Measuring AI Visibility

Why Traffic Is No Longer the Right KPI

When traffic drops, many teams conclude their content no longer works. That is usually a misjudgment. The problem is not the content, but the metric we use to judge it.

The wrong conclusion

It is not the content that fails - it is the measurement

Content teams see declining visits and conclude that AI-assisted content no longer pays off. In most cases that conclusion is wrong. Today, a drop in clicks says nothing reliable about the value of a piece of content.

The reason is a structural shift in search, not a quality problem with your pages.

Why the metric broke

The link between traffic and value has decoupled

For twenty years traffic was a good proxy for content value because value and clicks moved together. That link is broken. Industry data for early 2026 shows: 68% of US Google searches ended without a click - up from 60% in 2024. The main driver is AI Overviews.

According to an Ahrefs analysis, AI Overviews can reduce clicks to the top result by up to 58%. Worse, Search Console cannot tell whether a click came from classic search or from an AI summary. When clicks fall, it stays unclear whether the cause is an AI summary, a ranking loss, or answers read-but-not-clicked.

What really happens

What actually happens to "failed" content

For one brand, Seer Interactive found that although Overview CTR fell by 61%, the actual number of clicks stayed almost unchanged. An analysis of around 846,000 search sessions also shows: when an AI summary appears, users slow down and weigh their options more carefully. The results page now does the job the landing page used to do.

Decisive for visibility: pages cited in AI summaries get about 120% more clicks per impression than uncited ones. Being cited is not a soft factor - it is the measurable lever.

Not every page is affected equally

Category determines the impact

AI Overviews mostly affect informational research queries. Branded, local and high-intent transactional searches still perform well in organic. In e-commerce, "best of" lists and buying guides are hit hardest, while product and category pages keep converting. Publishers face the toughest situation.

That is why looking at total traffic is misleading: it mixes structurally affected pages with pages for which clicks are still the right measure.

What to do now

Act instead of retiring pages prematurely

Add a layer to your pages that an AI cannot generate: interactive graphics, video, downloadable content. Create memorable content that triggers branded search later. And do not retire a page just because its traffic dropped - first check what it actually achieves.

What to measure instead

A correlation dashboard, not a single traffic KPI

Rand Fishkin recommends tracking the publishing calendar together with several impact indicators instead of one traffic number. These indicators are exactly what TYS puts at the center - not rankings and click counts.

Branded search

Is the number of searches for your name growing? Direct evidence that content sticks.

Direct traffic

People who come directly remembered you - independent of the results page.

Visibility on AI surfaces

Is your brand cited in AI answers? Measured via the AI Authority Score across four systems (ChatGPT, Gemini, Claude, DeepSeek).

Visitor behavior

Reading depth, return visits, newsletter sign-ups - instead of a bare click count.

Actual outcomes

Conversions and lead quality. Early industry data suggests: AI traffic is lower in volume but often higher in intent.

Bottom line

Do not change the content first - change the measurement

You cannot decide based on a number that does not reflect what content actually achieves. Thinking only in clicks retires valuable content and starts over with every algorithm change. That is exactly why TYS measures AI visibility as a verifiable baseline - and ties it to branded search, direct traffic and real customer outcomes.

Sources

Measure what your content actually achieves

The Initial Check reveals your AI visibility as a baseline - the first indicator of the correlation dashboard.