Pre-flight checks
- robots.txt: ClaudeBot and Claude-User decisions documented (default: allow both).
- Authorship audit: does substantive content carry named or institutional authorship and dates?
- Baseline: ask Claude your category's analytical questions; note where it hedges.
Source priorities
- Primary documents — original data, methodology pages, real documentation. Claude's citation style prefers sources that originate claims.
- Precision passes — replace superlatives with numbers; hedged, verifiable claims match the house style.
- Institutional clarity — who published this, when, with what qualifications to know.
30-day plan
| Week | Moves | Checkpoint |
|---|---|---|
| 1 | Crawler + authorship audit fixes | All key pages carry authorship and dates |
| 2 | Publish one primary-source asset (data, method, documentation) | Asset live, structured, dated |
| 3 | Precision pass on top pages (numbers over adjectives) | Zero unsupported superlatives on money pages |
| 4 | Re-test analytical prompt set | Hedging reduced; citations documented |
Why bother, given Claude's smaller share
Claude's users write tomorrow's sources: analysts, developers, journalists. Being correctly understood here seeds the secondary literature every engine reads. It is the cheapest long-game in GEO because the work — primary sources, precision — improves your standing everywhere else simultaneously.
Frequently asked questions
What counts as 'primary material' for a normal business?
Your own measured data (any structured test with published method), your actual product documentation, your real process written down with specifics. The bar is origination, not scale — a 30-sample test with honest limitations is primary material.