How professional services firms can improve visibility in AI search

Professional services firms cannot compel ChatGPT, Perplexity or Google’s AI features to cite or recommend them. They can, however, make their websites eligible, clearer and more useful as potential sources by publishing distinctive expertise, substantiating important claims, maintaining technical access and governing content carefully.

A referrer might ask:

Which Australian advisory firms have experience helping family-owned manufacturers prepare for succession, can work alongside an existing accountant and solicitor, and publish enough information for me to assess their approach before making an introduction?

An answer could emerge through ChatGPT Search, a Google AI Overview, Google AI Mode, Perplexity or an ordinary search result. It may draw on several sources: a firm’s website, practitioner profiles, public registers, independent publications or other pages. The firm cannot determine which sources will be selected, how it will be described or whether it will appear at all.

What it can improve is the source material available: whether important pages can be discovered, whether the firm’s services and experience are clear, whether professional claims can be checked, and whether the content remains accurate. Those improvements also make the website more useful to the buyer or referrer who visits it.

1. What AI-search visibility actually means

AI-assisted search is not one uniform channel.

ChatGPT Search can search the web for current information and present links to sources. OpenAI says placement is determined using multiple factors and is not guaranteed.

Google’s AI Overviews and AI Mode are features within Google Search. AI Overviews provide summaries with supporting links for some searches. AI Mode is designed for more exploratory or complex questions. Google states that both remain rooted in its established Search ranking and quality systems.

Perplexity searches for information, synthesises an answer and provides links or numbered citations to sources.

For a professional services firm, visibility can therefore mean several different things:

• a page is eligible to be discovered;
• information from the page contributes to an answer;
• the page is cited or linked;
• a user follows a referral to the website;
• the firm, service or practitioner is described more accurately; or
• the visit leads to a useful action, such as reviewing a profile, downloading a resource or starting an enquiry.

A citation is not an endorsement, professional verification or persistent ranking. It does not establish that the cited firm is suitable for a particular matter. Results can also vary with the precise question, platform, date, location, available context and product configuration.

The practical objective is not to secure a permanent position. It is to make the firm a clear, useful and technically eligible potential source when its real expertise is relevant.

2. Start with buyer and referrer research tasks, not prompt lists

A useful content plan begins with what someone needs to resolve before choosing, shortlisting or referring a firm. In professional services, those research tasks often include:

• whether the firm handles the particular problem, transaction or matter;
• whether it understands the relevant sector, client type or operating context;
• which service or practice area applies;
• who is likely to lead or perform the work;
• whether required qualifications, registrations, licences or authorisations are current;
• how the firm approaches the engagement;
• whether it can collaborate with existing advisers;
• what is included, excluded or referred elsewhere;
• what relevant experience or evidence is available; and
• what happens before an enquiry, introduction or formal engagement.

These are not merely keywords. A compound question can require information from several connected sources.

Full Steam's guide to how specific a professional services website should be explains how to organise precise service, situation and evidence pages without narrowing the whole firm unnecessarily. The referral guide also shows what referral partners assess before recommending a firm when they need to verify fit, people, boundaries and process.

The succession example in the opening might depend on a primary succession-advisory page, a manufacturing-sector page, profiles for the relevant practitioners, an explanation of the firm’s collaborative process and an article addressing a recurring decision in family-business succession.

The answer is not to create a separate page for every wording variation. Google’s current generative-AI optimisation guidance specifically advises against producing pages for every possible fan-out or long-tail variation.

Instead, build a coherent body of source material around durable research tasks. Each important page should have a distinct purpose, while internal links help readers and search systems move between the relevant service, context, practitioner and evidence.

3. Make the firm, service, sector and practitioner relationships clear

Many professional services websites contain the right facts but leave their relationships ambiguous.

A broad “what we do” page may carry ten loosely defined services. A sector specialisation may appear only in one biography. Two names may be used for the same practice area. An article may have a byline without explaining why the author or reviewer is relevant. Old profiles, capability PDFs and current service pages may contradict one another.

A clearer website answers basic relationship questions without requiring the reader to assemble the firm from fragments:

• What is the firm, and which disciplines or practice areas does it offer?
• What does each canonical service page cover?
• Which sectors, client types or recurring problem contexts are relevant?
• Which practitioners are responsible for particular work?
• Where can the firm operate, and which locations or jurisdictions matter?
• What methodology, experience or evidence supports the service?
• Which resources were authored or professionally reviewed by relevant people?

The page architecture should reflect those relationships.

A practitioner profile should link to the services and sectors for which the person has a genuine role. A sector page should lead to relevant services and evidence rather than repeat generic firm copy. An article should identify its author or reviewer where that context is useful, then link to the relevant canonical pages. Breadcrumbs and navigation should reinforce the same structure.

