What AI search does to clicks and to cases

How much traffic has actually gone

The cleanest measurement I have is the Ahrefs study of click-through rates. They took 150,000 keywords with an AI Overview, then compared the set to 150,000 informational keywords with none. They then pulled the Search Console click-through rate for December 2023 and December 2025. The December 2023 baseline came just before Google took that feature to the public¹.

About 60% of the Google searches in Datos and SparkToro's tracking end with no click to any third-party site².

Right now the trend is bad and only getting worse.

So the question really is where the damage ends and the searches that bring you cases start.

The searches that sign cases are the least affected

This is the part that changes minds, and there are real legal searches tested for exactly this. Whitespark ran 540 searches in Houston, Phoenix, and Denver. They were done by hand across six industries including personal injury lawyers³.

The first row, that's people ready to hire. If they want a lawyer, they will see the map pack 93% of the time and an AI Overview 15% of the time.

For certain fields like healthcare and education, AI citations usually match organic rankings better than citations elsewhere. BrightEdge noted that overlap between citations and rankings ran in healthcare at 75.3%. Education worked out at 72.6% and the overall figure was 54.5%⁷. If legal has healthcare-like overlap, prioritise rankings. Legal got no category in the study, so this is an inference. The full reasoning is on the AI Overviews page.

What being named inside the answer is worth

Google is claiming the remaining share of clicks are better quality. If a person clicks through from an AI Overview, for example, they will be more likely to spend real time on the site⁵. Google says it has seen over 10% more use on queries with AI Overviews, but how do we even know⁶?

Like almost all vendor and platform claims, take these with a grain of salt. Either way, the underlying story is largely the same.

Why traffic has become a dangerous number to report

I have seen this scenario happen again and again.

Sessions drop 30% on the year and the managing partner sees the chart. The partner questions the spend and cuts the work. Meanwhile signed cases stay flat or go up for the same period; mostly because the lost sessions were explainer readers and unlikely to call.

This is worse in legal both in how enquiries arrive and in how firms track them.

BrightLocal found that 54% of local business Google Analytics properties run with no goals at all¹³. For many firms, sessions is the only number on the chart.

Volume was always the wrong yardstick here

I think a lot of the panic about lost clicks is because it's a volume worry with fairly little connection to cases. Legal search has been about small numbers. "Law firm seo" is a term with 8,900 national searches a month, and a $40 cost per click. A city-plus-practice term gets a few hundred monthly searches⁹.

You chase the smaller term, because a term with 90 searches a month and a $90 click is worth more than one with 9,000 searches and a $2 click. The maths is on the keyword research page.

Most of the session mismeasuring happened before AI answers existed.

What still earns clicks, and why firm sites are better placed than publishers

The Zyppy study included 400 websites from Google's twelve-month winners and losers. They ran a correlation for five features and the traffic outcome¹¹. The two that tied most strongly to gaining are these.

• Selling a product or service, Spearman correlation 0.391
• Letting a visitor complete their task on the site, 0.381

The five features also compound one another. Only 30.7% of three-feature sites ended up gaining traffic. A fourth feature changed the picture and lifted the gaining rate to 68.1%¹¹.

Law firm sites that come to market with both of the top two features have a significant advantage. Having a service to sell helps, especially if visitors can book or call straight from it. The losing sites were mostly informational publishers. That is the model law firm blogs have copied.

Never underestimate the speed of people deciding on a lawyer¹².

• 56% of legal consumers acted within a week of realising they had an issue, and 16% within a day
• Of those who found an attorney online, 82% used reviews
• 40% called reviews their primary source for deciding who to contact

What I would do about it

1. Fix the measurement before changing anything else. Phone enquiries should be attributed as tracked calls. Each form should be tracked separately as an event. Create a field that enables the firm to link each enquiry to a signed matter. Once you skip the tracking you cannot price the click loss.
2. Agree the reporting order in writing: signed cases, qualified enquiries, response time, then traffic as context. I put the order in writing routinely at the start of the engagement. There is a template on the intake and measurement page for this.
3. Separate your pages into two piles. One pile is the hire-ready practice area pages and location pages. The other main pile is informational. Look for a twelve-month trend to check in each pile instead of the site total. Different piles have different trend lines.
4. Stop producing thin informational posts whose entire answer fits in an AI Overview. They are cheap to make (plus they are worthless now).
5. Keep the informational pages that can win a citation. Make it clear what the specific answer is at the top, and, more importantly, how the heading matches the question. The actual method is only set out on the writing page.
6. Put the things only your site can do in front of people: booking, calling, a deadline checker, a fee calculator. Task completion remains the second strongest of the gaining features¹¹.
7. Then fix the phone. I will say that 39% of the firms out there (at least in Hennessey's 2025 study) take more than two hours to answer an online enquiry, or never answer it¹⁰. Every argument on this page is a moot point for that group.