Measurement
Nearly everything in AI SEO that deserves the word "known" came from measurement. This pillar covers what a small business, or anyone, can observe: crawler visits, referral clicks, prompt-set results, citation patterns. It is also where this site publishes its own instruments and findings, because we hold ourselves to the standard we describe.
- Measure the pipeline as separate stages. Retrieval, citation, answer influence and real custom are different outcomes; collapsing them into one number destroys the information. Strongly supported
- Your server logs already contain evidence: which AI crawlers fetch your pages, verified against the operators' published identities. GA-style analytics cannot see this. Strongly supported
- You can count ChatGPT referrals. OpenAI tags applicable referral links with utm_source=chatgpt.com. Official recommendation
- Repetition is mandatory. A single-run prompt check measures noise. Our protocol runs every question repeatedly, in fresh sessions, and reports the variation. Strongly supported
In this pillar
- Prompt-set testinglive
- Reading your crawler logslive
- Tracking ChatGPT referralslive
- Why citation counts aren't enoughlive
- Search Console for AI features · Bing AI Performance · Branded-search measurement · Misleading visibility scoresplanned
What bernard does about this pillar
Every hosted site gets verified AI-crawler observation as standard: which operators fetched which pages, identity-checked against published IP ranges. verified
Per-customer citation tracking, meaning your business's own visibility across engines over time, is designed and not yet shipped. we're building this