Your website might be winning on Google and losing everywhere else

A general counsel has an urgent matter. 10 years ago, they'd have called a colleague for a recommendation. Today, before they've even opened a browser tab to search, they open ChatGPT and type something like: "Which firms have deep expertise in M&A with a track record in the mid-market?"

The model answers with confidence. Three or four firm names. No links, no page to scroll through, just a synthesised answer built from thought leadership, press mentions, directory rankings and attorney bios it has already read.

If your firm isn't one of the names in that answer, you don't find out. There's no failed click, no bounce, no line in Google Analytics that flags it. You simply don't get the call.

Start here: run the test yourself

Before anything else, do this. Open ChatGPT, Perplexity or Gemini and ask it a question a real client might ask, not "who are the best firms for X" but something closer to how a client actually thinks: "what compliance gaps commonly trip up investment products" or "what clauses tend to cause post-closing disputes". See whether your firm appears, and in what context.

This takes five minutes and it tells you more than most quarterly SEO reports. If your firm shows up, you'll know roughly what it's being credited for. If it doesn't, you've just found the blind spot most law firm marketing teams don't know they have: not a ranking problem, but a visibility problem in a channel nobody's been measuring.

Everything below explains why that test matters, and what to do once you know the answer.

What's actually changed

Search hasn't disappeared. But a growing share of legal research and firm discovery now happens inside AI tools that answer the question directly instead of sending someone to a results page. Marketers have started calling this GEO: generative engine optimisation. In simplified terms: traditional SEO earns you a place on a page of links. GEO earns you a mention inside the answer itself. The mechanics are related (both depend on structure, authority and trust signals), but GEO adds a layer that most firm websites weren't built for. At this year's Legal Marketing Association Annual Conference, Jennifer Lill of One North put it plainly: the firms that win now are the ones that treat this as measurable, not mysterious. She told the room that the priority now is investing in the ability to learn, meaning benchmarks, dashboards and a habit of checking, not a one-off audit.

Why some firms show up and most don't

Once you know whether you appear, the more interesting question is why. A separate LMA session, led by Robyn Addis and Joseph Giovannoli of 9Sail, unpacked what's actually happening behind the scenes when a model decides which firms to name. The answer isn't mysterious, and it isn't really new; it's the same trust signals search engines have rewarded for years, just read by something that can't ask a follow-up question.

A model can only cite what it can verify. That's the whole logic of it. Schema markup, the code that tells a machine "this page is about M&A, this person is a partner, this firm is based here", exists precisely so a system doesn't have to guess. Most firm websites have some of this. Very few have it consistently across every practice page and every attorney bio, which means the model ends up more confident about some parts of the firm than others. Confidence, not accuracy, is what gets a firm named.

The same logic explains why some content earns a citation and other content gets skipped entirely. A page that opens with the answer, in plain language, in the first two sentences, is easy for a model to lift and attribute. A page that opens with three paragraphs of firm history before it gets to the point is not. And original material (a proprietary survey, a jurisdiction-specific breakdown, a genuinely distinctive case study) gets cited for the same reason a journalist quotes it: because nowhere else has it.

There's a quieter signal too, and it's the one firms most often get wrong without realising: identity consistency. If a firm's name, address and phone number appear slightly differently on its own website, its Google Business Profile, and half a dozen legal directories, a model reads that as ambiguity rather than error. It doesn't flag the inconsistency. It just becomes less sure which listing is authoritative, and quietly favours a competitor whose details all agree with each other.

Underneath all of it sits something less glamorous: whether the site can be read cleanly in the first place. Slow pages, tangled site architecture, a robots.txt file blocking the wrong crawlers. These don't just cost a firm in traditional SEO. They mean an AI system can't fully index what's there, so it works from an incomplete picture of a firm that might otherwise be exactly the right fit for the question being asked.

Why this belongs on the leadership agenda, not just the marketing one

Managing committees think in billable hours and origination credit, not citation rates. Two arguments tend to land.

The first is about return. 

An AI recommendation reads as an independent endorsement rather than an advert. The audience is smaller than a paid search campaign, but it arrives already convinced, which tends to mean better-qualified enquiries at a lower cost per lead.

The second is about risk.

AI tools are already describing your firm to prospects, whether or not you've checked what they're saying. Wrong practice areas. Outdated credentials. An attorney confused with someone at a different firm entirely. Without measurement, a firm has no way of knowing how it's being represented in a conversation it isn't part of.

The idea worth sitting with

Most firms still measure their website by whether people click on it. That's the wrong unit now. The click is being replaced, in a growing number of cases, by the citation: a firm's name surfacing inside an answer, with no click involved at all. A website can be technically excellent by every old metric and still be invisible in the one channel that's quietly deciding who gets shortlisted before the RFP is even drafted.

That's not a reason to panic or to rebuild everything this quarter. It's a reason to start treating the website the way the best-run firms already treat their finances or their risk register: as something you check on deliberately, on a schedule, rather than something you built once and assumed would keep working. The firms doing that now aren't necessarily the biggest, or the ones with the flashiest sites. They're the ones who ran the five-minute test above, didn't like everything they found, and treated that as useful information rather than bad news.