Guide
How to measure AI visibility
Guessing from one ChatGPT session is not a measurement system. Use a repeatable scan, fixed pillars, and evidence you can re-run after changes.
A repeatable measurement loop
- Baseline — run a full scan before a campaign of site changes
- Diagnose — separate crawl blockers from content gaps from answer absence
- Fix — ship the highest-leverage items from the action plan
- Re-scan — compare pillar scores and evidence, not vibes
What BatSignal measures
| Pillar | Weight | Evidence |
|---|---|---|
| Crawl access | 25% | robots + live AI bot fetches |
| Content readiness | 20% | HTML, meta, OG, JSON-LD, sitemap, llms.txt |
| ChatGPT search | 30% | mentions / recommendations / citations |
| Open-web coverage | 20% | Open-web answer evidence |
| Training presence | 5% | Common Crawl signal |
Avoid these measurement mistakes
- Only testing branded prompts that already include your name
- Changing five things at once so you cannot attribute lifts
- Treating a single answer as a durable ranking
- Ignoring crawl failures while rewriting blog copy
Deep dive: methodology. Product surface: features. Start a project from signup.
FAQ
How often should I re-scan?
Re-scan after robots, template, or major content changes, and at least monthly if AI visibility is a tracked KPI. Avoid reading noise into day-to-day answer variance.
What is a good score?
Scores are relative to BatSignal’s four-pillar model, not a universal industry grade. Use your baseline and deltas after fixes; read the methodology for weights.