How are we going to handle attribution as clicks disappear?

How are we going to handle attribution as clicks disappear?

AI search is going to reduce clicks. That part is not controversial.

The uncomfortable consequence is what follows: fewer clicks means fewer people trigger web analytics, which is still how most organisations measure performance and assign value. The inevitable outcome is that SEO will look like it is getting worse, even when it isn’t.

...likely one of the most important problems we need to solve in 2026.

This is not a tooling issue, it’s an attribution problem - and it’s likely one of the most important problems we need to solve in 2026.

If discovery, evaluation, and even decision-making increasingly happen with AI search, then traditional performance signals degrade. Not because demand disappears, but because it no longer expresses itself through visits in the same way.

Attribution alternatives for AI Search/GEO

Here are some of the approaches that are emerging, none of which are perfect:

Educate stakeholders that clicks matter less

This has to happen to some extent. The mechanics of AI search mean fewer users will reach websites, even when brands are influencing decisions earlier in the journey. That reality needs to be clearly explained.

The problem is that education alone rarely protects budgets.

Asking senior leadership to accept declining traffic without a credible replacement signal quickly turns into “trust us, it’s working.” Without additional evidence, SEO and AI-driven discovery risk being perceived as cost centres rather than demand drivers.

Adopt proxy metrics for traffic and exposure

If clicks are no longer the primary output, we need other indicators of presence and demand. Impressions from Google Search Console, Google Trends data, and visibility metrics from rank or AI tracking tools can help show whether a brand is appearing at key moments in the user journey.

These metrics don’t map cleanly to revenue, but they do demonstrate whether demand is being captured or lost. Over time, organisations may need to treat these exposure signals as leading indicators, rather than treating traffic as the only proof of value.

Correlate exposure with revenue

In principle, this is straightforward: if exposure increases and revenue follows, there is a relationship worth understanding.

In practice, it’s difficult.

This approach works best when exposure can be isolated to specific sections of a site, product lines, or markets where other variables are relatively stable. On large or complex sites with overlapping channels, long consideration cycles, and multiple teams influencing outcomes, correlation can quickly become noisy and contested.

Media Mix Modeling (MMM)

MMM offers a more formal way to understand how different channels contribute to outcomes, using historical data across spend, channels, and seasonality.

In theory, it’s well suited to a world where direct attribution breaks down.

The challenge is that AI search is not yet a clearly observable channel. Until AI exposure can be measured with more consistency, integrating it into an MMM framework requires assumptions that need careful validation rather than simply adding another row to the model.

Model SEO and AI search impact on Direct

Instead of treating Direct as a standalone channel, it can be more useful to see it as intent finally realised.

Direct traffic often reflects the cumulative impact of earlier exposure across search, AI interfaces, and paid media. By combining indicators of brand demand, AI visibility, and high-level paid activity, it’s possible to build lighter-weight models that estimate how SEO and AI search influence top-level demand. These models won’t be precise, but they can be directionally valuable.

Clickstream-based forecasting

Third-party clickstream providers such as Similarweb or Datos can be used to model user journeys across search engines and AI platforms they are able to observe. This can help estimate shifts in behaviour and provide directional forecasts, sometimes even tying them back to revenue.

However, this approach comes with important caveats. Questions around data sourcing, representativeness, and validation need to be addressed, and any outputs should be treated as indicative rather than definitive.

None of this is comfortable. It requires more data literacy, more cross-team alignment, and more tolerance for uncertainty than most organisations are used to.

SEO isn't Dead, Attribution Is

But pretending clicks still tell the full story is worse. That’s how channels get defunded while still doing their job.

AI search isn’t killing SEO. It’s killing lazy attribution.

The organisations that navigate this well won’t be the ones with the fanciest AI dashboards. They’ll be the ones that accept attribution is becoming less precise, build trust in directional signals, and adapt how performance is evaluated before the numbers force the issue.

AI search isn’t killing SEO. It’s killing lazy attribution.

What matters now is whether we evolve our measurement models fast enough to keep up.

Even if platforms like OpenAI or Google release better data in the future, the core issue remains the same: attribution systems built around clicks are no longer sufficient.

This is a problem we need to actively solve, not wait out. How are you thinking about attribution as we move into 2026?