Getting cited when clients ask the AI first

A general counsel (GC, the top in-house lawyer at a company) needs a firm for a messy cross-border restructuring. In 2020 she would have called two trusted partners and asked who they rate. In 2026 she opens an LLMLLMEin Large Language Model ist ein KI-System, das auf riesigen Textmengen trainiert wurde, um Sprache vorherzusagen und zu erzeugen. Damit sind Aufgaben wie Schreiben, Zusammenfassen und Beantworten von Fragen möglich.Vollständige Definition ansehen → (large language modellarge language modelEin Large Language Model ist ein KI-System, das auf riesigen Textmengen trainiert wurde, um Sprache vorherzusagen und zu erzeugen. Damit sind Aufgaben wie Schreiben, Zusammenfassen und Beantworten von Fragen möglich.Vollständige Definition ansehen →, the technology behind tools like ChatGPT, Claude, and Gemini) and types: "Who are the best firms for a cross-border restructuring involving German and Brazilian entities?"

The model answers in seconds. It names three firms and two individual partners. Those names came from somewhere. This lesson is about making sure your firm is one of them.

Why this changed how firms get shortlisted

The old buying journey started with a human recommendation, then moved to the firm website and league tables. The new journey often starts with a synthesis: the LLMLLMEin Large Language Model ist ein KI-System, das auf riesigen Textmengen trainiert wurde, um Sprache vorherzusagen und zu erzeugen. Damit sind Aufgaben wie Schreiben, Zusammenfassen und Beantworten von Fragen möglich.Vollständige Definition ansehen → reads across thousands of sources and hands the buyer a pre-filtered shortlist.

That matters because of a simple behavioral fact: the shortlist anchors everything after it. If your firm is not in the model's first answer, you are now arguing your way onto a list the client already trusts.

Two things decide whether you appear:

1. What the model was trained on (or can retrieve in real time via web search).
2. How authoritative and specific your firm looks on the exact topic being asked about.

Both are influenceable. Neither is magic. It is marketing, applied to a new reader: the machine.

How LLMs actually pick names

You do not need to understand transformers to market well here. You need to understand three practical mechanics.

1. Frequency plus specificity

Models surface entities that appear often AND in context. A firm mentioned once as "a top restructuring shop" is weaker than a firm mentioned across fifty client alerts, ranking directories, and conference agendas specifically tied to "cross-border restructuring" and "Chapter 15" (the US law provision for recognizing foreign insolvency proceedings).

The lesson: own a narrow topic loudly, rather than claiming everything quietly.

2. Structured, quotable sources

Models favor content that is clean, attributed, and easy to extract. League tables and directories like Chambers and Partners and The Legal 500 are gold because they are structured, ranked, and widely cited. When a directory says "Band 1, Restructuring, Partner Jane Okafor," that is a clean fact the model can repeat.

3. Retrieval-augmented answers

Many 2026 legal buyers use tools that search the live web before answering (retrieval-augmented generation, or RAG). This means fresh content counts. A client alert you published last week on a new EU insolvency directive can be pulled into an answer today, even if the base model never "learned" it.

The three assets that get you cited

Thought leadership that names a human

LLMs cite people, not just logos. A GC asking "who should I call" wants a name.

Publish under named partners, consistently, on a defined topic. A partner who writes four substantive pieces a year on cross-border restructuring, always bylined, builds a machine-readable identity: "This person equals this expertise."

Concrete moves:

• Byline every piece with the partner's full name and practice area.
• Repeat the topic phrase the buyer would type ("cross-border restructuring," "distressed M&A," "scheme of arrangement," the UK court process for restructuring debt).
• Publish somewhere crawlable: your firm site, LinkedIn articles, and reputable legal media, not a gated PDF the model cannot read.

Gated content is invisible to most crawlers. If the model cannot open it, it cannot cite it.

Client alerts as timely signals

Client alerts (short updates a firm sends when a law or ruling changes) are your fastest citation lever. When a regulator issues new guidance, the firm that publishes a clear alert within 48 hours becomes the source everyone else, humans and models, quotes.

Make alerts extractable:

• Lead with the change and its practical effect, not "We are pleased to note."
• Use a clear headline with the actual topic and jurisdiction.
• Include a named contact partner at the bottom.

An alert titled "New German StaRUG Amendments: What Cross-Border Groups Must Do Now" is far more citable than "Recent Developments in Restructuring Law." (StaRUG is Germany's preventive restructuring framework.)

League-table and directory presence

This is the highest-trust signal a model can find, because it is independent and structured.

