Transparency · updated every 15 minutes
AI crawl statistics
How often the crawlers behind ChatGPT, Claude, Perplexity, Gemini and the other assistants read this site. Counted at our origin, so every figure is a floor.
- AI/LLM requests · last 24h
- 87,997
- AI/LLM requests · last 30 days
- 919,972
- Assistant-company crawlers only · 30d
- 580,495
- Tools re-checked · last 24h
- 1,182
By company, last 30 days
Every crawler a company documents is counted under that company. Companies below 1,000 requests are pooled.
| Company | Requests · last 30 days |
|---|---|
| OpenAI (ChatGPT) | 545,131 |
| Amazon (Alexa) | 132,190 |
| Anthropic (Claude) | 98,608 |
| ByteDance | 63,374 |
| You.com | 43,750 |
| Perplexity | 36,408 |
| Other AI crawlers | 511 |
Google’s Gemini has no crawler of its own and cannot be counted; Googlebot is a search crawler and is never included.
Per day, last 30 days
AI/LLM crawlers (emerald) against classic search-engine crawlers (zinc). Same floor caveat.
How we count
- 01Unit: one HTTP request to rightaichoice.com from a named AI or LLM crawler that our origin served (refused and dead requests are excluded), counted by user agent.
- 02Windows: rolling 24 hours and rolling 30 days in UTC, measured in whole-hour buckets (the current partial hour is included). Recomputed on demand, at most every 15 minutes; this page refreshes within the hour.
- 03Counted: the AI-assistant and AI-training agents we recognise — gptbot, oai-searchbot, chatgpt-user, claudebot, claude-user, claude-searchbot, anthropic-ai, perplexitybot, perplexity-user, applebot-extended, amazonbot, bytespider, ccbot, meta-externalagent, meta-externalfetcher, mistralai-user, duckassistbot, cohere-ai, youbot, ai2bot, diffbot. Search-engine crawlers (Googlebot, Bingbot) and SEO tools are never included.
- 04A floor, not a total: requests served from the Cloudflare edge cache never reach our origin, so every figure here undercounts.
- 05Forged user agents are filtered: security scanners that impersonate AI assistants while probing for secrets (dotfiles, path traversal) are bucketed apart and excluded from every figure since 2026-09-24. Probe requests recorded before that date cannot be re-classified, so older 30-day figures may run slightly high until the window rolls past the fix.
- 06Human traffic is never included. The same numbers are published as JSON so anyone can check them.
Source: the hourly `crawler_hits` rollup written by our request proxy since 30 Aug 2026, classified by user agent in this codebase. Read the same figures on /submit and in /api/vendor/live-stats.
Questions
What exactly is one "request" here?
One HTTP request to rightaichoice.com from a user agent we recognise as an AI or LLM crawler, counted at our origin server by user-agent string. HEAD and GET both count. Human browsers are never counted.
Why do you call the numbers a floor?
Because requests served from the Cloudflare edge cache never reach our origin, and only origin requests are counted. The true number is higher; we do not estimate how much.
What about bots that fake an AI user agent?
Security scanners impersonate AI assistants while probing for secrets (/.git, /.env, path traversal). Since 2026-09-24 those requests are bucketed apart at the recorder and excluded from every figure on this page. Probe requests recorded before that date cannot be re-classified, so 30-day figures include a small legacy overcount until the window rolls past the fix.
Which windows, and how often do they refresh?
Rolling 24-hour and rolling 30-day windows in UTC, recomputed at most every 15 minutes. The page and the JSON refresh on the same schedule.
Are Googlebot and Bingbot included?
No. Search-engine crawlers and SEO tools (Semrush, Ahrefs) are excluded from every AI figure. The daily chart shows them separately so the two can be compared.
Can I use these numbers?
Yes — CC BY 4.0, cite rightaichoice.com/ai-crawl-stats. The JSON at /api/vendor/live-stats carries every variant, the bot list, the exclusions and the method, and is the source to cite for a figure at a given time.