We put our own GEO measurement under review.

The starting point

We used molthub's own site to document how a small GEO studio builds a measurable foundation. The public experiment began in August 2026; the dated five-query web-search observation was recorded on 2 September.

What the evidence showed

That search observation returned 0/5 molthub mentions and 0/5 citations. It was a web-search observation, not a consumer ChatGPT or Gemini test.

On 3 October, machine requests from the audit environment returned HTTP 403 for the homepage and several public paths. This demonstrated a failure in that environment at that time, not a block affecting every visitor or verified search crawler.

What we implemented

The October update repaired how the scanner and event tracking handle incomplete evidence.

• Unreadable, redirected or challenged robots requests remain unverified and earn no passing score.
• Robots checks use the applicable crawler group and path rules; policy permission stays separate from actual access.
• Referral and campaign fields distinguish observed AI referrals, tagged campaigns and unknown traffic.
• Failed event writes return a failure instead of reporting an unsaved event as successful.

What was verified

The measurement fixes were checked and deployed in the October update. The public research page retains the baseline, access record, measurement template and limitations.

Those checks verify the scanner's behavior. They do not establish Google indexation, improved AI citations, production conversion attribution or payments.

The next measurement

Confirm current access and indexation with search URL inspection, run fresh cross-platform buyer questions and verify that production referral and conversion events are saved. Keep the historical query set separate from new questions so the comparison remains interpretable.