What law firms reveal about marketing in the AI era

Intro

Marketers tend to look for AI lessons in the obvious places: the companies shipping the most content, running the boldest experiments, posting the loudest results. Regulated professions rarely make that list. They move slowly, they buy cautiously, and they talk about risk before they talk about reach.

That is exactly why they are worth watching. A law firm cannot afford to be confidently wrong, and neither, increasingly, can a search result. As AI overviews and chat assistants take over more of the discovery layer, the constraints that have always governed legal marketing, namely accuracy, trust and demonstrable expertise, are becoming the constraints that govern everyone's.

In other words, the most cautious buyers in the market are running the experiment the rest of us are about to be graded on.

How law firms actually use AI, and the bar they hold it to

Inside a firm, AI is already doing real work, just not the work the headlines imply. The practical uses are narrow and supervised: summarising long documents, surfacing relevant case law, producing first drafts of routine correspondence, and triaging inbound enquiries so a paralegal is not reading every form by hand.

What is striking is the acceptance threshold. A marketing team might tolerate an AI tool that is right eighty per cent of the time, because the cost of a weak subject line is trivial. A litigator cannot. A fabricated citation is not a quirky failure mode, it is a professional conduct problem, and there have been enough sanctioned cases of invented case law to make every partner nervous.

So firms have settled on a pattern worth copying. AI accelerates the first ninety per cent and a qualified human owns the last ten, which is the part that carries liability. The tool drafts, the lawyer verifies, and the verification is non-negotiable. That division of labour is not timidity. It is an honest read of where current models are reliable and where they are not.

How their clients now search, and what it does to discoverability

The bigger shift is happening outside the firm, in how prospective clients look for help. Someone with a problem used to type "unfair dismissal lawyer Melbourne" and scan a page of blue links. Now a meaningful share of them ask a question in full sentences, "can my employer make me redundant while I am on leave", and read the synthesised answer before they ever see a directory of firms.

This changes the discoverability problem in two ways. First, the answer is increasingly assembled by a model rather than chosen by a click, so being on page one matters less than being the source the model trusts enough to quote. Second, the queries are longer, more specific and more situational, which rewards content that genuinely resolves a narrow question over content that targets a broad head term.

For a regulated firm this is awkward, because the honest answer to most legal questions is "it depends". But that constraint produces better material. The pages that get cited tend to be the ones that lay out the actual considerations, cite the relevant legislation, and are clearly attributable to a named, credentialed author. Vague, hedged, anonymous content does not survive the summarisation step.

Why high-trust verticals reward expertise over volume

For a decade, a lot of SEO advice quietly assumed that more was better: more posts, more pages, more coverage of every long-tail variation. In low-stakes categories that often worked. In high-trust verticals it never really did, and AI search is now exposing why.

When a model assembles an answer about a medical, financial or legal question, it is making an implicit bet about which sources are safe to repeat. The signals it leans on are the ones that map to genuine credibility: clear authorship, professional credentials, citations to primary sources, consistency with other reputable material, and a track record of being right. None of those are gameable at volume. You cannot mass-produce demonstrable expertise.

This is also where firms are quietly upgrading their stack. The AI tools for lawyers that get adopted are the ones built around verification and source attribution rather than raw output speed, because in this setting a fast wrong answer is worse than no answer at all. The buying criteria tell you something. When the cost of error is high, people stop optimising for quantity and start optimising for provenance.

The lesson generalises cleanly. As AI mediates more queries across every category, the sites that win citations will be the ones that look, to a model, like the sources a cautious professional would rely on.

The practical lesson for anyone optimising for AI search

You do not need to market a law firm to use what they have learned. A few moves travel well to any vertical that AI search is reshaping.

• Put a real, credentialed author on substantive pages, and make the credentials checkable. Author entities are becoming a trust signal models can read.
• Answer the specific question the way a knowledgeable person actually would, including the caveats. Hedged generality reads as low confidence to both humans and models.
• Cite primary sources and link to them. Provenance is what makes content safe for a model to repeat.
• Treat factual accuracy as a ranking input, not just an editorial nicety. Being wrong is now a discoverability cost, not only a reputational one.
• Stop measuring content by raw output. A handful of genuinely authoritative pages will out-earn a hundred thin ones in AI-mediated discovery.

None of this is exotic. It is closer to old-fashioned editorial discipline than to anything labelled growth, which is rather the point.

Where this leaves the rest of us

The useful thing about regulated professions is that they have been forced to confront, early and under real consequences, the question the rest of the market is only starting to ask: what does AI have to get right before you are willing to put your name on its output?

Their answer has been consistent. Use the tools where they are strong, verify where they are weak, and never let speed override accuracy in a context where being wrong is expensive. AI search is steadily making "wrong" expensive for everyone, because a model that will not trust your content will not surface it, and a reader who catches one fabricated fact will not return.

Law firms reached that conclusion before most marketers because they had no choice. The smarter move is to reach it on purpose, while it is still a competitive advantage rather than a remedial scramble.