In technical language, this creates clearer entities, canonical pages and structured relationships. Those terms are secondary to the underlying requirement: a person should be able to understand who does what, for whom, under which circumstances and on what basis.

This is not an AI-only construction. It reduces ambiguity for buyers, referrers, staff maintaining the site and systems attempting to interpret it.

The same principle shapes Full Steam's broader approach to a professional services website: architecture and content should help people assess the firm even when no AI-generated answer is involved.

4. Publish evidence that generic capability statements cannot provide

Professional services websites frequently rely on claims that are positive but almost information-free.

Statements such as “we deliver tailored solutions”, “our experienced team helps clients navigate complexity”, “we are trusted advisers” and “we take a collaborative approach” could apply to thousands of firms.

More useful source material explains the substance behind the claim.

Distinctive first-party expertise can include the situations the firm is equipped to handle; the stages, inputs and outputs of an engagement; the criteria used to compare approaches; common trade-offs and limitations; what the firm needs from the client; and what normally happens next.

It may also include original research, anonymised datasets, diagnostic tools, practical frameworks or observations developed through real work.

The strongest material often explains boundaries. A firm can state when a problem falls outside its scope, when independent legal, tax, financial or technical advice is required, or when an apparent solution is unlikely to suit a particular context. This is more useful than implying universal capability.

Hypothetical worked example: succession advice for a manufacturer

Consider a hypothetical multidisciplinary advisory firm that assists family-owned manufacturers with succession planning. Its website should not invent an “AI answer page” repeating that phrase in several forms.

A more useful source set might include:

• a succession-advisory page defining the situations covered, engagement stages, expected inputs and boundaries;
• a manufacturing page explaining issues that arise in asset-intensive, family-owned operations without claiming experience the firm cannot substantiate;
• practitioner profiles showing who leads valuation, governance and implementation work;
• a methodology page explaining how commercial, ownership, family and transition questions are sequenced;
• a practical article comparing an immediate sale, staged transition and management succession, with qualifications rather than a universal recommendation; and
• a clear explanation of how the firm works with the client’s existing accountant, solicitor and licensed advisers.

Together, those pages help answer the compound research question because each contributes a different part of the evidence. They are also useful when no AI-generated answer is involved.

The objective is not to manufacture tiny “answer blocks” or paraphrase regulators, professional bodies and competing firms. It is to publish information the firm is qualified to provide and that adds something substantive.

First-party expertise is valuable precisely because it reflects the firm’s real decisions, methods, limits and experience.

5. Make identity and expertise verifiable without inventing ranking factors

Important professional claims should be accurate, specific and capable of being checked where appropriate.

That includes practitioner names and roles, qualifications, registrations, licences or authorisations, professional memberships, areas of responsibility, relevant sector experience, and article authorship or review.

Consistency reduces avoidable ambiguity. A practitioner should not have different job titles across a biography, article byline and downloadable capability statement. A former practice name should not remain on active pages unless the relationship is explained.

Where a public register is relevant, a genuine link can help a buyer or referrer perform their own verification. Genuine external publications or professional profiles may provide additional context.

These steps are about clarity and substantiation, not a documented “entity authority” formula.

Exact name consistency is not a published standalone citation factor. Schema.org’s sameAs property identifies a reference page that unambiguously indicates identity; it does not prove expertise. A public-register link does not automatically create authority. Author markup does not guarantee source selection.

Credentials and external mentions should never be added, exaggerated or manufactured for search purposes. Professional obligations differ by discipline, jurisdiction, licence, service and circumstances. The firm’s appropriately qualified reviewer should decide which claims may be made and how they should be framed.

6. Keep the technical route open

Useful content cannot contribute through a platform if the necessary pages are inaccessible to it. Technical eligibility is therefore a foundation, not a promise of inclusion.

ChatGPT Search

OpenAI identifies OAI-SearchBot as the crawler used to surface websites in ChatGPT search features. Its crawler documentation recommends allowing the bot in robots.txt and permitting requests from OpenAI’s published SearchBot IP ranges.

A public page can still be unavailable if a firewall, content delivery network or security rule blocks those requests.

OAI-SearchBot is distinct from GPTBot. The former concerns search inclusion; the latter is used for content that may help train OpenAI’s generative foundation models. OpenAI states that the controls are independent, so a firm can allow search crawling while disallowing training crawling.

ChatGPT-User, used for some user-initiated actions, is another separate agent and is not the control for Search eligibility.

OpenAI’s publisher guidance also documents utmsource=chatgpt.com in referral URLs from ChatGPT search results. Analytics rules should retain and report that parameter rather than stripping it during redirects.

Google

For Google’s generative-AI features, ordinary Search foundations still apply. Google says a page generally needs to be indexed and eligible to display a snippet.

Its current guidance emphasises crawlability, internal links, page experience, visible textual content and consistency between structured data and what users can see.

