Most teams discover their AI search visibility the same way: a senior leader opens ChatGPT, types in a question about their category, and checks whether the company name appears. If it does, they breathe a sigh of relief. If it doesn’t, someone gets a Slack message.
That’s not an audit. That’s a spot check — and it tells you almost nothing useful.
A real AEO audit produces three things: opportunities (attributes where your visibility is low and improvable), objections (negative narratives that AI models repeat about you), and strengths (what’s already working that you can build on). Without all three, you’re flying blind.
Two surprises come up in nearly every first audit. First, most companies are significantly less visible than they assumed. Second, they often show up in categories or contexts they never intentionally targeted — sometimes in ways that don’t serve them. Both findings are useful. Neither shows up in a one-prompt check.
Here’s how to run the real thing.
Step 1: Discovery Visibility Audit
Start at the top level. What is your overall visibility rank and visibility score across the AI platforms you care about? These two numbers together tell you how prominently AI models feature you when buyers are exploring your category — and how consistently they do it.
Once you have the headline numbers, break them down attribute by attribute. Visibility isn’t monolithic. You might rank well for one capability and be nearly invisible for another, even if both are core to what you do.
Reading the Signal: A 2×2 Framework
When you look at rank and score together for each attribute, four patterns emerge:
- High rank + high score — The AI has a settled, confident view of you here. It surfaces you consistently and with conviction. This is a strength to protect and reference in your positioning.
- High rank + low score — You appear, but the signal is mixed. The AI mentions you sometimes, but the answers vary. This usually means inconsistent or thin content around this attribute.
- Low rank + high score — You have niche authority. When this topic comes up in a narrow context, you’re the answer — but the AI doesn’t connect it to broader discovery prompts. Expand the surface area.
- Low rank + low score — A comprehension gap. The AI barely registers you in this area. This isn’t a trust problem; it’s a foundational content problem. The model doesn’t have enough signal to form a view.
Watch for Category Label Variance
One of the most common audit surprises is how much visibility can vary based on the exact label used for the same capability. “Revenue intelligence” and “sales analytics” might describe the same product feature, but AI models can have dramatically different views of who leads each one. Run your audit across the labels your buyers actually use — not just the ones in your internal taxonomy.
Step 2: Citation Analysis
Visibility scores tell you whether AI models surface you. Citation analysis tells you why — and why not.
For this step, scope your analysis to discovery prompts only. These are the broad, category-level questions buyers ask early in their research. What pages and domains are feeding AI answers in your category?
Citation data answers three questions that visibility scores can’t:
- What content types are shaping your category? If third-party review sites and analyst reports dominate citations, that tells you where AI models are going to learn about your space — and where you need a presence.
- Which of your pages get cited, and for which attributes? You may find that one blog post is doing heavy lifting across multiple prompts, while your core product pages are never cited. That’s a structural problem.
- Where do competitors get cited that you don’t? These are your citation gaps — specific content opportunities where a competitor has established a foothold and you haven’t.
Citation gaps are often more actionable than visibility gaps. They point to specific content types, specific domains, and specific topics where you can make targeted investments.
Step 3: Validation Prompts as a Diagnostic
Discovery prompts tell you whether AI models surface you when buyers are exploring. Validation prompts tell you something different: what happens when a buyer already knows your name and asks the AI to evaluate you?
The combination of these two signals is a diagnostic tool.
Low discovery + low validation points to a comprehension problem. The AI doesn’t have a clear model of what you do, who you serve, or what category you belong to. It can’t recommend you because it doesn’t understand you. The fix is foundational: clearer, more consistent content that establishes your category, your use cases, and your differentiation.
Low discovery + strong validation points to a trust problem. The AI understands what you do — it just doesn’t reach for you when a buyer is exploring options. This is a different problem with a different solution. You need more presence in the third-party sources AI models treat as authoritative: reviews, analyst coverage, industry publications, and community discussions.
