AI overviews changed the shape of the click, not the work

Google's AI overviews sit above organic results and answer the question directly. The measurable consequence is straightforward and unpleasant: for informational queries, you keep the impression and lose the click.

The unmeasurable consequence is more interesting.

What it does to your reporting

Position stays the same. You still rank third. The overview is not an organic result and does not change your organic position.

Impressions stay the same. You were still shown.

Clicks fall. The answer was given before anyone reached you.

So the metric that moves is click-through rate at a fixed position — the one number most reports do not track, because it is derived rather than reported. A site can lose a third of its organic traffic while every headline number in its SEO report looks unchanged.

Which queries are affected

Overwhelmingly informational ones. "How do I", "what is", "why does" — questions with a settled answer that fits in a paragraph.

Much less affected: transactional queries, comparisons where the answer is genuinely contested, anything requiring a specific local answer, and anything where the user needs to do something on a site rather than know something.

That distinction is the whole strategy. Content whose entire value was answering a question in a paragraph is now competing with a paragraph that appears above it and requires no click.

What actually responds

Be the source, not the competitor. Overviews cite. Being cited is worth less than being clicked, but it is worth considerably more than being neither, and citation correlates with being clearly structured and unambiguously stating the answer.

Write the thing a paragraph cannot replace. A number specific to your data. A comparison table with your prices. A calculator. Anything where reading the summary makes someone more likely to click, not less.

Stop measuring informational content by traffic alone. If a page's job is to make you the obvious source, its traffic will decline while its contribution does not. This is a genuinely hard reporting problem and pretending otherwise is dishonest.

The check to add

Track click-through rate per query, at a fixed position band, over time. That single series shows you the effect clearly, and no standard report contains it.