Found, cited, visited: how to separate the outcomes of AI search

SEO has not stopped mattering. What has changed is that a second competition is now happening alongside the first: besides showing up among search results, a page can be one of the sources used to build an AI-generated answer. This difference interests me because it changes what I need to look at when I evaluate how a piece of content performs.

What Google says about the foundation

In Google’s case, the official guidance is direct. The Search Central guide on generative features states that these features rely on the same core ranking and quality systems as Search, and that SEO best practices remain the foundation (Google Search Central, AI optimization guide). The page on AI Overviews and AI Mode adds that there are no additional requirements to appear in these formats (Google Search Central, AI features and your website).

This applies to Google. It does not mean every search engine and assistant uses the same criteria, and each environment needs to be observed on its own terms. I wrote more about this question of scope in GEO, AEO and LLMO: scope comes before the acronym.

Three different events

Being found, being cited and receiving a visit are distinct events, and one does not guarantee the other.

A page can be found by the system and not appear in the answer. It can appear as a source and receive no clicks. It can receive a visit that leads to no action that matters for the business. Treating these three moments as a single “AI visibility” metric mixes signals that call for different readings.

One data point helps gauge the distance between them. In a study of 900 adults in the United States, Pew Research Center analyzed almost 69,000 Google searches made in March 2025. When the results page showed an AI-generated summary, users clicked on a traditional result in 8% of visits, compared with 15% when there was no summary. Clicking on a link inside the summary itself happened in only 1% of visits (Pew Research Center, 2025).

The study has its own scope (one country, one month, one sample) and should not be generalized to every industry. The useful point is a different one: being among the sources of an answer and receiving a visit are things that can be quite far apart.

Four questions to track

Instead of a single metric, I prefer to split tracking into a sequence of questions.

question	what to look at
Was the page found?	indexing, impressions, queries it appears for
Was it used as a source?	presence in generated answers, with or without a link
Was there a visit?	clicks and sessions coming from these surfaces
Did the visit lead to anything?	reading, contact, sign-up or another action defined in advance

Some of these questions already have tools. Google states that traffic from AI Overviews and AI Mode is included in Search Console’s overall performance report, and Search Console Help describes a specific generative AI performance report that shows impressions of the site’s links in these features, still rolling out gradually and with data limitations (Search Console Help, Generative AI performance report). For other assistants, observation tends to be more manual and less stable.

Be careful with the one-off screenshot

The most common way to report AI presence is still the screenshot: someone asks an assistant a question, the brand shows up, and the image circulates as proof. The problem is that generated answers vary with the wording, the history, the timing and the system itself. A screenshot shows that something happened once, not that it happens consistently, and much less that it generated a business opportunity.

That does not make observation useless. It makes it necessary to record the question used, the date, the environment and the repetitions of the test, and to treat the result as a signal, not as evidence of performance.

Where the work stands

My reading is that the work expands rather than being replaced. Taking care of the site, its structure, and the clarity and usefulness of its content is still necessary. Alongside that, it now makes sense to investigate how this content appears in other search experiences, with separate questions for each stage. This approach seems more useful to me than celebrating every brand mention as a success or abandoning SEO because the search interface changed.

References

• Google Search Central. Google’s guide to optimizing for generative AI features on Google Search. Accessed September 29, 2026. Link
• Google Search Central. AI features and your website. Accessed September 29, 2026. Link
• Google. Generative AI performance report (Search). Search Console Help, accessed September 29, 2026. Link
• Pew Research Center. Google users are less likely to click on links when an AI summary appears in the results. July 22, 2025. Link