Here's what most guides on this topic have in common: they're vague enough to be useless and specific enough to sound credible.
"Optimise for E-E-A-T." "Create quality content." "Be authoritative."
Cool. Extremely actionable.
I've been running my own checks on this for a while — not from a live campaign, but from genuine curiosity about where search is actually going. Checking which pages get cited, which don't, and what the patterns look like when you sit with them long enough. Here's what I've actually noticed.
1. It's not about authority. It's about clarity.
The pages getting cited in AI Overviews aren't always the highest-authority domains. They're the ones that answer a specific question most directly.
Google's AI isn't surfacing the most trustworthy page on the internet for a given query. It's surfacing the most legible one.
That distinction matters more than most people realise. I've watched pages with mid-tier domain authority get consistently cited because they structure content around a clear question and give a direct answer in the first paragraph. No preamble. No "in this article we'll explore..." No long introduction about why the topic matters. Just: here's the question, here's the answer, here's why.
That's most of the work.
2. The format that gets pulled
The pattern I keep seeing across pages that show up in AI Overviews:
A direct, two to three sentence answer appears within the first 150 words. Not below the fold. Not after a definition and a brief history of the subject. Right at the top, plainly stated.
After that, the page goes deeper — but the entry point is always that clean, specific summary. Think of it as writing for two audiences simultaneously: the AI that needs to extract a citable snippet, and the human who wants the full picture.
The pages that do both from the first paragraph almost always get cited. The ones that bury the answer in credentials and context usually don't.
3. What entity signals actually do
There's a lot of conversation about entity building in SEO circles right now. Most of it is right, but the framing is slightly off.
It's not that being a recognised entity makes you more likely to get cited. It's that entity clarity makes it easier for the AI to understand what you're saying and why it's relevant to a specific query. If Google's knowledge graph doesn't have a coherent model of who you are and what you cover, the AI won't confidently associate your content with a query — even if your content is genuinely the best answer.
Practically, this means your About page matters more than it used to. Your structured data matters more. Third-party mentions on credible sites matter more. You're not just optimising a page anymore. You're helping an AI system build an accurate picture of what you know.
4. What I've stopped believing
That FAQ schema is the unlock. It helps, but I've seen plenty of pages with perfect FAQ markup that never get cited, and pages without it that consistently do.
That ranking first guarantees a citation. It doesn't. The AI surfaces what it needs, not what ranked first.
That longer content wins. Some of the most consistently cited pages I've looked at are under 600 words. They're just completely, specifically right about one thing.
5. The uncomfortable part
Most pages aren't getting cited in AI Overviews because they weren't written to answer a question. They were written to rank for a keyword.
Those are related goals. They're not the same goal anymore.
A keyword tells you what someone searched. A question tells you what someone needs to know. The pages built around the second tend to be the ones that show up.
This isn't a new insight. But it's one most content strategies haven't actually internalised. The AI is reading your page looking for an answer to give someone. If your page is an answer, it gets cited. If it's an optimised document, it might rank. Different things.
Write the answer first. The rest follows.