A practical review should cover:

• successful HTTP responses;
• robots and noindex controls;
• canonical tags;
• duplicate URLs;
• logical internal links;
• XML sitemaps;
• rendered content; and
• JavaScript dependencies.

Search Console should be used to inspect crawl and index status. A canonical tag or sitemap is not a substitute for resolving contradictory versions, but redirects, canonicals and sitemap inclusion can reinforce the preferred URL.

Google is rolling out a separate Search generative AI control to a subset of Search Console properties. Where available, inclusion is the default; exclusion prevents a site’s links and content from appearing in the covered generative-AI features. It is distinct from ordinary crawl and index controls.

Perplexity

Perplexity’s crawler guidance identifies PerplexityBot as the crawler intended to surface and link websites in Perplexity search results. It recommends allowing that agent in robots.txt and permitting its published IP ranges.

The documentation separately identifies Perplexity-User for actions initiated by a user. It is not an automatic web crawler or a foundation-model training bot, and its treatment of robots.txt differs because a user requested the retrieval.

A site using a web application firewall may need to distinguish legitimate Perplexity traffic from spoofed user-agent strings by checking the current published ranges.

None of this documentation publishes a citation-selection formula. A firm should keep intended public source pages accessible, but it should not expose confidential portals, client documents, restricted reports or identifying matter information merely to make them crawlable.

Technical eligibility does not guarantee discovery, inclusion, citation or accurate representation.

7. Use structured data as description, not a citation engine

Structured data provides machine-readable descriptions of visible content.

Legitimate implementations may use BlogPosting or Article for an article, Organization or an accurate subtype for the firm, Person for a practitioner, BreadcrumbList for visible breadcrumbs, and Service where the page genuinely describes a service. sameAs may connect an item to a reference page that unambiguously represents the same identity.

The markup should reflect the page, not compensate for missing or vague public information.

Google’s structured-data guidance requires markup to represent visible content, while its generative-AI guide states that structured data is not required for AI Overviews or AI Mode and that no special AI schema is needed.

FAQPage is not a citation mechanism. An FAQ section does not automatically justify that type, and Google stopped showing FAQ rich results from 7 May 2026. Valid markup is evidence only that the syntax and vocabulary have been implemented correctly; it is not evidence that an AI-assisted search system will select the page.

The llms.txt proposal describes a voluntary file intended to help agents use website information. It is not a universal web standard or a documented ranking factor. Google says it does not use llms.txt for Search and that the file neither helps nor harms Search visibility. OpenAI’s and Perplexity’s current crawler guidance does not list it as a search-eligibility requirement. A firm may maintain one for another supported purpose, but ease of implementation alone is not a reason to prioritise it.

8. Govern professional content so it remains accurate

A technically sound page becomes a poor source when its claims are stale.

Common risks include:

• practitioner departures;
• changed roles or credentials;
• licence and authorisation changes;
• narrowed service scope;
• renamed practices;
• office or jurisdiction changes;
• superseded guidance;
• outdated case evidence; and
• duplicate biographies.

Old PDFs are especially easy to overlook because they may remain indexed after the page that linked to them has changed.

Each important content group should have a named owner and an appropriately qualified professional reviewer. Reviews should occur on a schedule and when an event changes the underlying facts.

Canonical sources should be established for repeated information such as names, roles, offices and service descriptions. Superseded pages and documents should be updated, redirected, archived or removed according to their continuing purpose.

Publication and review dates are useful where recency matters, but a fresh date should not be applied without substantive review. Confidentiality, professional claims and client-identification risks need explicit checks before publication.

Generative AI can assist with research, comparison, structure and drafting. It must not invent a firm’s experience, credentials, clients, outcomes or professional opinions.

Google’s guidance on generative-AI content similarly focuses on accuracy, quality and value rather than the mere use of a tool. Qualified human review remains necessary before professional content is published.

9. Measure observable signals, not a fictional AI ranking

Measurement should separate what the firm can observe from what a platform has not disclosed.

For Google, monitor crawl and index status first. Google is rolling out a dedicated generative AI performance report to a subset of Search Console properties.

It currently reports impressions from AI Overviews and AI Mode, with page, country, date and device dimensions. Not every property has access, and the report does not disclose a citation formula or stable position.

For ChatGPT, retain the documented utm_source=chatgpt.com parameter and examine the landing pages, engaged sessions, enquiry starts and completed enquiries associated with those visits.

Review Perplexity and other identifiable referral domains on the same basis. Server logs can help diagnose crawler access and confirm whether security systems are blocking known agents, but a crawl is not proof of source use.

At the website level, useful measures include:

• engagement with service, sector and practitioner pages;
• enquiry starts, completions and assisted conversions;
• corrections to inaccurate or contradictory pages;
• the proportion of priority content that has a current owner and completed review; and
• resolution of crawl, index, canonical and rendering problems.

