Let the crawler in
Claude's web search relies on its own user agents. If robots.txt blocks them, you cannot be cited regardless of content quality. Review your directives and confirm nothing is silently excluding the search crawler along with training crawlers.
Write for a careful reader
Claude's answers tend to hedge and qualify. Content that already contains the caveats - conditions, exceptions, ranges rather than absolutes - is easier to incorporate faithfully than content full of unqualified superlatives.
Make claims checkable
Attach dates, methodologies, and sources to your numbers. A statistic with a stated method survives scrutiny in synthesis; an unattributed round number gets dropped or attributed to a more transparent source.
Structure for extraction
Question-shaped headings, standalone answers, tables for comparisons, and a summary block near the top. The same structure that wins featured snippets wins Claude citations, because both depend on a passage that survives being lifted out.
Strengthen who you are
Organization schema, an author page with real credentials, and consistent descriptions across the web make it easier for the model to name you rather than describe your category generically.
Measure it deliberately
Run a fixed prompt set in Claude with search enabled, three runs per prompt, and log mentions and cited URLs. Track the trend monthly - single-run results are non-deterministic and will mislead you.