How to Measure SEO ROI When Most Queries Never Get Clicked

The question of whether SEO pays is reasonable and the standard answers to it are not. Most ROI models assume a click leads to a session leads to a conversion, and each of those links is weaker than it was.

We have our own evidence on this and it is uncomfortable. In the period we measured, the overwhelming majority of queries where this site appeared on page one produced no clicks at all.

That does not mean the presence was worthless. It means a measurement model built on clicks would have reported it as worthless, which is a different and more fixable problem.

Why are clicks a bad measure of SEO?

Short answer: Because an increasing share of searches are answered on the results page. Pew Research found users clicked a result on 8% of visits where an AI summary appeared and 15% where none did, so the same visibility now produces roughly half the clicks it used to.

Ahrefs found the same direction of effect on click loss for top-ranking pages when an AI Overview is present. Two independent sources pointing the same way is about as settled as this industry gets.

The consequence for measurement is specific. If clicks fall while impressions hold, a click-based model reports a decline where the underlying visibility is flat.

That is not a reason to stop measuring clicks. It is a reason not to make them the first number, because the first number is the one that drives the decisions.

What does our own data show?

Short answer: That click-based measurement would have reported almost nothing. Across the queries where this site reached page one in the measured period, 99.2% produced zero clicks, which is a figure we published rather than one we would have chosen to publish.

We are using our own data here because it is the only data we can characterise without speculating about anyone else. It is also a worked example rather than a general law.

The shape of it is what matters. A site can be present, relevant and visible while generating almost no clicks, and the gap between those two facts is where most SEO arguments go wrong.

If your own data looks like this, the correct response is not to abandon search. It is to measure the presence directly and to value it separately from the traffic it does or does not send.

What should I measure instead?

Short answer: Three things in order: impressions by query, average position by query, and cost per qualified enquiry against a pre-work baseline. The first two come free from Search Console and answer different questions; the third is the one a finance team will accept.

Notice that four of the six are free and most businesses track none of them. The paid tools in this category mostly repackage the free data with a nicer chart.

The distinction between the first two rows is the single most useful diagnostic in search. Impressions without position means relevant but not credible, which is an authority problem.

How do I build an ROI case that survives finance?

Short answer: With a dated baseline, a defined attribution window and an explicit statement of what you cannot attribute. A model that claims to trace every pound of revenue back to a specific keyword will be dismantled in the first serious review, and it will deserve to be.

1. Take the baseline before anything starts. Impressions, positions, branded search volume and enquiry count, exported and dated.
2. Define a qualified enquiry in writing, with your sales team, before you start counting them. Otherwise the definition will drift toward whatever makes the number look better.
3. Pick an attribution window and state it. Thirty or ninety days, consistently applied, beats a sophisticated model nobody trusts.
4. Report what you cannot attribute. The enquiries that came through a referral after someone read a page are real and uncountable, and saying so buys credibility for the parts you can count.
5. Compare against the baseline, not against the previous month. Month to month comparison in search is mostly seasonality.

Step four is counterintuitive and it is what makes the rest believable. An analyst who volunteers the limits of their own model is harder to dismiss than one who does not.

Is SEO ever not worth measuring that hard?

Short answer: Yes, when the spend is small enough that the measurement costs a meaningful fraction of it. A business spending a few hundred a month should track enquiries and impressions and stop there, because a full attribution model would cost more than the programme.

This is a real tension in small-business marketing. Proper measurement has a fixed cost that does not scale down.

The pragmatic version is the baseline plus two free numbers. It is not rigorous and it is enough to tell whether to continue.

Above roughly the full-programme band, the measurement becomes worth building properly, because the decisions it informs are large enough to justify it.

Frequently asked questions

What is a good ROI for SEO?

There is no benchmark worth quoting, because it depends entirely on your margin and the value of an enquiry. The useful comparison is against your other channels on cost per qualified enquiry, using the same definition of qualified for all of them.

How long before SEO ROI is measurable?

Impressions move first and are measurable within weeks. Enquiry-level effects take longer and are noisier. We target and measure both rather than promising a payback period, because the honest variance between businesses is wider than any average.

Should I use a rank tracking tool?

As a secondary measure at most. Search Console gives you the position you actually held for queries you actually appeared for, which is better data than a simulated check from an arbitrary location.

How do I attribute an AI-generated answer?

Mostly you cannot, which is the honest answer. Brand mentions inside an assistant answer leave no referral trail, so the proxy is branded search volume and direct enquiry growth, both imperfect and both better than nothing.

Does my agency track this or do I?

The data should be in properties you own, so you can always check. An agency that reports from accounts you cannot access is asking you to take its measurement on trust, which is the one thing measurement exists to avoid.

Sources and further reading

• Google Search Central, getting started with Search Console
• Pew Research Center, Google users are less likely to click on links when an AI summary appears (22 July 2025)
• Ahrefs, AI Overviews reduce clicks to top-ranking pages
• Google Search Central, AI features and your website
• Google Search Central, do you need an SEO?

Want a baseline taken properly?

We take and date a baseline before any work starts, in properties you own, because without one there is no honest argument about what changed. Ask us for it even if you hire someone else.