Don’t treat these two patterns the same way. Throwing more content at a trust problem won’t fix it. Chasing citations when you have a comprehension problem won’t fix that either. The diagnostic step is what lets you act on the right lever.
Step 4: Competitor Head-to-Heads
Once you understand your own position, you need to understand it relative to the alternatives buyers are being shown.
Run direct head-to-head comparisons across platforms — not just on one. When you ask an AI model to compare you to a specific competitor, the answer often varies significantly between ChatGPT, Gemini, Perplexity, and Claude. Those differences aren’t random. They reflect different training data, different citation sources, and different levels of content coverage on each platform.
Platform-level differences point to specific gaps. If you lose a head-to-head on one platform but win it on another, look at what content is indexed and cited differently. That’s your roadmap.
Examine the Buying Framework
Beyond individual comparisons, pay attention to how AI models frame the buying decision itself. What criteria does the AI use when it evaluates options in your category? What questions does it tell buyers to ask?
If the buying framework AI constructs naturally favors your strengths, that’s a significant advantage — and one worth reinforcing. If it consistently emphasizes dimensions where competitors are stronger, that’s a positioning problem that no amount of content optimization will fully solve. You may need to actively shape how the category is defined.
Step 5: Agent Analytics
AI visibility isn’t only about what models say — it’s also about what they do. As AI agents become more capable of browsing, researching, and acting on behalf of users, your site’s relationship with AI crawlers becomes a measurable signal.
Three metrics matter here:
- Fetchable visits — Bot visits where the AI crawler successfully retrieved your page. This is the baseline: can AI agents access your content at all?
- Chosen visits — Cases where the AI crawler fetched your page and then cited or used it in a response. This is the signal that your content is not just accessible but useful.
- Extractable content — Whether the content on your pages is structured in a way that AI models can parse and use. Pages that are fetchable but not extractable are a wasted opportunity.
Pair these bot metrics with your AI referral traffic — visits that arrive from AI platforms directly. Together, these numbers tell you whether AI models are not just mentioning you but actively sending buyers your way. A high visibility score with low AI referral traffic suggests your mentions aren’t driving action. That’s worth investigating.
Sharing Audit Results Across Teams
An AEO audit isn’t just a marketing deliverable. The findings touch every team that shapes how your company is perceived and described — and each team has a specific role to play in acting on them.
- Customer Success — Objections and recurring negative narratives in AI responses often mirror what buyers raise in sales conversations. CS teams can validate whether AI-surfaced objections match real-world friction and help prioritize which ones to address first.
- Product — Comprehension gaps frequently trace back to how the product is described, categorized, and positioned. If AI models don’t understand what you do, product teams need to be part of the conversation about how capabilities are named and framed.
- Product Marketing — PMM owns attribute coverage and positioning. The audit tells them which attributes are well-covered, which are thin, and where competitors have established stronger narratives.
- Brand — Narrative consistency across AI platforms is a brand problem as much as a content problem. If different platforms tell different stories about you, brand needs to understand why and help establish a more consistent signal.
- Demand Generation — Citation gaps and AI referral traffic are demand gen metrics. Where are buyers being sent that isn’t you? What content investments would close those gaps?
Share the audit findings in terms each team can act on — not just as a visibility score, but as a set of specific, prioritized problems with clear owners.
The Audit Is a Baseline, Not a One-Time Event
Everything above gives you a starting point. That’s exactly what it is — a starting point.
AI models update continuously. New content gets indexed. Competitors publish. Analyst reports shift. The buying frameworks AI constructs evolve as the category matures. A visibility score that’s accurate today may look very different in 90 days.
The value of running a rigorous first audit isn’t just the findings — it’s establishing the baseline that makes future measurement meaningful. You can’t track improvement without knowing where you started. You can’t prioritize investments without understanding where the gaps are.
Knowing where you stand in AI search is the first step to improving it systematically. Now you know how to find out.