The work here is submission discipline. Directories run annual research cycles with deadlines. Firms that submit strong, well-evidenced entries (with client referees and specific deal matters) get ranked. Firms that skip the cycle vanish from the exact structured sources the model trusts most.

Practical checklist:

• Track Chambers and Legal 500 deadlines by practice group.
• Name individual partners in submissions, not just the firm.
• Secure client referees early; they carry disproportionate weight.

Measuring whether it is working

You cannot manage what you do not test. Run a simple quarterly audit.

Pick the ten questions your ideal clients would actually ask an LLMLLMEin Large Language Model ist ein KI-System, das auf riesigen Textmengen trainiert wurde, um Sprache vorherzusagen und zu erzeugen. Damit sind Aufgaben wie Schreiben, Zusammenfassen und Beantworten von Fragen möglich.Vollständige Definition ansehen →. Prompt several models with each. Log the results.

Prompt: "Best law firms for cross-border restructuring in Germany and Brazil?"

Track per model:
- Is our firm named?           Y / N
- Is a named partner cited?    Y / N
- Position in the list         1 / 2 / 3 / not listed
- What source did it cite?     (directory / alert / news / none)

Do this across three or four LLMs, because they draw on different sources and update on different schedules. Repeat every quarter. Rising presence is your KPIKPIKey Performance Indicator, ein messbarer Wert, der zeigt, wie wirksam Sie ein bestimmtes Ziel erreichen, über die Zeit verfolgt und gegen einen Zielwert gemessen.Vollständige Definition ansehen → (key performance indicatorkey performance indicatorKey Performance Indicator, ein messbarer Wert, der zeigt, wie wirksam Sie ein bestimmtes Ziel erreichen, über die Zeit verfolgt und gegen einen Zielwert gemessen.Vollständige Definition ansehen →, a metric you track to judge success).

One caution: model outputs vary run to run, and firms cannot verify training data. Treat this as directional evidence, not a precise scoreboard. And never claim a ranking the model invents; models sometimes fabricate ("hallucinate") citations, so verify anything you repeat in your own marketing.

Putting it together: a 90-day plan

You do not need a new department. You need focus on one practice area first.

Weeks 1 to 2: Pick the battle. Choose one topic where you are genuinely strong and where clients ask LLMs first (restructuring, data privacy, and cross-border M&A are common). Write the ten buyer prompts. Run your baseline audit.

Weeks 3 to 6: Fix the assets. Ungate your best thought leadership. Add named bylines and clear, topic-specific headlines. Confirm your site is crawlable (no blanket blocking of bots in your robots.txt file unless legal reasons require it).

Weeks 7 to 10: Publish on rhythm. Ship two sharp client alerts on live developments. Get one bylined partner article into reputable legal media. Consistency beats volume.

Weeks 11 to 13: Lock the directories. MapMapEinsatz von Software, um wiederkehrende Marketingaufgaben und Kampagnen zu automatisieren und Personalisierung in großem Maßstab über Kanäle wie E-Mail, Web und Social zu ermöglichen.Vollständige Definition ansehen → submission deadlines. Draft entries naming specific partners and matters. Line up client referees.

Then re-run the audit and compare.

A word on ethics and accuracy

Marketing rules still apply. Most bar associations (the bodies that regulate lawyers) prohibit false or misleading claims and unverifiable comparative statements like "best" without substantiation. Do not seed AI-facing content with claims you could not defend in a print advertisement. The channel is new; the professional conduct rules are not.

If you are unsure whether a claim complies, treat it the way you would any advertising: check it against your jurisdiction's rules before publishing.

Key Takeaways

• The shortlist forms before the phone rings. Buyers now ask an LLMLLMEin Large Language Model ist ein KI-System, das auf riesigen Textmengen trainiert wurde, um Sprache vorherzusagen und zu erzeugen. Damit sind Aufgaben wie Schreiben, Zusammenfassen und Beantworten von Fragen möglich.Vollständige Definition ansehen → first, so your goal is to be one of the two or three names it surfaces on your core topics.
• Be narrow and loud. Frequency plus specificity wins. Own a defined topic across bylined articles, timely alerts, and directory rankings rather than claiming broad, generic expertise.
• Name the humans. Models cite people. Consistent partner bylines and named directory rankings turn individuals into machine-recognizable experts.
• Make content extractable. Ungated, crawlable, clearly headlined content gets cited; gated PDFs and vague titles do not.
• Audit quarterly, and stay compliant. Test real buyer prompts across several models to track your presence, verify any citation before reusing it, and keep every claim within your bar's advertising rules.