Over the last year I built an AI named “Sarah”
Not because I wanted to be in the AI product business. Because I was doing AEO audits manually, and the most time-consuming part wasn’t the analysis — it was the systematic review of schema, content structure, and entity signals across dozens of pages. It was pattern recognition work. The kind of work AI is actually good at.
So I built Sarah.
Sarah is named after the telephone operator at the Mayberry exchange — that cheerful, warm voice who knew everyone in town and connected the right people to the right conversations. I wanted an audit tool that felt like it understood your site, not one that generated an impersonal wall of technical warnings.
Sarah ingests a site’s schema exports and content structure, runs analysis against AEO signal patterns, and produces a diagnostic report that reads like it was written by someone who actually read your site. Because in a meaningful sense, she did.
After running her on dozens of sites across industries, I can tell you the five things she finds almost everywhere.
1. The FAQ Schema Gap
On site after site, Sarah finds FAQ content with no FAQPage schema. The questions exist — sometimes on a dedicated FAQ page, sometimes embedded in service pages, sometimes inside blog posts. The structured data that would make those questions machine-readable to AI systems doesn’t.
This is the most common finding and the highest-impact fix. It’s also the most avoidable problem. FAQ schema is free, implementable without a developer, and validated with free tools. The barrier to implementation is awareness, not capability. Most sites just don’t know they need it.
2. Answer Blocks That Are 3x Too Long
The second most consistent finding is that pages are answering questions — they just take too long to get to the answer. The actual response to the implicit question the page promises to address is buried inside a paragraph that runs 80, 100, 150 words.
AI extraction systems are looking for concise, direct answer blocks. Under 40 words, ideally. When Sarah flags an answer block as too long, the fix is almost always surgical: pull the core answer sentence to the top of the section, then let the existing explanation follow. The content doesn’t change. The extractability changes dramatically.
3. Entity Inconsistency Across Platforms
Entity signals are how AI systems verify that the brand claiming to be an authority actually is who and what it says it is. Organization name, address, phone number, founding date, social profiles — these signals appear in schema, in Google Business Profile, in LinkedIn, in directory listings, in press mentions.
Sarah routinely finds inconsistencies across these signals. The brand name is formatted differently in three places. The address is abbreviated one way on the website and written out another way in GBP. The phone number format doesn’t match.
Individually these feel trivial. Collectively they create entity uncertainty — and AI systems don’t cite sources they’re uncertain about.
4. Topically Strong Content With No Structural Signaling
Some of the sites Sarah audits have genuinely excellent content. Thorough, accurate, well-written. And structurally invisible to AI systems because the content was written for human navigation rather than machine extraction.
Good human-navigation content has narrative flow, builds context, and rewards a reader who starts at the top and reads through. Good AI-extraction content leads with the answer, uses question-format headings, and marks up Q&A pairs explicitly.
The solution isn’t to make the content worse for humans. It’s to add an extraction layer on top — question-format H2s, 40-word answer blocks, FAQ schema — that makes the same content accessible to both audiences.
5. No AI Visibility Baseline
The fifth finding isn’t a technical problem. It’s an awareness problem. Most sites Sarah audits have never checked whether they appear in AI-generated answers at all. They’re managing SEO metrics they can measure without asking whether those metrics translate to AI visibility.
They often don’t.
A site can have first-page rankings, strong traffic, and solid conversions in traditional search, and still have a complete Retrieval Gap in AI search. The two visibility systems are related but not identical. A site with strong SEO has a better foundation for AEO — but that foundation doesn’t automatically produce AI citations.
Establishing an AI visibility baseline is the first step in any AEO program. You need to know where you’re starting before you can measure whether you’re improving.
What Sarah Doesn’t Do
I want to be honest about what the audit is and isn’t.
Sarah diagnoses. She doesn’t prescribe. The audit report tells you exactly what AEO problems exist on your site, ranked by impact, with enough context to understand why each problem matters. It doesn’t hand you a copy-paste fix, because AEO problems are site-specific and the right solution depends on your content architecture, your CMS, and your competitive landscape.
The audit is the diagnosis. What you do with it is the treatment. Some clients implement fixes themselves. Some bring in a developer. Some come back to me for implementation support.
What every client has after the audit is clarity — something most sites don’t have going in. They know exactly why AI search can’t see them, and exactly where to focus their energy to change that.
That’s what a $450 AEO audit is: professional clarity on an invisible problem that’s costing you visibility you’ve already earned.
If you want Sarah to take a look at your site, you know where to find me.