When a B2B SaaS product doesn't appear in Claude citations, the temptation is to immediately rewrite landing pages. It's a premature reaction. The bottleneck may be in access, in documentation, in product identity, in the lack of a suitable source for the test question, or simply in the variation of answers. Diagnosis must eliminate these causes one at a time.

Before any big editorial project, save a set of prompts and a baseline. Without them, you won't know if the rewrite changed anything or if you're just seeing normal variation.

Clarify what "we are not quoted" means

It may mean that the domain does not appear at all, that a third-party source appears, that the documentation is cited but not the product page, or that the answer mentions the product without its own source. These situations call for different actions.

Define the finding in observable terms. "Poor visibility" is too vague to diagnose.

First check access to pages that should be sources

Inspect HTTP status, robots, noindex, authentication, and any protection that may prevent access. Technical documentation, integration pages, and security resources are often more valuable to B2B inquiries than the homepage.

If the page is only available after login, it does not mean that the system can use it as a web source. Keep a public version of the information you want a buyer to be able to verify.

Compare the question with the page you want it to cite

A prompt about integrating with a CRM needs a source that explains the integration. A generic landing page can be very well written and still not be the most relevant piece of evidence.

Construct an array prompt -> desired source. If there is no page that responds directly, the problem is coverage, not wording.

Separate the entity issue from the content issue

If the product has a similar name to another company, if the brand and product name are used inconsistently, or if purchases have changed identity, a response may confuse the entities. Check Organization data, company pages, and product-organization relationship.

Don't add sameAs or markup just to force association. Use references that represent the real identity and are consistent with visible pages.

Control third party sources

In B2B, marketplaces, partner pages, review sites and repos can describe the product. Sometimes they are more specific than the website itself. Identify which source appears and if it is current.

If a third-party page contains outdated information, don't assume that rewriting the homepage will solve the problem. It may be necessary to correct the listing or clarify the official documentation.

Check factuality before tone

A citation is unsuccessful if the response misrepresents the product. For each example, compare the statement with the current documentation. A material factual issue takes precedence over the absence of branding in a generic response.

Maintain list of sensitive claims: pricing, security, compliance, plan limits, availability and integration.

Don't use a single prompt as a verdict

Answers may vary. Run a steady set and keep the variation. For each question family, look at the frequency of source types, not just an isolated occurrence.

If you change the wording of the prompt, make a note of it. A paraphrase may change the intent or the sources used.

Diagnostic tree

  1. Is there a public page suitable for the test question?
  2. Is it accessible and indexable under normal web conditions?
  3. Is the product and company identity consistent across the site?
  4. Are there third party sources that are more specific or more current?
  5. Is the answer factually correct even if it does not cite its own domain?
  6. Is the finding repeated in several runs?
  7. If the problem appears to be editorial, what specific passage or page is missing?
  8. After the fix, repeat the set without changing the definitions of success.

When content is worth rewriting

Rewriting is warranted when the relevant page exists but does not clearly answer the question, when the product terms are ambiguous, when the documentation is fragmented, or when important information is hidden in hard-to-use formats.

It is not justified just because the domain does not appear in a single answer. First, it removes technical and coverage bottlenecks.

Prioritize by buying journey

A missing citation to a generic question may be of little value. An error on a question about security, compatibility or pricing can affect a buying committee. It links the finding to the actual selling stage.

This prioritization prevents optimizing for vanity metrics and helps content, product marketing, and sales engineering work on common problems.

A quick test for coverage before rewriting

Take ten questions from sales calls or demos and check if there is a public page that can support the answer. Do not rate the style yet. Just note the existence of the source, specificity and freshness. If half of the questions do not have a suitable source, the problem is coverage. If the sources exist and are clear, but the finding persists, you can investigate entity resolution, provenance, or platform variation. This short test prevents large editorial projects started from a poorly defined symptom.

Claim ledger

  • FACT/EVIDENCE: Anthropic documents web search and citations in its tools, without guaranteeing citation of a specific domain.
  • FACT/EVIDENCE: Google documents Organization structured data and web access, but these are not declared citation control mechanisms Claude.
  • PRACTITIONER GUIDANCE: prompt-source matrix and diagnostic tree are proposed auditing methods for B2B SaaS.
  • INFERENCE: specific documentation and consistent identity may reduce some ambiguities, but the effect on citations should be noted.
  • NOT PROVEN: that rewriting landing pages directly produces more citations or pipeline.

Sources reviewed