Every legal marketing newsletter in your inbox has now run the same story. Google’s AI Overviews are eating law firm websites. Legal queries trigger them more than any other industry. Traffic is down nineteen percent even where rankings held. Clicks to the top organic result have fallen fifty eight percent.
In other words – act now, and here is a retainer agreement.
Some of it is true but measures something other than what you think. And one part of it is repeated so confidently, by so many people, that almost nobody has noticed it is contradicted by the very dataset it is drawn from.
Meanwhile, a firm that builds law firm websites for a living went and looked at what its clients’ analytics actually said.
Is AI Search Actually Killing Law Firm Website Traffic?
Great Jakes reviewed traffic patterns across its client portfolio over a full year and published the findings in May. The headline result was that among the law firm websites they looked at, traffic remained steady.
Their own summary is that the more dramatic predictions about answer engines draining law firm traffic have not materialised, at least not yet.
It is interesting to examine the supporting numbers. AI referrals accounted for 0.47% of sessions on average across their client sites, with larger firms seeing even less. LinkedIn referred 187% more sessions than every AI answer engine combined. Google was still sending, on average, 48% of all sessions, and more than 60% at some firms.
Before anyone in the AEO business writes in: Great Jakes are not sceptics. The same piece argues that AI’s influence on how clients find and assess firms will certainly increase, notes that AI referrals grew 220% over the year even at that tiny base, and sets out six things firms should do about it.
Hold 0.47% in your head because it is the most useful figure in this entire debate, and it is the one nobody quotes.
What Percentage Of Legal Searches Trigger An AI Overview?
One of the most repeated announcements on legal searches is that roughly 78% of legal queries trigger a Google AI Overview, the highest rate of any industry.
The first half is well supported. Several vendor analyses land near 78%, and legal informational queries clearly trigger overviews at a high rate.
The second half does not survive contact with the data. BrightEdge’s vertical tracking for February 2026 puts healthcare at 88%, education at 83% and B2B technology at 82%, all above legal. So legal is high.
Legal is not the highest of any industry, and every agency deck that says so is quoting a claim that its own preferred source contradicts.
The “highest of any industry” line is the one doing the emotional work in the sales pitch so when that is stripped out you have a sector that is somewhat above average on a metric that is rising everywhere.
What Does The 58% Click Loss Figure Actually Measure?
The Ahrefs study is real, large and honestly reported. It is also almost universally misread.
Ahrefs compared 150,000 keywords that trigger an AI Overview against 150,000 that do not, using aggregated Search Console data for December 2023 against December 2025. Position one click-through rate on the AI Overview keywords fell from 0.073 to 0.016.
The forecast for where those same keywords would have landed without AI Overviews was 0.037. The gap between 0.037 and 0.016 is the 58%.
The 58% figure is not the fall from 2023. Click-through rates were already collapsing on keywords with no AI Overview anywhere near them, and comparable non-overview keywords fell nearly as far over the same two years. AI Overviews accelerated a decline that was already running down the track.
It is also concentrated. The damage sits on informational queries. Branded and transactional queries hold up. Chartbeat, looking at more than 2,500 publisher sites, found small sites down about 60% while large ones fell 22%, with total traffic to those publishers holding roughly steady as other channels absorbed the difference.
Which brings us to the part where the two irreconcilable datasets turn out not to be in conflict at all.
Why Do The Two Datasets Disagree?
They are measuring different businesses.
The apocalypse numbers overwhelmingly come from the consumer intake end of the market such as the personal injury, family, immigration, criminal. Those practices built their entire acquisition model on informational content. “How does a deposition work.” “What is adjustment of status.” “How long does a divorce take in Texas.”
Those are precisely the queries an AI Overview answers completely, and precisely the traffic that is not coming back.
The steady numbers come from firm websites, where the traffic is largely branded and transactional. Nobody asks ChatGPT to explain a cross-border carve-out and then hires whoever it names. The general counsel who lands on a firm site typed the firm’s name, or a partner’s name, or followed a link from a directory, a ranking table, a LinkedIn post or an article like this one.
