Short answer: if a professional services firm doesn't appear in Claude citations, it doesn't automatically start rewriting articles. Anthropic documents web search and citations, but does not publish a formula by which a particular page format achieves selection. The correct diagnosis first checks the identity of the company, the ownership claims, the first-party sources, the relevant external profiles and the indexability of the pages.

Failure mode 1: firm and experts are mixed

An article can be signed by a consultant, while the service page belongs to the organization. If profiles, bylines, and the company page do not clarify the relationship, public sources may describe different entities.

Failure mode 2: The same offer has three descriptions

Marketing, a partner's profile and an external directory may describe the service differently. Before you optimize the citation, define which first-party page owns the current scope.

Failure mode 3: experience claims without owner

"20 years of experience", "over 100 projects" or "presence in 12 markets" must make exact sense and have a verifiable internal source. If they cannot be supported, remove false precision.

Failure mode 4: service pages are too commercial

A page may list benefits but not explain the problem, methodology, or limitations. The lack of a citation does not prove that this is the reason, but the page may be weak to the user before any GEO issues.

Failure mode 5: thought leadership competes with service page

A large article can accidentally become the owner for the question that the service page should clarify. It defines roles: the article explains, the service describes the offer, the methodology can have a separate owner.

Failure mode 6: stale external profiles

Professional directories, association profiles or speaker pages may retain old roles and services. Only prioritize sources relevant to the buyer journey and those you can legitimately manage.

Failure mode 7: case studies without context

A case may be cited externally, but it does not say what service was delivered, under what conditions, or what limitations the result has. Don't make up details to make the study more "extractable".

Failure mode 8: divergent bios

The same expert may appear with different roles. It separates the legitimate career change from a stale page.

Failure mode 9: technical SEO is faulty

Noindex, wrong canonical, redirects, broken pages or inaccessible links are more direct problems than "AI visibility". Fix the base before copy.

Failure mode 10: The same definition is cloned

Professional services use methodology terms. If the definition is copied to many pages and then updated in only one place, the conflict occurs.

Failure mode 11: a single query becomes a verdict

Answers may vary. Use sets of questions on awareness, evaluation, comparison and implementation. Keep the same language and period.

Failure mode 12: the absence of owned citation is treated as an error

If the answer is correct and uses a current independent source, the absence of the own domain is not automatically a problem.

Reproducible decision tree

  1. Are the firm, the expert and the service clearly separate entities?
  2. Is there a first-party owner URL for the verified claim?
  3. Is the page indexable and canonical?
  4. Do commercial claims have provenance?
  5. Do thought leadership and service pages have distinct roles?
  6. Are the priority external profiles current?
  7. Is the author identity consistent?
  8. Do case studies explain scope and limits?
  9. Is the query set versioned?
  10. Is the observed problem factual or just lack of proper citation?
  11. Is the answer repeated in several observations?
  12. Is there a verifiable action for finding?

If the first four answers are negative, the rewrite for Claude is premature.

How do you build the baseline

For 20-50 priority questions save the material answer, sources, factual accuracy, brand mention and timestamp. In parallel, it maintains the first-party claims registry and critical external profiles.

Do not collect personal customer data or confidential projects for this analysis.

How do you prioritize findings

P0: mistaken identity or potentially misleading professional claim. P1: stable service or role. P2: Inconsistent relevant external source. P3: cosmetic variations and isolated observations.

How do you measure after remediation

Track factual accuracy, first-party conflict rate, critical-profile consistency, source mix and time-to-resolution. Referral or leads can be added separately with their own denominator.

Do not aggregate into a single "Claude authority score".

An example

A consulting firm has a partner listed as the owner of a methodology in an article, but his profile says a different role and the service page uses an older version of the methodology. Claude quotes an external interview. The problem is not automatically the interview. First align the ownership and first-party claims, then rerun the observation.

Stop criterion

The diagnosis enters monitoring when entities and owners are clear, P0/P1 conflicts are closed, priority external profiles have status and the query set no longer produces new material errors.

How do you check if the finding is worth escalating

Not all deviations are equal. If Claude cites a correct external source and the response is factual, keep the observation, but do not automatically open an incident. Escalate when mistaken identity, stale role, misassigned service or a material claim that can influence the customer's choice occurs.

Keep for each finding query, claim, source, severity, owner, observed_at and verified_at. This structure allows comparison between rounds and prevents an isolated catch from becoming the general conclusion.

Claim ledger

  • FACT/EVIDENCE: Anthropic documents web search and citations.
  • FACT/EVIDENCE: Google documents Organization, ProfilePage and Article structured data.
  • PRACTITIONER GUIDANCE: the diagnosis must separate firm, person, service and external source.
  • NOT PROVEN: a public formula by which a firm guarantees citation in Claude.

Conclusion

In professional services, the real bottleneck is often the ownership of the information, not the format of the article. If your company, experts, and services are described consistently, you can measure citations without turning every variation into a rewrite project.

Sources reviewed

Related insights