The verdict in three sentences
A general counsel in Kendall Square facing a licensing dispute with a European partner now asks an AI assistant like ChatGPT which Boston firms handle international arbitration, before even calling their network. The firms that get named are those whose partners, seats, institutions and sectors are described precisely and consistently, on the site and by third parties, with no superlatives. No provider can promise a recommendation, but a dated record of answers shows what AI assistants say about you and what needs fixing.
The questions buyers ask, stage by stage
Questions follow the life of a matter. Realistic examples for the Boston market in 2026:
| Stage | Typical question to an AI assistant | What the assistant looks for |
|---|---|---|
| Contract drafting | 'Which Boston firm drafts arbitration clauses for biotech licensing deals with Swiss partners?' | Corporate and life sciences pages |
| Dispute arises | 'Boston lawyers experienced in ICC arbitration for pharmaceutical licensing disputes' | Seats, institutions, sectors, named partners |
| Investment arbitration | 'Boston firms that have represented parties in ICSID arbitrations' | Public experience, publications |
| Languages | 'Arbitration lawyer in Boston who speaks Portuguese for a construction dispute in Brazil' | Working languages per partner |
| Budget | 'How much does international arbitration counsel in Boston cost for a USD 50 million dispute?' | General explanations of fees and billing models |
| Enforcement | 'Who in Boston enforces a foreign arbitral award in Massachusetts courts?' | Award enforcement pages, New York Convention |
| Comparison | 'Boston or New York firm for a cross-border construction arbitration?' | Factual positioning, team size |
If your pages stay generic ('we guide clients through complex disputes'), the assistant will cite a more explicit competitor.
What gets a firm cited in an answer
| Signal | What the assistant understands | Common gap at firms |
|---|---|---|
| Detailed partner bios | Who argues, under which rules, in which languages | Three-line bios, no languages |
| Pages by institution and seat | The firm practices before the ICC, LCIA, SIAC, HKIAC or ICSID | One catch-all 'Arbitration' page |
| Sector pages | Life sciences, energy, construction, technology | Sectors listed with no content |
| Third-party mentions | Articles, conferences, publications, real professional directories | No trace outside the site |
| Data consistency | Same partners and offices everywhere | A partner who left in 2024 still online |
| Structured data | Organization, people, addresses identified | Missing or broken markup |
| Freshness | Recent, dated analysis pieces | Last publication two years ago |
Ethics: what AI visibility does not change
Lawyer advertising rules still apply to everything you publish. In the US, state bar rules modeled on the ABA Model Rules prohibit false or misleading communications about a lawyer's services. In practice: no matter shown without client consent or before it is public, no invented ranking or award, no promise of outcome, care with words like 'specialist' or 'expert'. On the site itself, an accessibility target of WCAG 2.1 or 2.2 AA reduces exposure to ADA Title III suits, which are frequent in the US, and contact forms should stay lean on the data they collect. This is not legal advice: your ethics partner signs off on every piece of content.
There is good news here too: AI assistants reward exactly what the ethics rules require, verifiable facts rather than superlatives.
Market budget and how results are measured
The market is new and prices move. A 2026 ballpark based on typical quotes from agencies and consultants who serve US law firms:
| Line item | Scope in broad terms | 2026 ballpark (USD) | Indicator tracked |
|---|---|---|---|
| Diagnostic | 20 to 50 questions recorded, gaps, priorities | 4,000 to 15,000 | Share of answers naming the firm |
| Bio and practice page rework | Partners, institutions, sectors, languages | 6,000 to 30,000 (project) | Answer accuracy |
| Publications and third-party mentions | Articles, speaking, PR | 3,000 to 12,000 a month | Sources cited by assistants |
| Technical and structured data | Markup, consistency, performance | 2,000 to 8,000 (project) | Data errors fixed |
| Monthly dated record | Same questions, raw answers archived | 300 to 1,500 a month | Quarterly trend |
At Kolonell, this kind of engagement is on quote, after a diagnostic. Measurement rests on three numbers: the share of answers that mention the firm, the accuracy of what is said (partners, languages, institutions) and the kind of sources cited. These are cross-checked with inbound enquiries and the 'how did you hear about us?' question asked on the first call.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Mini case study
Illustrative example: Commonwealth Wharf Partners LLP, 12 lawyers in Boston's Financial District, with an arbitration and corporate practice. Across 25 recorded questions, the firm appears in 3 answers, or 12%, and 4 answers name a partner who left in 2024.
Budget under review: a USD 10,000 diagnostic, then 6 months of support at USD 5,000, so 10,000 + 30,000 = USD 40,000. If a mid-size commercial arbitration represents USD 400,000 in fees (market assumption), the budget equals 10% of a single matter. The partners approve, with a monthly record on the same 25 questions and a month-3 review focused on answer accuracy.
FAQ
What does an AI visibility diagnostic cost for an arbitration firm in Boston?
In 2026, expect roughly USD 4,000 to 15,000 depending on the number of questions (20 to 50) and practices covered. The monthly record that follows often costs USD 300 to 1,500.
Do the ethics rules prevent this work?
No, they frame it: no non-public matters, no invented rankings, no promise of outcome. Factual content across 5 or 6 practice pages respects those rules and feeds AI answers.
Do we need a page per arbitral institution?
Often yes, if the experience is real: a well-documented ICC page and ICSID page answer better than a generic one. Three to five targeted pages are enough for most firms of 10 to 30 lawyers.
How long before a visible change?
Bio corrections sometimes show within a few weeks, third-party mentions within 3 to 6 months. The dated record should be read over at least two quarters.
Can we be sure ChatGPT will recommend us?
No, nobody controls an AI assistant's answers. The work improves the quality of public information and measures change honestly on the same questions.
Find out how AI assistants describe your firm. We record what AI assistants like ChatGPT say today about your partners, institutions and sectors, then size the corrections worth making. Request a diagnostic quote: reply within 48 hours.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
