Google's AI Threat to Law Firms: What Gemini Enterprise for Legal Actually Changes

Google Cloud launched Gemini Enterprise for Legal on 25 August 2026. It is in preview, it ships with legal-specific agents, and Cleary Gottlieb, Freshfields, Weil and Williams & Connolly were named as launch firms (Artificial Lawyer, Unite.AI).

Within a day the coverage had settled into the usual shape: Google is coming for legal work, firms must respond, everything changes. That framing is not wrong, but it is imprecise in a way that leads firms to solve the wrong problem.

There are two separate threats here and they arrive on different timelines, at different firms, through different doors. Conflating them is how a five-attorney injury practice ends up worrying about brief-drafting agents while quietly losing the search visibility that actually feeds it.

What Google actually shipped

Not a separate application. Google is explicit that this is a legal plugin inside the existing Gemini Enterprise app, which matters because it means the entry point is an enterprise agreement a firm already has rather than a new product to evaluate.

The pre-built agents and skills cover:

• Legal brief drafting
• Citation verification
• Contract lifecycle management
• Regulatory horizon scanning
• Data subject access request (DSAR) fulfilment
• Motion to seal preparation
• NDA drafting

The integrations tell you more about the target customer than the feature list does. Named connectors include iManage and NetDocuments for document management, Docusign for agreement metadata, Everlaw and RelativityOne for e-discovery, Thomson Reuters for collaboration content, the Free Law Project's CourtListener for federal and state opinions and PACER dockets, plus Harvey, Legora, Solve Intelligence, Eudia and Courtroom5.

Thomas Kurian, Google Cloud's CEO, framed it as helping firms "manage the complex realities of modern legal work, while simultaneously providing critical grounding and governance capabilities."

Pricing was not disclosed. This is also Google's first industry-specific packaging of Gemini Enterprise, launched alongside a financial services edition, with healthcare and life sciences described as next.

Who this is actually built for

Read the integration list again. To get value from a citation-verification agent you need a document management system for it to read. To get value from an e-discovery agent you need RelativityOne or Everlaw. To get value from a regulatory horizon-scanning agent you need a regulatory exposure worth scanning for.

That is a description of large firms and corporate legal departments. The four named launch firms are all large international practices. Nothing here is priced, packaged or integrated for a small consumer-facing practice, and the near-term competitive pressure lands accordingly.

Threat one: AI doing legal work

This is the threat the headlines mean, and it is real, but it is narrower than it sounds and it is not new.

What these agents automate is document-heavy, procedure-heavy, high-volume work: first-draft NDAs, citation checks, contract review at scale, regulatory monitoring. That work has been under margin pressure for a decade from offshore review, contract lifecycle software and legal process outsourcing. Google has made the pressure faster and better funded. It has not introduced it.

It is also not unique to Google. Anthropic has Claude for Legal, OpenAI is building a legal vertical, Microsoft has a Legal Agent, Palantir is working with Kirkland, and Perplexity has moved on the sector too. When every major model vendor ships the same category in the same year, that is a market forming, not one company taking a sector.

For a consumer-facing practice, the honest read is this. The client with a broken wrist, a DUI charge or a custody dispute is not buying document production. They are buying judgment, advocacy and someone who answers the phone at a bad moment. No agent in that feature list touches the part they are actually paying for.

Threat two: AI answering the question your website used to answer

This is the one that reaches every firm, and it is already happening.

The same class of model powering Gemini Enterprise for Legal also powers the answer that appears above the search results, and the assistant a prospective client opens before they ever run a search. When someone asks how long they have to file a claim after a car accident, or what happens at a first DUI hearing, they increasingly get an answer rather than a list of links.

That changes the competitive question. It is no longer only "does my page rank for this term." It is "when an answer engine explains this, is my firm the source it draws on and names."

For a law firm this cuts both ways.

The bad half. Informational traffic that used to land on your practice-area pages gets answered upstream. A blog post that existed to capture "how long do I have to file" will bring fewer clicks than it did in 2023.

The good half, which most firms are missing. Legal queries carry unusually high commercial intent and unusually high trust requirements, so answer engines are conservative about what they cite for them. They favour sources with clear authorship, verifiable expertise, specific jurisdictional detail and an accessible structure. That is a smaller field to compete in than the open web, and a firm with genuine depth on a narrow practice area can win it in a way it could never win a national head term.

What to actually do about it

Nothing on this list is exotic. Most of it is what good law firm SEO already required, which is the point.

1. Check that AI crawlers can reach you

Many firms block them without ever deciding to, through a CDN or security setting that came on by default. Check for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended in both robots.txt and your CDN bot rules. Whatever you decide, decide it on purpose.

2. Answer the question in the first two sentences

Answer engines lift passages, not pages. A section that opens with two sentences of preamble before it gets to the point does not get quoted. One that opens with a direct, complete, self-contained answer does. This is a writing discipline, not a technical one.

3. Be specific where a model cannot generalise

A model can produce a generic explanation of comparative negligence in seconds. It cannot produce the filing deadline quirks of your specific county, the local judge's scheduling habits, or what actually happens at the courthouse you practise in. Jurisdictional specificity is defensible ground precisely because it is the part that is hardest to synthesise.

4. Make your entity signals unambiguous

Consistent firm name, address and phone across the web, accurate schema for the organisation and its attorneys, real author attribution on content, and a Google Business Profile that matches everything else. An answer engine that cannot tell who wrote something and whether they are credible will cite someone it can identify instead.

5. Stop measuring only clicks

If AI answers reduce informational clicks while your consultations hold steady or rise, your reporting will show a decline that your bank account does not. Track branded search volume, direct traffic and enquiry quality alongside sessions, or you will optimise your way out of work that is functioning.

The full method is on our AI search optimization page, and it shares a foundation with the law firm SEO work rather than replacing it.

The budget question

The queue of coverage around this launch pushes firms toward a reallocation. Before you make one, be clear about which threat you are responding to.

If you are a large firm with a document-heavy practice, the delivery-side question is a technology and staffing question, and it belongs with your operations budget rather than your marketing budget.

If you are a consumer-facing firm, the delivery-side threat is mostly noise for now, and the demand-side threat is the one with your name on it. The response is not a bigger budget. It is a better-weighted one: substantive content over thin content, technical foundations over volume, and a website you own over channels you rent. We set out what that costs on our pricing page, published rather than quoted on a call.

For a fuller treatment of how much to spend in the first place, see what percentage of revenue should go to marketing, by practice area.

The honest summary

Google shipping legal AI agents to four of the largest firms in the world is genuinely significant, and it is significant mostly for firms that look like those four.

The change that reaches everyone else is quieter and older: search is becoming answering, and the firms that get named in answers will be the ones with real depth, clear authorship and an open door to the crawlers. That work was worth doing before 25 August 2026. It is just harder to postpone now.

Inovista builds law firm websites and the organic and AI search visibility behind them. We do not sell Google Ads, which is worth knowing when you read anyone's take on where your budget should go. If you want a straight read on whether your site is positioned for AI search, get in touch and we will tell you what we see.