Original data
AI citation statistics: provenance remeasurement in progress.
In July 2026 we retired every historical citation-rate and source-graph figure that could not be verified under our new per-source provenance contract. The answer-coverage, mention, and crawler figures below remain available with their denominators. Status updated 2026-07-21.
3,206
historical AI answers measured across 10 engines during a 90-day dogfood window ending 2026-07-18.
42.4%
of 3,206 historical answers mentioned the tracked dogfood brand. This is text matching, not a citation claim or an industry benchmark.
10
distinct AI-associated crawler user-agent strings observed in edenrank.com server logs over 9 days (GPTBot, PerplexityBot, ClaudeBot and others).
~226/day
requests bearing AI-associated user agents on edenrank.com (2,038 hits in 9 days of logs).
Historical mention coverage by engine
Share of answers whose text mentioned the tracked dogfood brand, per engine. Each row includes its measured-answer denominator. These figures do not depend on citation-source provenance.
Measurement honesty: this is one brand's historical prompt set, not a representative market sample. It describes what those 3,206 answers said; it does not establish lift, traffic, or commercial impact.
Citation metrics: retired and being remeasured
Historical citation rates and the source graph are not displayed. Older runs did not preserve enough per-item provenance to distinguish an AI-answer citation from an organic search fallback. Because that distinction cannot be reconstructed honestly, those figures remain retired rather than being estimated or backfilled.
Remeasurement uses contract provider-effective-path-v1 for answer surfaces and niche-ai-answer-probe-v1 for the niche graph. New figures will return only after they pass the current provenance and publication gates.
AI crawlers on a real site
Requests by AI-associated crawlers in 9 days of edenrank.com server logs (63,306 total requests in the window). ChatGPT-User and Claude-User are user-agent labels associated with live assistant fetches. User-agent matching alone does not cryptographically verify the operator, so these rows describe observed request labels, not authenticated crawler identities.
| Crawler | Operator | Hits / 9d |
|---|---|---|
| GPTBot | OpenAI (training/index crawl) | 652 |
| PerplexityBot | Perplexity (index crawl) | 409 |
| OAI-SearchBot | OpenAI (search index) | 214 |
| Amazonbot | Amazon (Alexa/AI) | 190 |
| Meta-ExternalAgent | Meta (AI crawl) | 167 |
| Applebot | Apple (Siri/Spotlight/AI) | 145 |
| ClaudeBot | Anthropic (index crawl) | 123 |
| ChatGPT-User | OpenAI (live fetch during a user chat) | 118 |
| Bytespider | ByteDance (AI crawl) | 11 |
| Claude-User | Anthropic (live fetch during a user chat) | 9 |
Methodology
- Answers: 3,206 AI answers collected over a historical 90-day window ending 2026-07-18 across 10engines for one tracked brand's prompt set (EdenRank's own dogfood workspace). Mention and answer-count figures are specific to that prompt set - they are a historical measured example, not an industry average.
- Mentioned: an answer counts as a mention when the tracked brand name appears in the answer text. Citation rates are intentionally absent until enough post-contract observations exist.
- Source graph: brand-agnostic probe queries now retain only sources explicitly cited by the AI answer and stamp the extraction method. Legacy and unknown methods are excluded. The public graph remains withheld while qualified data rebuilds.
- Crawlers: raw nginx access logs of edenrank.com, user-agent matched, 9-day window ending 2026-07-18. No sampling; operator labels are classifications, not verified identities.
- Refresh: figures are regenerated from production data and the date above changes when they do.
Check your own numbers
Run the free AI visibility checker to see whether AI engines can even read your site, browse our public citation evidence, or explore the free tools.