Is your law firm keyword growth declining? You’re not alone.
If you’ve opened Search Console this year and felt your stomach drop, you’re not imagining things. Across three of our larger client accounts- one criminal defense firm, one family law firm, and one estate planning firm- impressions and total clicks have been sliding since the first week of January. Same shape on all three charts. No single algorithm date to blame.
Here’s the part that doesn’t match the panic: none of those three firms are signing fewer cases. All of them are up significantly.
That gap between the traffic graph and the intake calendar is the most important thing happening in law firm SEO right now, and most firms are reading only half of it.
What’s Actually Declining (Hint: It’s Your Blog Posts)
When we broke the losses down by query type, the pattern was consistent across all three accounts. Almost all of the decline came from informational searches.
“How long does a DUI stay on your record?” “What happens to the house in a divorce?” “Do I need a will if I already have a trust?” Those are exactly the questions Google’s AI Overviews were built to answer, and Google now answers them on the results page without sending anyone anywhere.
The leading AI and SEO industry data lines up with what we’re seeing:
- Ahrefs analyzed 300,000 keywords and found a 34.5% CTR drop for the top-ranking page when an AI Overview was present.
- When they re-ran the study on December 2025 data, that gap had widened to roughly 58%.
Seer Interactive, working from 2.43 billion organic impressions across 53 brands, measured a 61% CTR decline on affected queries. - Pew Research found that only about 1% of users click a link inside an AI Overview itself.
- BrightEdge now tracks AI Overviews on roughly 48% of queries, up 58% year over year.
But look at where those overviews trigger. They show up on informational intent. They rarely show up for “criminal defense attorney Cheyenne,” “personal injury lawyer near me,” or somebody typing your firm’s name. Transactional and local searches are still a high-click zone.
So when your impression graph falls, the real question isn’t how far it fell. It’s what fell. In every account we’ve audited this year, the answer has been top-of-funnel blog content, not the pages that actually produce consultations.
Rankings for the Queries That Matter Went Up
While informational traffic softened, something else happened in the same accounts. Rankings improved for the terms tied to actual case value.
Three things drove it, and none of them are exotic:
Depth over breadth on landing pages. Instead of one page for “criminal defense,” we built out separate pages by case type, by jurisdiction, and in some cases by both. A page targeting “felony drug possession defense in Laramie County” competes in a much thinner field than a general practice area page, and the person searching it is far closer to picking up the phone. These pages gained in both ranking position and volume.
Schema and technical signals. Attorney schema, LegalService markup, FAQPage markup, review markup, and proper entity connections between the firm, its attorneys, and its locations. Google deprecated FAQ rich results in May, but the structured data still helps AI machines understand what a page is and who wrote it. That matters more when AI platforms are deciding whether to cite you (or someone else).
Blogs as an internal linking engine. We stopped treating blog posts as traffic destinations and started treating them as authority routing. Every new post links into the relevant practice area page and the relevant geographic page with real anchor text. The post itself may only pull a hundred impressions a month now. The page it points at is ranking for terms worth thousands.
Lead volume followed the deep pages, not the traffic total.
AI Referral Traffic Is Tiny, and It Converts Like Nothing Else (especially in legal)
This is the number that surprised me the most.
Referral traffic from ChatGPT, Gemini, Perplexity, and Claude has stayed relatively flat across these accounts over the past twelve months. What changed is the mix. ChatGPT’s share is down. Gemini, Perplexity, and Claude are all up. BrightEdge shows the same redistribution industry-wide, with Gemini climbing from 11.6% of AI referrals in Q1 2026 to 13.2% in April. People are spreading out across tools, not abandoning them.
Now the interesting part. In all three accounts I spot-checked, AI referral traffic had the highest key event rate of any channel. Not slightly higher. Multiples higher (like 500% + higher).
- AI referrals: roughly 12% to 16% key event rate (form fill, phone call, or chat)
- Organic search: roughly 2% to 4%
- Paid search: roughly 5%
Every published study we could find points in the same direction, even if the magnitude varies by industry. Semrush measured AI visitors converting at 4.4x the rate of standard organic. Opollo’s benchmark across 312 firms found 14.2% for AI referrals against 2.8% for Google organic. Seer Interactive clocked ChatGPT referrals at 15.9% versus 1.76% for Google organic. Ahrefs found that AI search made up 0.5% of their traffic but drove 12.1% of signups.
Most of that published research covers B2B software and ecommerce rather than legal, so it isn’t a perfect comparison. What matters is that our legal accounts land inside the same range as datasets running thousands of domains. Different vertical, same mechanism, same direction.
And the accounts posting the strongest numbers aren’t the average ones. They’re the firms with the heaviest investment in deep landing pages, structured data, and deliberate AI visibility work. Those accounts are beating the published benchmark, not just matching it.
