How AI Overviews choose sources

What the system appears to do

Google describes AI Overviews as generated with the help of its search systems, which find relevant pages and use them to support the summary. In practice, the linked sources tend to be pages that already rank reasonably well for the query or for closely related queries. A query can also be broken into several sub-questions behind the scenes, which is why you sometimes see a source that ranks for a narrower version of the question.

The important consequence: if Google can’t index your page, or doesn’t consider it relevant and trustworthy for the topic, it is unlikely to be linked in an AI Overview. There is no side door.

Signals you can influence

• Indexability. The page must be crawlable, indexable, and not blocked by robots.txt or a noindex tag. Check it in Google Search Console with the URL Inspection tool.
• Relevance to the exact question. Pages that address the specific question, ideally in a heading and a direct answer, have a clearer claim than a general page that mentions it in passing.
• Answer shape. Overviews often summarize definitions, steps, and comparisons. A short definition under the heading, followed by numbered steps or a compact table, gives the system clean material to support a summary.
• Evidence of experience. Google’s guidance on helpful content stresses first-hand experience and clear authorship. Show who wrote the page, why they know, and what they actually tested.
• Freshness where it matters. For topics that change, such as pricing, software features, or regulations, keep the page current and say when it was last updated.

What probably doesn’t help

Some tactics get passed around that have no public support. There is no “AI Overview schema.” Stuffing a page with question phrases doesn’t make it more likely to be cited, and it makes the page worse for people. Generating many near-duplicate pages to cover every variant of a question tends to dilute rather than help.

Structured data is still worth using for what it is designed for, such as Organization, Product, FAQ, or Article markup, because it helps Google understand the page. Just don’t expect it to be a switch for AI Overviews.

A practical workflow for a small site

• Pick ten informational queries that matter to your business and that currently show an AI Overview.
• For each one, record which sources the Overview links and where your page ranks in the regular results.
• If you don’t rank in the top results at all, treat it as a normal SEO problem first: a better page, better internal links, and a few relevant mentions.
• If you rank but aren’t linked, compare your page to the linked ones. Look for a missing direct answer, a missing definition, outdated facts, or a structure that’s harder to summarize.
• Make one focused change per page, then re-check over the next few weeks. Overviews vary from search to search, so look for a trend rather than a single result.

Reporting this to clients

AI Overviews make traffic reporting messier. Clicks from a linked source in an Overview are counted in Search Console as regular web search traffic, so you can’t cleanly separate them there. A simple, honest approach is to report two things side by side: whether the client is linked in the Overview for their priority queries, checked on a schedule, and the organic clicks to those pages over the same period.

A worked example

Say you run a small bookkeeping firm and the query is “how to reconcile a business bank account.” The Overview links a large accounting software help page, a government small-business page, and a blog post. Your page ranks seventh. It opens with a story about your founder and gets to the steps in the fourth paragraph. The linked pages all open with a one-sentence definition followed by numbered steps.

The fix isn’t a trick. Move the definition and the numbered steps to the top, add a short “common mistakes” list based on what you see with real clients, and add your name and credentials as the author. Then check the query again over the following weeks.

Where AI Cite Check fits

AI Cite Check currently checks Perplexity and ChatGPT with web search. Google AI Overviews support is coming soon. The same page improvements tend to help across all three, because each one rewards pages that are easy to retrieve and easy to quote.