How AI Overviews Work

Based on observable behavior, Google AI Overviews tend to draw from service pages that open sections with direct answers, use clear heading structure, carry valid structured data in the initial HTML, and come from domains with established topical authority. Most service pages fall short on several of these dimensions — because they were built to convert, not to be extracted.

The retrieval pipeline and where service pages tend to fall short

Google's AI Overview system uses retrieval-augmented generation (RAG): it retrieves candidate pages, selects specific passages, synthesizes them into an answer, and attributes sources. Each stage presents challenges for service pages.

Retrieval stage — where JavaScript can reduce visibility

AI crawlers retrieve pages based on semantic relevance to the query. But if the page's content is rendered by JavaScript rather than served as static HTML, some crawlers may see an empty shell. The page may never enter the candidate pool regardless of how good the content is.

Passage selection — where conversion structure reduces extractability

From retrieved pages, the system selects specific passages — not whole pages. It looks for content that directly addresses the query in the first one or two sentences of a passage. A service page that opens with "Transform your business with our award-winning SEO services" provides nothing extractable. A page that opens with "SEO services typically include technical auditing, on-page optimisation, content strategy, and link acquisition" provides a direct, usable answer.

Structured data — where valid markup can improve interpretability

Valid structured data like FAQPage and Service schema helps label content explicitly for machine reading. Pages without schema require the AI to infer content structure — which may be less reliable. Pages with correctly implemented schema in the initial HTML can be parsed more accurately and with clearer context.

What queries surround your client's service pages?

Every service category has a surrounding cloud of informational queries that AI Overviews may answer. "What does SEO include?" "How long does SEO take?" "What is the difference between on-page and off-page SEO?" "How much should I pay for SEO?" These are not the client's money keywords — but they are the queries that introduce their potential customers to the category. Being referenced in these answers builds brand authority at the top of the funnel.

Identifying and structuring content for these surrounding queries is a core part of service-page AEO — and it requires understanding both the query landscape and the content transformation workflow needed to address it.

A simple retrieval example

A weak page may mention the answer somewhere in the body, but surround it with vague headings, long introductory copy, and support details that are hard to isolate.

A stronger page surfaces the answer earlier, uses headings that separate supporting ideas clearly, and places evidence where it can be identified without guessing.

That does not guarantee inclusion in AI-generated answers. But it does make the page easier for retrieval systems to parse, segment, and reuse.

This is one reason structure matters so much. The issue is not only whether the answer exists. It is whether the page makes that answer easy to extract.

Common failure patterns

In practice, across large sites and platform work, the same retrieval failures appear repeatedly when service pages are not structured for AI extraction.

• The page has relevant content but no passage that directly addresses the query within the first two sentences
• Headings use marketing labels instead of question-based or topic-based language
• Structured data is injected by JavaScript and not present in the initial HTML response
• The page is semantically relevant but structurally indistinguishable from dozens of competing pages
• Supporting evidence and specifics are mixed into the same passage as the core claim, reducing extractability

The AI Overview Presence Checklist

Before assuming a page has been ignored by AI Overviews, run it through the AI Overview Presence Checklist. The checklist isolates the structural conditions an answer engine needs in order to surface a page as a cited source:

• A direct answer to the page's primary query within the first viewport, written as a single extractable passage
• Heading structure that names the question, not the brand or the section
• FAQ block that covers the surrounding queries the AI Overview is likely to fan out to
• Schema that confirms the entity, the service, and the FAQ — and matches the visible content
• Trust signals (author, last updated, source references) on the page itself, not buried in the footer
• A meta description that reads as an on-intent summary the engine can lift verbatim

AEO Pro Lab runs the AI Overview Presence Checklist against any live URL and returns the specific conditions the page is missing — bound to evidence on the page, not generic best-practice advice.

About the author

A.L. MacFarland is the founder of AEO Pro Lab and writes about SEO, AEO, AI search visibility, and the structural side of modern discoverability. Connect on LinkedIn.