Measuring AI visibility is harder than measuring traditional SEO because AI search is probabilistic. The same prompt at the same time on two devices can return different citations. Measurement has to account for variance, not just point-in-time results.
The full protocol covers baseline capture, prompt sets, variance accounting across runs, position scoring, and a 90-day measurement cadence. That writeup is in progress. In the meantime, here is the 5-test mini-protocol that catches most real-world AI visibility issues.
Test 1: The Direct-Ask Test
Open ChatGPT, Claude, and Perplexity in separate tabs. Ask each: “What does the business at [yourdomain.com] do?” Score 1 point for each system that returns an accurate answer. Zero out of three means you are invisible at the retrieval layer.
Test 2: The Category-Recommendation Test
Same three systems. Ask each: “Recommend a [your industry] in [your city or region in the Philippines].” Run the prompt three times per system. Document which businesses are cited and whether you appear. Citation rate below 33% means you are not in the active retrieval set.
Test 3: The Google AI Overview Test
Search the top 5 commercial-intent queries your customers actually type. Note whether Google generates an AI Overview, and if so, which sources it cites. If a Google AI Overview appears and you are not in it, that is the primary citation gap you need to close.
Test 4: The Crawler-Access Test
Open yourdomain.com/robots.txt directly in a browser. Look for any Disallow rule targeting GPTBot, ClaudeBot, PerplexityBot, Google-Extended, anthropic-ai, or ChatGPT-User. Any block here means AI cannot read your site at all, which is the most common hidden cause of invisibility in 2026. The fix is documented in Cloudflare Is Blocking AI From Reading Your Website.
Test 5: The Corroboration-Search Test
Google your exact business name in quotes. Count the top 20 results that are not your own properties (website, social profiles you control). Fewer than five means AI has no third-party signal to corroborate that you are a real entity. This is the entity-clarity gap from Reasons 4 and 5 in the diagnostic.
Scoring
Tally a score out of 5. Tests 1, 2, and 3 capture live AI behavior. Test 4 captures access. Test 5 captures entity authority. A score of 5/5 means your AI visibility is in good shape. A score below 3/5 means you have at least one structural gap, usually crawler access, entity confusion, or content extractability.
Re-run the full protocol every 30 days. Document scores over time. The compounding view is more useful than any single run, because AI search variance averages out across enough samples.
Where to Go From the Scores
If Test 4 fails, fix crawler access first. Everything else is wasted effort if AI cannot read your site. See the Cloudflare unblocking guide for the three-layer fix.
If Tests 1, 2, or 3 fail, the issue is on-site: missing schema, weak entity signals, content that does not answer directly. Run the 8-reason diagnostic to identify which gap is yours.
If Test 5 fails, the issue is off-site: you need corroborating mentions from third-party sources. This is the slowest gap to close because it depends on other sites referencing you, but it is also the most defensible once built.
This is the mini-protocol. The complete measurement methodology, with prompt sets, variance accounting, scoring rubrics, and 90-day cadence will replace this page when published. Bookmark the URL if you want the upgrade in place.