The Search Console generative AI report, read honestly
What Google's generative AI performance report counts, what it leaves out, how its impressions aggregate, and how to use it without overreading.
measurement
8 posts
Measuring AI visibility without overstating it
A measurement stack for AI search: readiness audits, engine-reported impressions and citations, referrals, and a citation log reported as ranges.
Building a defensible 0–100 score
How to design a readiness score that survives scrutiny: published signals and weights, deterministic inputs, truthful not-scored states, and honest bands.
Crawl, fix, re-crawl: proving a change moved something
A baseline crawl, a dated change log, a like-for-like re-crawl and a diff: how to prove a fix landed, and how to keep readiness deltas apart from traffic.
What prompt tracking can and can't tell you
Prompt tracking samples a probabilistic system with prompts nobody typed. What the method can support, what it can't, and how to read a report honestly.
Reading server logs for AI bot traffic
Find AI crawlers in access logs, verify them against published IP ranges, read CDN and origin logs differently, and build the analysis in a spreadsheet.
Prioritizing AEO fixes: severity × page value
A reproducible way to order AEO and GEO fixes: rate severity, page value, and effort in a spreadsheet, rank the list, then re-crawl to prove each fix landed.
What GEOBee measures, and what it deliberately doesn't
What GEOBee's three lenses score, why the results are local and re-runnable, and what it does not measure: citations, prompt visibility, or rankings.