So both camps are telling the truth about their own clients and neither is telling the truth about yours. A consumer PI firm that reads the Great Jakes numbers and relaxes is making an expensive mistake. A commercial firm that reads the vendor decks and panics is being sold a solution to somebody else’s problem.
There is a third group, and it is the one that should actually be worried: the mid-market commercial firm that spent five years building a content library of explainer posts on the theory that topical authority would compound.
That library was aimed at informational queries, and informational queries are where the water is going out.
Who Benefits From The Panic?
It would be naive to pretend the incentives here are neutral. Almost every alarming figure in circulation was produced by a company that sells the remedy for it. That does not make the figures wrong, but it does mean the framing around them is not disinterested, and it explains why the reassuring dataset in this article had to come from an agency reporting against its own commercial interest.
The cost of the panic is real money. Personal injury firms are reportedly paying up to 568% more per click than they were in 2021, with some legal keywords now clearing $1,000 a click.
Meanwhile only around 14% of solo attorneys and 32% of small firms have a formal marketing budget at all. The firms least equipped to evaluate the claim are the ones being sold hardest.
What Should Law Firms Actually Do About AI Search?
Three things, in order, and none of them require a panic.
Measure your own site before you buy anything.
Google Search Console has had dedicated generative AI performance reporting since June 2026. Pull your own AI referral share and your own branded versus non-branded split. If your number looks like 0.47% and your traffic is flat, you have a monitoring job, not an emergency. If you are a consumer practice and your informational pages are down forty percent, you have a genuine problem and you now know where it is.
Get cited, because citation is the whole game.
Seer Interactive found brands named inside an AI Overview earned around 35% more organic clicks and 91% more paid clicks than uncited competitors on the same queries.
Semrush put AI-referred visitors converting at 4.4 times the rate of ordinary organic visitors, which is the same finding from the other end – someone who read a synthesised answer, saw your firm named and clicked anyway is close to picking up the phone. Fewer, better visitors is a real pattern, but only for firms that get named.
Understand that the website is not the whole surface.
AI systems break a question into sub-questions and answer from wherever the best material sits, which is often not your site at all. It is directories, rankings, earned media, LinkedIn, and legal trade coverage. Consistency of description across those sources now does work that no amount of on-site optimisation replicates.
Everything else is housekeeping that was worth doing in 2019 anyway: clean architecture, fast pages, proper schema, specific attorney bios instead of the same three paragraphs every commercial litigator in the country has published, and content that answers a question rather than performing expertise at it.
The Honest Version
AI search is going to matter enormously, and currently it does not matter very much. That may seem ironic, but both are true and the gap between them is where the entire legal marketing industry is presently making its living.
The correct posture is neither panic nor complacency. It is a spreadsheet. Look at your own numbers, work out which half of this market you are actually in, and treat anyone who tells you the answer before asking that question as what they are.
Write For LawFuel
A note on our own interest, since this article has just spent several hundred words on other people’s.
One of the few things both camps in this argument agree on is that AI systems assemble answers from multiple sources, and that a firm described consistently across credible third-party publications is easier for those systems to recognise and cite than one that only ever talks about itself. Trade coverage, directories, ranking tables and bylined commentary all sit in that set. Your own website does not.
LawFuel has published legal industry journalism for twenty five years across the US, UK, Australia and New Zealand, and we take contributed articles from practitioners at a reasonable cost. If you have something genuinely useful to say about a development in your field, we would like to read it. We publish under your name and your firm’s, with proper author attribution and structured data, which is the form that both search engines and answer engines can actually use.
What we are after is analysis a lawyer would want to read: a clear argument, something you know that we do not, and a view. What we are not after is a press release with the word “insights” on the front.
Pitches and drafts to john.bowie@lawfuel.com. Guidelines here.
