Click-through rate (CTR)

Click-through rate is clicks divided by impressions for the same set of results over the same window. Google's Search Console help states the definition literally: CTR is "the calculation of (clicks ÷ impressions)".

The two terms are the whole subject, and each has a rule that decides what the ratio means.

An impression is "how often someone saw a link to your site on Google". Google counts one "whenever an item appears in the current page of results, whether or not the item is scrolled into view, as long as the user need not click to see more results". Visibility in the human sense is not required; being on the page that was served is. Independently scrolling or expanding widgets are the exception — inside a results carousel or an expanding FAQ result, the item "must typically be scrolled into view within the carousel, or expanded by a click, to register an impression".

A click is "how often someone clicked a link from Google to your site", and Google draws the boundary at its own platform: "Any click that sends the user to a page outside of Google Search, Discover, or News is counted as a click; clicking a link that stays inside the Google platform is not counted as a click."

So CTR is not a measure of attention. It is a measure of departure, divided by exposure.

How aggregation changes the number

The same traffic produces different rates depending on how rows are grouped. Google aggregates by property or by page: aggregated by property, one impression is counted for an entire element that contains several of your links; aggregated by page, impressions are counted separately for each distinct URL within that same element. The click count does not change, so the identical week can show two legitimate and different CTRs.

Position behaves the same way. Search Console reports the topmost position occupied by a link to your property or page, averaged across all queries — an average of a ranking, not a ranking. Averaged positions and averaged ratios do not compose: a page can improve on every query it serves and still show a worse average CTR because the mix of queries moved.

Sampling matters too. The Search Analytics API's rowLimit accepts 1–25,000 and defaults to 1,000, and Google states plainly that the API "does not guarantee to return all data rows but rather top ones". A site-wide CTR computed from a default query is a CTR over the head of the distribution.

Why the ratio decays as a goal

CTR was a good proxy for reach when the only way to get an answer was to leave the results page. That link is what generated surfaces break.

Google reports impressions inside AI Overviews and AI Mode — its help defines them as how many times links to your site were shown in a generative AI feature on Search, and says the insights were rolled out to all websites worldwide as of August 31, 2026. Those impressions land in the denominator. The corresponding read, however, often happens inside the answer, and a click is only counted when the user leaves Google. The predictable result is a falling ratio produced by a format change rather than by anything you did.

Google's own guidance on AI features says there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary", and that you "don't need to create new machine readable files, AI text files, or markup to appear in these features". There is no CTR lever hidden in that surface; there is a different relationship between being read and being visited. That is the subject of zero-click search.

Failure modes

• Treating a declining CTR as a page defect. Check whether impressions grew in a surface that answers in place before rewriting a title.
• Comparing property-aggregated and page-aggregated rates. They are different denominators over the same clicks.
• Averaging averages. A site-level CTR is a ratio of sums, not a mean of query-level ratios; the two diverge whenever query volumes are uneven.
• Reading a default API pull as complete. Without paging past 1,000 rows you are measuring your best-known queries.
• Extending CTR to assistants that do not report it. Outside Google Search there are no impressions to divide by. If the question is whether a model named you or cited you, the measurement is a sample of answers, not a ratio — see source citation.

What to do with it instead

Use CTR where its definition holds: comparing a single page's listed result against itself, before and after a title or description change, on a stable set of queries and a fixed aggregation. It is a good instrument for that and a poor instrument for anything that spans surfaces.

For reach in generated answers, measure the answers. Whether a model names you, and which pages it cites when it does, are separate observations from whether anyone clicked, and neither can be recovered from a ratio Google computes about its own results page.