AIBL Research
Evidence worth examining.
Investigations into what AI visibility measurements can—and cannot—tell us.
Research library
02 published notes- Brand identity
- Recognition across saved answers
- Measurement
- Completion labels and partial evidence
- Our approach
- Findings, methods and limits together
The research library
Two investigations. Evidence you can inspect.
Read the finding, examine the records, and see where the conclusion stops.
AI identified our business.
Two weeks later, it couldn’t.
Thirteen saved answers, one recurring question, and a correct identification that did not persist. Inspect the evidence and why the effect of our site changes remains unknown.
13 saved answersTimeline · Original answers · Limitations
Read the case study →
When a completed AI scan
isn’t a complete answer.
Of 1,593 non-mock question records marked completed, 253 carried an explicit truncation flag. A retrospective audit of our collection history, with methods, limitations and downloadable metadata.
1,625 saved question recordsMethods · Metadata · Limitations
Read the audit →
Evidence, not a sales score
These notes describe the records we inspected. Their limits are part of the finding.