---
title: Count AI bot visits
description: See how often ChatGPT, Gemini, Claude, Perplexity, Copilot and other AI bots request your site, from Cloudflare, Vercel or any log drain.
sidebar:
  label: Overview
---

AI assistants send bots to your site for two reasons. Crawlers read your pages ahead of time to build an index. Live fetchers open a page while they write an answer for a user. Both show that an assistant can reach your content. Zero visits from a bot usually means it is blocked or cannot find your pages.

Sommet counts these visits per assistant and per day, over the last 30 days. You see them in the **AI bots** line of the overview, and assistants can read them through the [MCP server](/docs/mcp/tools).

## Bots Sommet counts

Sommet matches the user agent of each request against these names. The match ignores case.

| Assistant  | User agents                                                                       |
| ---------- | --------------------------------------------------------------------------------- |
| ChatGPT    | `GPTBot`, `OAI-SearchBot`, `ChatGPT-User`, `ChatGPT Agent`                        |
| Gemini     | `Googlebot`, `Gemini-Deep-Research`, `Google-CloudVertexBot`                      |
| Claude     | `ClaudeBot`, `Claude-User`, `Claude-SearchBot`, `anthropic-ai`                    |
| Perplexity | `PerplexityBot`, `Perplexity-User`                                                |
| Copilot    | `bingbot`                                                                         |
| Meta AI    | `meta-externalagent`, `meta-externalfetcher`, `FacebookBot`                       |
| Apple      | `Applebot`                                                                        |
| Other AI   | `MistralAI-User`, `DuckAssistBot`, `Amazonbot`, `Bytespider`, `CCBot`, `Timpibot` |

:::note
Google and Microsoft have no separate crawler for their assistants. AI Overviews, AI Mode and Gemini answer from the pages Googlebot indexes, and Copilot from the pages Bingbot indexes, so Gemini and Copilot also count classic search crawling. Crawlers of search engines without an assistant, such as Baiduspider or PetalBot, are not counted.
:::

## Choose a source

A workspace has one source. Sommet reads the response headers of your homepage and selects the source that matches your host. Only the workspace owner can connect it.

| Your site runs on          | Use                                 | How it works                                               | Delay                    |
| -------------------------- | ----------------------------------- | ---------------------------------------------------------- | ------------------------ |
| Cloudflare (proxied)       | [Cloudflare](/docs/ai-bots/cloudflare)   | Sommet reads request counts with a read-only API token     | About 1 hour             |
| Vercel (Pro or Enterprise) | [Vercel](/docs/ai-bots/vercel)           | Vercel pushes request logs to a drain URL                  | A few minutes            |
| Anything else              | [Custom log drain](/docs/ai-bots/custom) | Your server, CDN or log shipper pushes logs to a drain URL | As fast as you send them |

If your site is on Vercel behind the Cloudflare proxy, use Cloudflare. It sees every request first.

## What Sommet keeps

Sommet keeps one number per day and per assistant. It does not store paths, IP addresses or the log lines it receives. To drop duplicate deliveries, it remembers request IDs for 6 hours.

## Check that it works

The **AI bots** line shows the date of the last logs received. If it says "Waiting for the first logs":

1. Request your homepage with a bot user agent. Replace the domain with yours:

   ```bash
   curl -A "Mozilla/5.0 (compatible; GPTBot/1.2; +https://openai.com/gptbot)" https://example.com/
   ```

2. Wait for the delay of your source, then reload the overview.
3. If the count stays at zero, check that your `robots.txt` and your firewall do not block the bot. See [Let AI crawlers in](/docs/improve/technical-checklist).