Global law firms: which ones ChatGPT, Gemini and Google AI Mode recommend

We asked ChatGPT, Gemini and Google AI Mode 12 questions that in-house legal teams of large companies ask when they choose an international law firm: cross-border M&A, banking, litigation, tax, employment, global coverage. Every answer names at least one of the 11 firms we tracked. A handful of transactional firms take most of the answers, and the global network firms come up mainly when the question is about size and reach.

Latham & Watkins is named in three answers out of four

Latham & Watkins is named in 27 of the 36 answers, and in 11 of Gemini's 12. A&O Shearman follows with 20, then Clifford Chance (16) and Freshfields (15). Kirkland & Ellis and Skadden come up less often; on debt restructuring for private equity, all three AI models name Kirkland & Ellis. On cross-border banking and finance, the three models give the same three names: Clifford Chance, Latham & Watkins and A&O Shearman.

1. Latham & Watkins27 · 75%
2. A&O Shearman20 · 56%
3. Clifford Chance16 · 44%
4. Freshfields15 · 42%
5. Kirkland & Ellis11 · 31%
6. Skadden9 · 25%

The global networks are named for their footprint

Baker McKenzie is the network firm named most, in 18 answers, ahead of DLA Piper (16). Hogan Lovells, Dentons and Norton Rose Fulbright follow far behind. Dentons, the largest firm in the world by number of lawyers, is named in 6 answers. When it is named, it is usually first (average position 1.5): on the question about deeply local advice backed by global scale, all three AI models open with Dentons. On the other eleven questions, it comes up three times.

1. Baker McKenzie18 · 50%
2. DLA Piper16 · 44%
3. Hogan Lovells7 · 19%
4. Dentons6 · 17%
5. Norton Rose Fulbright4 · 11%

ChatGPT, Gemini and Google AI Mode, October 22, 2026. One answer can name several global network firms.

ChatGPT, Gemini and Google AI Mode

The three models agree on the leaders but not on the rest. Latham & Watkins is in 6 of ChatGPT's answers and 11 of Gemini's. Norton Rose Fulbright is named by ChatGPT only (4 answers). Google AI Mode names an elite transactional firm in all 12 of its answers.

What each kind of firm is named for

On the three finance questions (ESG and sustainable finance, banking regulation, debt restructuring), none of the 9 answers names a global network firm. On global footprint, all 6 answers do. Specialist practices (real estate, IP and tech, employment, tax) are shared: Baker McKenzie leads on employment and tax, Latham & Watkins on IP and tech.

Chambers is the source behind the answers

The answers cite 65 different websites. chambers.com is cited in 25 of the 36 answers, far ahead of everything else; legal500.com comes up in 5. The firms' own sites follow: lw.com in 11 answers, bakermckenzie.com in 10, aoshearman.com in 9. Rankings and the firms' practice pages are what the AI models read.

Elite firm websiteNetwork firm websiteOther website

1. chambers.com25 · 69%
2. lw.com11 · 31%
3. bakermckenzie.com10 · 28%
4. aoshearman.com9 · 25%
5. dlapiper.com6 · 17%
6. freshfields.com6 · 17%
7. cliffordchance.com5 · 14%
8. legal500.com5 · 14%
9. skadden.com5 · 14%
10. dentons.com4 · 11%

What this means for law firms

• Directory rankings carry the answers. Chambers is cited in more than two answers out of three. A firm's band in each practice area is the first thing an AI model reads before it recommends anyone.
• Size alone does not get a firm named. The largest network is named in 6 answers out of 36. For each practice, the answer goes to the firm whose pages and rankings describe that practice, not to the one with the most offices.
• Each firm is known for a few things. Finance and restructuring go to the transactional firms, footprint to the networks. A firm that wants to be named on another subject needs pages, matters and rankings on that subject.
• The models differ. A firm named by ChatGPT can be missing from Gemini, and the reverse. Checking one assistant gives a partial picture.

Method

12 questions in English, written as an in-house legal team at a large company would ask them, from the practice areas of one of the firms tracked (Dentons). Each question was asked once to ChatGPT (gpt-6, with web search), Gemini (Flash-Lite) and Google AI Mode, from the US, on October 22, 2026: 36 answers. We counted the answers naming each of 11 tracked firms (names and common spellings, Freshfields Bruckhaus Deringer counted as Freshfields), their position among the tracked names of the answer, and the websites cited.

This is a snapshot: three models, one day and 12 questions. AI answers vary from one session to the next, so a difference of one or two answers between two firms is not significant. Firms we did not track, such as Linklaters or White & Case, are not counted.