# See which AI crawlers read your site, from Cloudflare

> Connect Cloudflare to Picked and see which AI crawlers, GPTBot and ChatGPT-User among them, open your pages. Read-only, 30 days back, no code to add.

Canonical: https://picked.so/integrations/cloudflare
Last updated: 2026-10-01

Connect Cloudflare and Picked reads your site's analytics to show which AI crawlers, like GPTBot, ChatGPT-User and ClaudeBot, opened which pages, how often, and which requests were turned away. It needs no code on your site and reads the last 30 days as soon as you connect.

- **30 days** of AI crawler visits read the moment you connect
- **Hourly** sync after that, until you disconnect

> Count the fetches, not the training crawls.

GPTBot and ClaudeBot hits mean your page may end up in a training set someday. ChatGPT-User and Claude-User hits mean a person asked a question just now and the assistant opened your page to answer it. [Anthropic says](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler) Claude-User fetches a page when a person asks, and [Perplexity says](https://docs.perplexity.ai/guides/bots) the same of Perplexity-User. Picked splits your hits by purpose so a training crawl doesn't look like a lead.

## What does Picked read from Cloudflare?

Picked reads your zone's request analytics: which user agents asked for which paths, with what status, each hour. It keeps the AI crawlers and drops the rest.

| On the Analytics screen | What it tells you |
| --- | --- |
| **Pages** | Which of your pages the AI bots opened, and how many times |
| **Bots** | Which bots came and which company owns each, split by why they came: a live answer, an AI agent, a search index or training |
| **Turned away** | Requests a bot made that were answered 401, 403 or 429: a firewall rule or rate limit standing between the bot and your page |

Cloudflare classifies verified bots itself, so these hits carry Cloudflare's verdict on whether the visitor is who it says. Nothing about your visitors' identities is read: it is counts by user agent, path, status and country.

## How do you connect Cloudflare?

1. **Open Analytics.** In your project, choose Connect Cloudflare.
2. **Sign in to Cloudflare and approve.** Picked asks for zone read, analytics read and offline access, which lets the hourly sync run without you signing in each time. It cannot change anything on your account.
3. **Pick the site.** If the account has several, choose the one that serves your domain.
4. **Wait a few seconds.** Picked reads the last 30 days, and then syncs every hour.

To stop, disconnect from the same screen. Picked revokes its access with Cloudflare and deletes the hits it stored.

## What should you know before connecting?

**30 days back.** Cloudflare keeps about a month of these analytics on the Free plan, and Picked reads 30 days when you connect. Older visits aren't there to read.

**Estimates on busy sites.** Cloudflare samples the analytics of high-traffic zones, so counts are close, not exact.

**Visits from AI answers need referrers.** Picked also counts people who clicked through from ChatGPT, Perplexity or Claude, but only when the zone's analytics include the referrer and query string. Cloudflare's Free plan doesn't share them, so on Free you see the crawlers and not those visits.

**Only traffic that goes through Cloudflare.** A site not proxied by Cloudflare has nothing to read.

**One source per project.** A project reads its crawlers from Cloudflare, Vercel or your server, not two at once, or the same visit would count twice. Connecting one removes the other.

**Cloudflare's own verdict.** Hits pushed from a server or Vercel are checked against the IP ranges the AI companies publish. Cloudflare hits are trusted as Cloudflare classifies them.

## When this won't work for you

**Your site isn't behind Cloudflare.** Use [Vercel](https://picked.so/integrations/vercel) or [your own server](https://picked.so/integrations/server) instead.

**You need more than 30 days of history.** Picked can't read further back than Cloudflare keeps.

**You can't sign in to the Cloudflare account.** The person who connects needs access to the site's zone.

**You expect to fix something here.** This shows what the bots do. Blocking or allowing them is a setting in Cloudflare or your `robots.txt`, and Picked doesn't change either.

## Which bots should you not block by accident?

Look at Turned away before anything else. Blocking `OAI-SearchBot` can keep your pages out of ChatGPT's search answers, and [OpenAI's crawler docs](https://developers.openai.com/api/docs/bots) say `ChatGPT-User` isn't used for that decision. [Anthropic says](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler) blocking `Claude-User` or `Claude-SearchBot` may reduce your visibility in Claude's search answers. Blocking training crawlers is your call. Crawler visits are half the picture: an [AI visibility tool](https://picked.so/) also tells you whether the answers name you.

## How we checked

**What Picked reads and how often.** the product as built on September 30, 2026: Cloudflare's GraphQL analytics, one query set per hour after the first read.

**What each bot is for.** [OpenAI](https://developers.openai.com/api/docs/bots), [Anthropic](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler) and [Perplexity](https://docs.perplexity.ai/guides/bots), read September 29, 2026.

**Setup steps.** [AI crawlers in the docs](https://picked.so/docs/ai-crawlers).

## Frequently asked questions

### What does Picked need from my Cloudflare account?

Zone read, analytics read and offline access. It reads request analytics for the site you pick and can't change your Cloudflare settings.

### Do I need to add code to my site?

No. Cloudflare already sees every request, so Picked reads the analytics from the outside.

### How far back can it read?

Thirty days, from the moment you connect. After that Picked keeps syncing every hour, so your history grows from there.

### Can it tell a real GPTBot from a fake one?

Cloudflare classifies verified bots, and Picked shows that classification. Anyone can send `GPTBot` as a user agent, so an unverified hit proves little.

### Does it cost extra?

Picked charges nothing extra for crawler analytics. See the [pricing page](https://picked.so/pricing) for plans.

### How do I disconnect?

From the same Analytics screen. Picked revokes its access with Cloudflare and deletes the hits it stored for that project.

### My site isn't on Cloudflare. What now?

Connect [Vercel](https://picked.so/integrations/vercel) if it runs there, or add the package to [your own server](https://picked.so/integrations/server). It works with Next.js, Express and any server that can make an HTTP request.
