three levels:
- Mention: your brand or product name appears in the answer, without a link.
- Citation: the model links to or explicitly attributes a claim to your domain.
- Recommendation: the model suggests you as a fit, often with a reason.
Mentions are common and cheap. A brand can be named in a list it never gets linked from. That matters most at awareness and consideration, where being in the set is the goal. Citations matter more at decision, because a buyer checking pricing or integrations wants a source they can verify, and a citation carries the model's implicit endorsement of that source.
The gap between the first and second level is where most teams lose ground. Quantifying the gap from mention to citation is worth working through with your own data, since the pattern tends to show that brands get named before they get sourced. In practice, this means a mention-before-citation sequence: the model knows you exist from training or from third-party coverage, but it has no page on your site it considers quotable. Your fix is usually on-page, not in PR.
Score each prompt on the three levels, then average across runs. Report the average by stage, not just the overall number. A 40 percent mention rate at awareness and a 5 percent citation rate at decision tells a very different story than a blended 22 percent.
Mapping Prompts to Pages You Already Own
With prompts tagged and scored, the next question is which page should answer each one. This is where most teams find they already have content that could serve a prompt but doesn't serve it well. Map each high-value prompt to one owned URL. If no page fits, that's a content gap. If a page fits but isn't cited, that's an extraction problem.
Be honest about the second case. A page can rank well and still never be quoted. Ranking rewards relevance and authority across a query. Citation rewards a passage a model can lift cleanly: a direct answer, a specific figure, a clearly labeled definition. What makes a page citable versus just rankable covers that distinction in detail. When a decision-stage page has the right facts buried in a long paragraph, how to structure content for LLM extraction is the fix to make first.
For B2B SaaS AI search prompts, the pages that most often need this work are pricing pages without plain-language tier descriptions, integration pages without stated supported versions, and comparison pages that argue rather than describe. Those are the pages a model reaches for at decision time, so they deserve the clearest writing you have.
Where Each Stage Breaks and How to Prioritize Fixes
Each stage tends to fail in a different way, and the fixes differ too.
Awareness usually breaks on coverage. The model explains the problem without naming any vendor, or names competitors who wrote more about it. The fix is a clear, definitional page that answers the problem prompt directly, plus third-party mentions that establish your name in the category.
Consideration usually breaks on positioning. You're mentioned, but described vaguely or grouped with the wrong alternatives. Here the fix is comparison content that states your actual differentiators in plain terms, and consistent descriptions of your product across the sites that get cited.
Decision usually breaks on verifiability. The model either can't find a current source for your pricing or integrations, or it finds an outdated one. The fix is maintenance: keep decision pages current and make the answer extractable.
When you prioritize, weight by two things: the commercial value of the stage and the size of the gap. A decision-stage citation gap on a high-intent prompt usually beats a broad awareness mention gap, even if the awareness prompt has more volume. Start there, re-run the same prompts in a month, and compare. Because answers vary run to run, look for consistent shifts across several runs rather than a single change.
Frequently Asked Questions
What is the AEO buyer journey?
The AEO buyer journey is the sequence of questions a prospect asks an AI model as they move from recognizing a problem to choosing a vendor. It's mapped by prompt rather than by page, and each prompt is tagged as awareness, consideration, or decision. Tracking which prompts name your brand, cite your domain, or recommend you shows where AI answers influence your pipeline.
How do I build a prompt inventory?
Collect questions from sales calls, support tickets, Search Console queries with question or comparison phrasing, and category forums. Rewrite them as full, conversational prompts. Aim for 50 to 150 prompts in the first pass, and log the exact wording so you can re-run them later. Keep the list focused on questions a real buyer would ask.
Why do AI answers vary between runs?
Models sample responses, so the same prompt can produce different lists, citations, and wording across runs or over time. Models also update, and retrieval may pull different sources depending on the moment. For that reason, run each prompt several times and report the rate at which your brand appears, not a single yes or no.
Which AI engine should I track first?
Track the engines your buyers actually use. ChatGPT, Claude, and Perplexity handle sourcing differently, so a tactic that earns citations in one may not work in another. Perplexity tends to show sources more visibly, while other tools may name brands without linking. Measure each engine separately rather than assuming one result carries over.
Do I need citations at every funnel stage?
No. Mentions matter most at awareness and consideration, where being in the shortlist is the goal. Citations matter more at decision, when a buyer is checking pricing, integrations, or fit and wants a verifiable source. Score each stage on the level that matters there, and don't treat a missing citation on an awareness prompt as a failure.
Bottom Line
Mapping the AEO buyer journey means accepting that buyers now complete part of their research inside a model's answer. You can't see that journey through traffic alone, so build a prompt inventory from real questions, tag each prompt by stage, and score mentions, citations, and recommendations separately. Then connect each high-value prompt to a page you own and check whether that page is written to be quoted. Expect noise, run prompts repeatedly, and look for consistent movement. Start with decision-stage gaps, since they sit closest to revenue, and keep the rest of the funnel on a regular re-check schedule.