- Professional services firms get recommended by AI when a model can describe them in one confident sentence and find two independent sources that agree.
- Gartner’s 2025 survey of 645 B2B buyers found that 45% used generative AI during a recent purchase, mainly to gather information on vendors and products.
- The Describability Test (Exhibit 1) gives five questions that show how clearly a firm reads to a machine.
- Partners are the product in a firm that sells judgement, so each partner needs a consistent, findable public profile.
- Confidential work still produces public evidence through mandates, publications, speaking, awards and rankings.
What does it mean to be recommended by AI?
Being recommended by AI means that an AI assistant names a firm in answer to a buyer’s question, such as "which Singapore law firms advise on cross-border funds" or "who can help us restructure a regional business". The answer is assembled from what the model has learned and, for tools with live search, from sources it retrieves at the time. The firm does not submit an application. The model includes the firms whose descriptions it can state with confidence.
This differs from search ranking. A search engine returns a list of pages and the buyer chooses which to open. An AI answer returns a short list of names with a sentence on each, and the buyer often starts the conversation with those firms. The shortlist forms before any firm’s website is visited.
For firms that sell judgement, the effect is concentrated. A buyer cannot inspect legal advice, strategy or architecture before buying it, so the buyer relies on reputation signals. An AI answer is a compressed reputation signal, and the model builds it from whatever the public record says.
Professional services buyers increasingly meet a shortlist before they meet a firm
Gartner’s May 2026 release reports that 45% of buyers used generative AI during a recent purchase, primarily to gather information on vendors and products. Buyers in the same survey reported using an average of seven information sources in a recent purchase (Gartner, 20 May 2026).
The same release shows where human contact sits in the process. Sixty-nine per cent of buyers prefer to validate AI-generated insights with sales representatives, and Gartner advises sellers to engage where they can help buyers validate information, reduce risk and move forward with greater confidence. For a professional services firm, this describes a two-step sequence. The AI answer builds the shortlist, and the partner meeting confirms it.
The survey covers B2B buyers in general and does not isolate law, accountancy, consulting or architecture. The direction is clear, and the exact share for each profession is not yet measured.
The AI answer builds the shortlist. The first partner meeting confirms it.
A model can only recommend what it can describe, and a clear description starts with the firm
A model describes a firm by combining the firm’s own pages with what others have written about it. The clearer and more consistent those materials are, the more confident the description. A firm that says it offers "integrated solutions for complex challenges" gives the model nothing specific to repeat. A firm that says it advises family-owned manufacturers on cross-border expansion in Southeast Asia gives the model a sentence it can use.
Vantage’s own tracking shows the mechanism from the other side. In its AI visibility monitoring, an engine described Vantage as an FMCG and packaging agency, repeating wording that a third-party directory had copied from another listing. The correct description exists on Vantage’s own site. The engine reproduced the directory’s version because the directory was an independent source and the model treated it as corroboration. For a deeper account of why models weight outside sources this way, see why AI does not cite your website.
The practical lesson is that a description needs to exist in one clear form across every place the firm appears. Our guide to getting a brand mentioned by ChatGPT covers the mechanics, and brand positioning supplies the sentence the description starts from.
Confidentiality limits private proof, and public evidence can fill the gap
Professional services firms often cannot name clients, and that constraint is legitimate. Models need evidence, and a firm has several ways to supply it without breaching a duty of confidence.
Public mandates are the first source. Transactions, appointments and filings that are already public can be described by the firm with the client’s consent. Published work is the second: articles, research, white papers and commentary in recognised outlets. Speaking engagements and conference programmes are the third. Awards and rankings from established directories are the fourth. Each one is a public, third-party record that a model can retrieve and compare with the firm’s own description.
Edelman and LinkedIn’s study of 1,300 business decision-makers, surveyed in late 2016, found that 37% of decision-makers said thought leadership had directly contributed to a previously unconsidered company being invited to pitch against incumbents (Edelman, 1 June 2017). The study predates AI assistants. It shows that published expertise has long influenced who is invited to bid, and AI tools now read the same record.
Firms in regulated professions should check the advertising and publicity rules of their professional body before publishing any of this. Rules differ between professions and between jurisdictions.
Exhibit 1: The Describability Test for a professional services firm
The test asks five questions. A "no" to any of them shows where an AI engine is likely to omit, generalise or misdescribe the firm. The engine behaviour listed is an observed pattern in Vantage’s tracking and carries no guarantee.
| Question | If the answer is yes | If the answer is no, the engine tends to |
|---|---|---|
| 1. Can someone outside the firm describe it in one sentence? | The model has a clear sentence to repeat | Generalise to a category such as "a full-service firm" |
| 2. Do two independent sources describe the firm in the same way? | The model treats the description as corroborated | Omit the firm or pick the description of whichever source it finds first |
| 3. Are the partners named and described consistently? | The model links the people to the firm and its specialisms | Recommend the firm without its experts, or attach the wrong expertise to the wrong person |
| 4. Is there publishable proof, such as public mandates, awards, rankings or publications? | The model can cite evidence of competence | Describe the firm in general terms with no evidence |
| 5. Does the description match across languages and profiles? | The firm reads the same on its site, LinkedIn, directories and in each language it serves | Misdescribe the firm using the version it finds in the language of the query |
Exhibit 1. The Describability Test. Vantage framework.
Most firms pass the first question and fail the second or the fifth. A firm can score itself in an hour and repeat the exercise each quarter. The measurement guide explains how to track the result across engines.
Partners are the brand, so each partner needs to be findable as an entity
In a firm that sells judgement, clients buy the people. An AI answer to "who advises on regional restructuring" names firms, and the follow-up question names people. The model looks for each partner’s profile, publications and role, and it combines them.
Each partner needs a consistent public profile: the same name format, the same title, the same description of specialism and the same link to the firm across the website, LinkedIn, professional directories and conference listings. Gaps and variations dilute the signal. A partner who appears as three slightly different people leaves the model unsure whether the expertise belongs to the firm.
This work belongs to the firm and is quick to start. A single standard biography, agreed by the partner and applied everywhere, closes most of the gap.
Do the clarity work first and the content work second
Firms often respond to AI visibility by publishing more articles. The order of work that produces results begins with clarity.
First, agree the one-sentence description and the three specialisms the firm wants to be known for. Second, run the Describability Test and correct the lowest-scoring answer, usually the consistency of third-party listings. Third, standardise partner profiles. Fourth, publish evidence: commentary, research and mandates that support the specialisms. Content published against a clear description compounds. Content published without one adds volume with no direction.
Buyers may also search in more than one language, and the answer can change with the language of the question. Our analysis of multilingual GEO across Southeast Asia shows how to check that each language version of the firm’s description matches. The buyer-side view of the same decision is in our guide to choosing a branding agency in Singapore.
Related Vantage insights
- How to get your brand mentioned by ChatGPT: the mechanics of earning a mention.
- How to measure brand visibility in AI search: tracking the result across engines.
- How to choose a branding agency in Singapore: the buyer-side view of the same decision.
- Brand positioning framework: the source of the one-sentence description.
- Why AI does not cite your website: how models weigh outside sources.
About Vantage Branding
Vantage is a Singapore brand consultancy specialising in brand research, strategy, and identity design for ambitious organisations across Southeast Asia, with particular depth in healthcare, finance, government, and cultural-institution branding. Vantage builds fewer, stronger brands, pairing research rigour with senior craft across strategy, identity, experience and activation. PMC-certified, with EDGE grant support for overseas expansion. For a conversation about how we could help your brand, get in touch.