The Lab
We run this site as a public GEO experiment. The Lab shows, in real and dated numbers, whether the LLMs cite us, which AI crawlers fetch us, how we score against our own GEO checklist, and which experiments we are running on ourselves. Where a measurement has not landed yet, we say so — never a fabricated zero.
Citation rate
0%
0/50 prompts · 1 Oct 2026
AI crawlers
10121
claimed requests · 24 crawler labels
Experiments
4
0 running · 1 planned · 3 closed
Automated self-audit
100%
9/9 pages passed the automated subset on 2026-06-08; current corpus: 10 live articles. Historical scope differs from the current corpus; this is not current corpus-wide compliance.
Citation panel: latest run and historical coverage
A diagnostic panel, separate from website visits or Bing AI reporting. New paid runs are paused pending provider and budget approval. Historical runs remain available; an API response is not proof that live search ran. dataset
0%
0/50 prompts · prompt set v1 (locked 2026-06-07) · as of 1 Oct 2026
| Engine | Rate | Cited | Mentioned | Valid / errors |
|---|---|---|---|---|
| ChatGPT | unavailable | 0 | 0 | 0 / 50 |
| Claude | 0% | 0 | 0 | 50 / 0 |
| Gemini | 0% | 0 | 0 | 50 / 0 |
| Perplexity | 0% | 0 | 0 | 50 / 0 |
Latest run: partial; providers with valid answers: 3/4. Valid responses: 150; errors: 50. The headline denominator counts prompts with at least one valid answer, not all provider responses.
Historical citations are present: 2026-09-01: 1/50 prompts. A later zero does not erase those observations.
Model identifiers, grounding coverage and costs are unavailable in this aggregate dataset. Errors are not zero citations. Runs with different coverage are not directly comparable; legacy rows lack response/error counts. Latest capture: 2026-10-01T02:00:06.822Z. No new run was made for the 21 October 2026 review.
Which AI crawlers fetch us
24 crawler labels were claimed; 10121 claimed requests in the retained dataset · updated 27 Oct 2026 · dataset
Generated 2026-10-27T02:15:02.458Z; 0 complete day(s) old.
Counters reconcile: claimed = verified + failed verification + unknown in the retained dataset.
How we check that a crawler is who it says it is
A User-Agent is a string the client chooses. We verify OpenAI and Perplexity against versioned operator IP evidence where available, and Google, Bing, Apple and Amazon by forward-confirmed reverse DNS. Anything we cannot check either way is reported as unverified — never as genuine, never as forged.
Verification runs from 18 Sept 2026 onward; earlier rows cannot be checked, because client addresses were never stored.
| Crawler | Operator | Purpose | 7d | 30d | Claimed | Verified | Unknown | First seen |
|---|---|---|---|---|---|---|---|---|
| ClaudeBot | Anthropic | training | 55 | 525 | 1906 | 0 · 14 failed verification | 1892 | 31 May 2026 |
| Googlebot | search | 218 | 546 | 1658 | 460 · 47 failed verification | 1151 | 31 May 2026 | |
| OAI-SearchBot | OpenAI | search | 246 | 1119 | 1364 | 956 · 19 failed verification | 389 | 6 Jun 2026 |
| GoogleOther | mixed | 285 | 321 | 993 | 300 | 693 | 5 Jun 2026 | |
| Amazonbot | Amazon | mixed | 39 | 187 | 834 | 194 · 24 failed verification | 616 | 6 Jun 2026 |
| bingbot | Microsoft | search | 54 | 202 | 793 | 235 · 13 failed verification | 545 | 1 Jun 2026 |
| ChatGPT-User | OpenAI | on-demand | 92 | 473 | 721 | 366 · 48 failed verification | 307 | 6 Jun 2026 |
| GPTBot | OpenAI | training | 30 | 75 | 340 | 168 · 24 failed verification | 148 | 6 Jun 2026 |
| Applebot | Apple | search | 70 | 84 | 330 | 58 · 23 failed verification | 249 | 4 Jun 2026 |
| Meta-ExternalAgent | Meta | mixed | 19 | 36 | 217 | n/a | 217 | 6 Jun 2026 |
| PerplexityBot | Perplexity | search | 21 | 36 | 196 | 9 · 37 failed verification | 150 | 6 Jun 2026 |
| facebookexternalhit | Meta | on-demand | 18 | 49 | 164 | n/a | 164 | 10 Jun 2026 |
| CCBot | Common Crawl | training | 17 | 33 | 111 | n/a | 111 | 6 Jun 2026 |
| Claude-User | Anthropic | on-demand | 5 | 8 | 104 | 0 · 4 failed verification | 100 | 31 May 2026 |
| Google-Extended | training | 0 | 0 | 88 | 0 · 13 failed verification | 75 | 6 Jun 2026 | |
| Bytespider | ByteDance | training | 19 | 24 | 79 | n/a | 79 | 6 Jun 2026 |
| cohere-ai | Cohere | training | 20 | 36 | 69 | n/a | 69 | 6 Jun 2026 |
| Perplexity-User | Perplexity | on-demand | 14 | 25 | 68 | 0 · 13 failed verification | 55 | 6 Jun 2026 |
| YouBot | You.com | search | 19 | 27 | 45 | n/a | 45 | 11 Sept 2026 |
| MistralAI-User | Mistral | on-demand | 0 | 0 | 23 | n/a | 23 | 17 Jul 2026 |
| anthropic-ai | Anthropic | on-demand | 0 | 0 | 14 | n/a | 14 | 6 Jun 2026 |
| Diffbot | Diffbot | mixed | 0 | 2 | 2 | n/a | 2 | 8 Oct 2026 |
| Applebot-Extended | Apple | training | 0 | 0 | 1 | 0 | 1 | 17 Jul 2026 |
| Claude-SearchBot | Anthropic | search | 1 | 1 | 1 | 0 | 1 | 21 Oct 2026 |
Traffic
We measure traffic with self-hosted, cookieless Plausible (no cookie banner, no third-party tracker). This page does not currently publish audience totals. Analytics visits are separate from crawler requests, citations and Bing AI visibility; a JavaScript event alone does not prove a human reader. QA must be labelled and excluded in a reproducible reporting layer.
What we've published
10 live articles · 6 glossary terms. Participation in a controlled experiment is listed separately in the registry.
For the narrative behind these numbers — the stack, the first measurements, and the honest caveats — read This site is our GEO lab.
Experiments
Full log →- exp-004planned
Does inclusion in llms-full.txt change crawl or citation?
Proposed — not started. Awaiting selection and control/treatment sign-off.
- exp-003inconclusive
Does an answer-first opening change how often we're cited?
Closed on 2026-10-21: inconclusive. The original review window ended 2026-07-08. A single unmatched-in-detail page pair, short baseline, historical identity…
- exp-002inconclusive
Does FAQPage schema raise AI crawl frequency or citation?
Closed on 2026-10-21: inconclusive. The original review window ended 2026-07-08. A single unmatched-in-detail page pair, short baseline, historical identity…
- exp-001concluded
Baseline established — the day-zero snapshot
At launch the live corpus was 2 articles (both cornerstone) plus 6 glossary terms; robots.txt welcomes the major AI crawlers and Cloudflare runs DNS-only, so…
Methodology & data
Every figure here is dated and traceable to a published dataset. The raw numbers are at /datasets/citation-rate.json and /datasets/ai-crawlers.json; a machine-readable summary is at /lab.md. How we test: methodology.