The mechanism makes sense either way. Somebody who asks ChatGPT, “What should I do if I got charged with a DUI in Wisconsin and I have a prior?” has already done their research inside the chat window. By the time they click through to a firm’s site, they’ve narrowed the field. You’re not catching them at the start of a research process. You’re catching them at the end of one.
Traffic volume from AI platforms is still small. Conductor puts AI referrals at about 1.08% of total sessions across nearly 14,000 domains. But if 1% of your traffic converts at five times the rate of the other 99%, that’s not a rounding error. That’s the highest-value channel most firms aren’t measuring separately. If you learned anything reading through all this data nonsense, it’s that improving your AI visibility is and should be part of your focus for the rest of 2026, and years to come.
Does Content Still Matter?
Yes. Arguably more than it did two years ago, just not for the reason it used to.
Content used to earn clicks. Now it earns citations, and citations earn the clicks. Seer’s data shows brands cited inside an AI Overview pull roughly 120% more organic clicks per impression than uncited brands on the same query. You cannot be cited for something you never published.
Legal content sits in Google’s YMYL category, which means E-E-A-T isn’t a nice-to-have. The firms holding up best right now share a few habits:
- Every post has a job beyond traffic. The strongest content programs we run don’t publish standalone posts and hope. Each piece answers a real client question, links into the practice area and geographic pages it supports, and gets built out with an FAQ section an AI model can lift a clean answer from. A post might only pull a few hundred impressions on its own. The page it feeds is competing for terms worth real cases.
- Attorney attribution, not anonymous publishing. Named author, real bio, bar admissions, jurisdictions, years in practice, linked from every post. Plenty of firms work with legal writers to produce drafts, which is normal and fine. What matters is that a named attorney reviews it and stands behind it on the page. Content published under “Firm Staff” or no byline at all throws away the strongest expertise signal you have.
- Real case experience on the page. Statutory citations, local court procedure, county-specific filing realities. An AI model can summarize a general DUI explainer from fifty sources. It can’t invent what happens in a specific courthouse on a Tuesday.
- Depth on the pieces that matter. A 600-word post restating what every other firm in your market already published isn’t worth the slot. That same effort spent expanding an existing post with a real FAQ section, a video, or three more specific sub-questions usually returns more.
- Refresh instead of churn. Updating a page that already has authority beats publishing a new one that has none.
What to Track Instead of Total Impressions
If you take one thing from this, change how you report (or ask your vendor to, too).
Segment your Search Console queries into informational and commercial buckets, and monitor them separately. Observing a mixed number completely obscures the story. In GA4, create a separate channel group for AI referral sources, instead of having them show up under referral or direct. Then report on key event rate by channel, not sessions per channel.
If the conversion rate is rising and the traffic number is going down, that’s usually a good sign that the strategy is working. If impressions are the only number on the dashboard, you can’t see it.
The firms that lost impressions aren’t the firms that are losing now. They are the ones who cut content when the drop happened and didn’t get to be the source the AI cites.
Common Questions on Legal Content and SEO Strategy
Less than you were told five years ago, and better. Better two to four meaty posts a month than twelve thin ones, and the math on why is different. AI Overview: This is a description of a frequently asked question. It replaces 600 words. What page gets cited instead? A 1,500-word article with a named attorney author, real statutory citations, and details about how something plays out in your county.
The volume just doesn’t work anymore. If you are only measuring the number of posts published per month as a metric of success, you are measuring effort and not outcome. Ensure each post has citations, creates internal links to your practice area pages, and gets qualified people who call.
Start with the pages that an AI model can actually check. Named attorneys, bar admissions & jurisdictions on the page; clear entity relationships between the firm and its locations; structured data on every practice area and location page; answers written so a single paragraph can stand on its own when lifted out of context.
Then check whether you’re being cited at all. Ask ChatGPT, Gemini, Perplexity, and Claude the questions your clients ask, in your market, and see who comes back. Most firms have never run that test. It takes fifteen minutes, and it tells you more than a rank tracker will.
We put the full process in our AI Citation guidebook, including the page-level checklist and the prompts to test with.
Because those two numbers now measure different things. Impressions are dominated by informational queries, which is exactly where AI Overviews absorbed the clicks. Leads come from commercial and local searches, which barely trigger an overview at all. Losing the first while holding the second is a normal 2026 pattern, not a warning sign. Losing both is the warning sign.
Yes, just for a different reason. Google retired FAQ rich snippets in May so that you won’t see the dropdowns in search results anymore. The markup still tells machines that a specific block of text is a direct answer to a specific question, which is useful when an AI model is deciding what to quote. Keep it on practice area pages and long-form posts.
Build a custom channel group and route the AI hostnames into it: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and copilot.microsoft.com. Left alone, GA4 scatters these across referral and direct, and you never see the channel. Once it’s separated, report key event rate rather than sessions. The volume will look insignificant. The conversion rate won’t.
