Every week someone asks us about AEO. Answer Engine Optimization. How do we get into the AI Overviews? How do we rank in ChatGPT? What's the trick? And every week I give them the same answer, which is not what they came for: you probably do not have an AI search problem. You have a proof problem.
Search is changing fast. That part is real. AI Overviews appear across Google results. ChatGPT cites sources. Perplexity links to specific pages. Businesses are paying agencies real money to optimize for AI search, and a small industry has grown around the promise: get the right schema, the right structure, the magic llms.txt file, and you will appear in the answers. Done.
The received wisdom says AEO is a new discipline requiring new tools, new tactics, and a new checklist. That most businesses are missing out on AI visibility because they have not bought the right package yet.
The actual truth is considerably more boring. AI systems cite things that are worth citing. They pull answers from sources that are clear, specific, factually grounded, and relevant. Google's own guidance comes down to three things: be unique, be useful, and be readable. Not by humans only. By machines making decisions about whether your content belongs in an answer at all.
The real problem is usually evidence
What's actually happening when a business does not appear in AI answers? It is almost never that they lack the right schema. It is that they do not have the kind of specific, citable proof that makes an AI system reach for them as the answer.
Their About page is vague. Their service pages describe what they do rather than demonstrating they have done it. Their case studies do not exist, or they are hidden, or they use language so generic that no model would trust it as evidence of anything.
This is what I mean by a proof problem. An AI search system is essentially asking: does this source know what it is talking about? Is there specific, verifiable evidence here that I can extract and cite with confidence? If the answer is no, if the page is full of marketing language, aspirational claims, and zero specifics, no amount of schema fixes that. The structure is perfect. The substance is not there.
What the proof layer looks like
Real client outcomes with numbers, not we helped our clients grow. Specific methodologies described in enough detail that an agent could summarise them accurately. FAQs written around actual questions actual buyers ask, answered properly, not great question, it depends on your situation.
Named services with clear scope, pricing signals, and what-happens-next. Transcripts, case studies, data points, process notes, before-and-after evidence, and original thinking that a language model can extract and attribute with confidence.
None of that is AEO. All of it is what AEO checks for.
Google is already telling you this
Google's generative AI search guidance says the same thing in calmer language. SEO basics still matter because AI features are rooted in Search ranking and quality systems. Google also says to create valuable, non-commodity content, avoid overdoing query-variation pages, and ignore AI search hacks like special llms.txt files for Google Search.
That does not mean technical structure is irrelevant. Crawlability, indexability, snippets, structured data, and clear page architecture still matter. But they are eligibility work. They help your evidence get found and understood. They do not create the evidence for you.
The listicle-farm AEO approach, where you generate fifty optimized FAQ pages and flood the index, has a short shelf life. Google's scaled content abuse policy already points in that direction. What survives is genuine documented expertise that would be worth citing even if AI systems did not exist.
The audit is brutally simple
If you are spending on AEO packages right now and wondering why the dashboards do not show citations improving, start here: open your own website and ask honestly, if I were an AI system deciding whether to pull an answer from this page, would I trust it? Is there anything on this page that is actually specific enough to be useful?
Most businesses already know the answer before they finish asking the question.
The fix is not a new tool. It is documented proof. What you have built. What you have measured. What happened. Who it happened for. Written clearly enough that a machine reading it at speed would find something worth keeping.
AEO without proof is scaffolding with no building behind it.
What should businesses document for AEO?
Start with the facts an answer engine can safely cite: what you do, who it is for, how the work happens, what results you have created, what constraints matter, what buyers ask before they commit, and what proof supports the claim. If those details are not on the page, the model has nothing solid to use.
Does schema help with AI visibility?
Schema can help search systems understand page context and qualify for rich results, so it is still worth doing properly. But schema does not make a weak claim stronger. It labels the content you already have. If the proof is thin, the label will not save it.
What should you fix first?
Fix the evidence before the embellishment. Publish real case studies, specific service pages, buyer questions, process detail, expert explanations, and source-backed claims. Then make the site technically clean enough for search systems and AI agents to retrieve that evidence without friction.
Evidence and further reading
- Google Search Central: Optimizing your website for generative AI features on Google Search
- Google Search Central: Creating helpful, reliable, people-first content
- Google Search Central: Guidance on generative AI content and scaled content abuse