I've realised ranking and being cited in AI Overviews aren't the same thing, so I've started tracking citations more systematically. It's still a small sample - not a big research report, just a practical framework.
Here's what I'm logging for each query:
🔍 the query itself
📊 ranking position of the page
🧠 whether an AI Overview appears
📎 whether the page is cited
📄 page type
💬 answer format
🔧 schema present or absent
⏱️ content freshness
🧩 whether the main answer is easy to find on the page
A few early patterns I'm watching:
➡️ Pages with direct answer sections are easier to evaluate. If a page answers a narrow sub-question clearly, it's obvious why it might be used or ignored.
➡️ Pages with clear entity/source context feel easier to understand - the page makes it obvious who the company is, what the page covers, and why it should be trusted.
➡️ Long pages aren't automatically better. Some bury the actual answer too far down, while shorter pages sometimes make the answer more extractable.
➡️ Ranking and citation seem related but not identical. A page can rank well and still not be the clearest source for the AI answer.
What I'm not claiming:
❌ There's no guaranteed formula here
❌ My sample size is still limited
❌ AI Overview behaviour keeps changing
❌ Schema alone doesn't cause citations
Right now I'm mostly trying to separate normal SEO visibility from AI citation visibility. Someone in the thread suggested logging external citation count for the cited page too - AI Overviews seem to lean heavily on pages with strong third-party mentions, not just on-page optimisation. That's a smart addition I'll start capturing.
Still doing manual spreadsheets and prompt snapshots for tooling. nothing automated feels reliable yet - the AI visibility tools are inconsistent and pricey given how much LLM outputs shift day to day.
Are you logging citations manually, using tools, or mostly watching traffic and impressions?