Limited manual checks can reveal material errors or omissions, but they should be recorded as dated observations.

Record:

• the exact question;
• the platform;
• the date;
• the location;
• the account state where relevant;
• the sources displayed; and
• any significant misdescription.

Do not call the result a ranking.

Treat unexplained “AI authority scores”, citation ranks, confidence scores, visibility percentages and share-of-answer claims cautiously. A third party may create a transparent and repeatable monitoring method, but it remains that provider’s methodology. It should not be represented as access to a private platform metric.

10. A prioritised professional-services AI-search readiness audit

A practice manager or marketing leader can work through the following sequence without beginning with schema or a large content-production program.

1. Define the important research tasks. List the material questions buyers and referrers must resolve for priority services and client contexts.
2. Confirm technical eligibility. Check crawl, index, snippet and relevant platform-inclusion controls, including firewall and CDN access.
3. Inventory canonical pages. Identify the authoritative firm, service, sector, practitioner, location, methodology and resource pages.
4. Verify professional claims. Check identities, roles, credentials, authorisations, scope, dates and external references.
5. Clarify page relationships. Improve navigation, internal links, breadcrumbs and page responsibilities so the website reflects how the firm works.
6. Replace generic claims. Add distinctive first-party information about situations, methods, decisions, boundaries and next steps.
7. Resolve contradictions. Update, consolidate, redirect, archive or remove stale pages, duplicate biographies and superseded documents.
8. Validate structured data. Confirm that each type and property has a legitimate purpose and matches visible content.
9. Assign governance. Name owners, qualified reviewers, review cycles and event-based triggers for each priority content group.
10. Configure measurement. Track crawl and index health, available platform reports, referral traffic, useful website actions and dated manual observations.

The order matters.

Schema cannot repair unclear services. More articles cannot compensate for contradictory practitioner profiles. Monitoring cannot create a meaningful baseline until technical access, source clarity and governance are sound.

A better potential source starts with a better website

AI-search visibility does not require a separate body of artificial “AI content”. It requires the firm’s real expertise to be expressed in material that people can understand, professionals can substantiate and search systems can access.

The strongest first priorities are clear relationships among the firm, its services, sectors and practitioners; distinctive first-party information; clean technical eligibility; fewer stale or contradictory claims; and disciplined ownership and review.

None creates a citation guarantee. Together, they produce a more useful professional services website—and a better potential source whenever an AI-assisted search system is looking for information the firm is genuinely qualified to provide.

Frequently asked questions

Can a professional services firm guarantee that ChatGPT, Perplexity or Google will cite or recommend it?

No. OpenAI states that placement in ChatGPT Search is not guaranteed, and Google states that meeting technical and content requirements does not guarantee indexing or inclusion in its generative-AI features.

Perplexity publishes crawler-access guidance, not a citation-selection formula. A firm can improve public accessibility, source clarity, distinctiveness and substantiation, but it cannot control a platform’s answer, sources or wording. A citation should not be presented as a recommendation or professional endorsement.

Does schema markup improve a professional services firm’s AI-search visibility?

Structured data can describe visible content more explicitly and can support legitimate Search features. It may therefore be a sensible part of a technically sound website.

It is not, however, a citation engine. Google says structured data is not required for its generative-AI Search features and that there is no special AI schema. Use accurate types such as BlogPosting, Person, Organization, Service or BreadcrumbList only where they represent the visible page and fit the site’s wider implementation.

Do practitioner profiles and professional credentials matter for AI search?

They matter because buyers, referrers and reviewers need to understand who is responsible for the work and whether relevant claims can be substantiated.

Accurate profiles can reduce ambiguity across services, sectors, articles and external references. That does not mean a qualification, register link, Person markup or consistent name independently causes citation. Publish only current, relevant and supportable credentials, explain the practitioner’s real responsibilities, and have discipline-specific claims reviewed by an appropriately qualified person.

Does a professional services website need an llms.txt file?

Not for Google Search. Google says it does not use llms.txt and that the file has no positive or negative effect on Search visibility or rankings.

The current llms.txt specification describes itself as a proposal, while OpenAI’s and Perplexity’s crawler guidance does not list the file as a requirement for search inclusion. A firm may use one where a particular tool or operational purpose supports it, but it should not displace work on public content, crawler access, canonical pages or governance.

How can a firm measure visibility in AI-assisted search?

Begin with observable data: crawl and index health, Google’s generative-AI report where available, ChatGPT referrals carrying utm_source=chatgpt.com, other referral domains, landing-page engagement, enquiry activity and assisted conversions.

Add server-log checks and a content-governance register. Manual searches can supplement this evidence when the exact question, platform, date, location, account state and displayed sources are recorded. Treat them as dated observations, not stable rankings, and do not confuse third-party scores with platform metrics.