Which AIs read your site? And what they bring you.
With the WordPress plugin or the server code, Snorklee counts every AI bot hit on your pages, under the name the bot declares. It also counts the visitors ChatGPT or Perplexity send you — at least those whose app passes on where they came from.
Three questions. Three measured numbers.
No made-up score: every number comes from your own traffic.
Which AIs read your site
17 AI bot families recognized by their user-agent, including GPTBot, ClaudeBot and PerplexityBot. 29 signatures in total.
How many times
Page by page, day by day, bot by bot.
How many visitors they send back
Every visit from ChatGPT, Perplexity or Claude is counted separately, when the app passes on where it came from. Many do not: this number is a minimum.
Your Google and Bing keywords, on the same screen as your visitors.
You connect your Search Console and your Bing account. Snorklee reads, and changes nothing.
Google Search Console
One-click connection, read-only. Snorklee asks for the "webmasters.readonly" scope, nothing more.
Bing Webmaster Tools
An API key is enough. Most analytics tools read Google only — Bing stays a blind spot.
Clicks, impressions, position
Your top keywords and top pages, with click-through rate and average position.
Refreshed every night
None of your visitor data goes to Google or Microsoft. Snorklee never pushes anything.
Your Google searches, your Bing searches, your visitors and your AI bots: one page to read.
The others ask the AIs questions. We read your server.
Peec, Profound and the rest send made-up questions to the AIs, then estimate your visibility. Snorklee counts what actually reaches you.
Tools that ask made-up questions ℹ
- 1,000 made-up questions, sent to the AIs
- An estimated presence percentage
- Manufactured data, outside your site
Snorklee
- The traffic your server receives
- Hits and visits counted one by one
- Your data, no cookies, hosted in Europe
Snorklee does not know what question was asked before a visit: nobody gets that information.
17 AI bot families. 29 recognized signatures.
Snorklee recognizes each bot by its user-agent: the name it declares itself.
A user-agent can be forged: these numbers say what the bots declare, not a proven identity.
OpenAI
3 user-agents
-
GPTBot— associated with model training -
ChatGPT-User— on-demand browsing -
OAI-SearchBot— SearchGPT engine
Anthropic
3 user-agents
-
ClaudeBot— Claude indexing crawler -
anthropic-ai— legacy Anthropic agent -
Claude-Web— on-demand browsing
-
Google-Extended— associated with Gemini / Vertex AI
Perplexity
2 user-agents
-
PerplexityBot— indexing crawler -
Perplexity-User— on-demand browsing
Mistral
1 family
-
MistralAI/Mistralbot— Mistral crawler (Le Chat)
Meta
2 user-agents
-
FacebookBot— associated with Llama / Meta AI -
Meta-ExternalAgent— Meta AI external agent
ByteDance
1 user-agent-
Bytespider— ByteDance AI crawler (Doubao)
xAI
2 user-agents-
Grokbot— indexing crawler -
Grok-User— on-demand browsing
Other LLMs & archives
9 user-agents-
Applebot-Extended— Apple Intelligence -
cohere-ai— associated with Cohere -
CCBot— Common Crawl archive (base dataset for many LLMs) -
YouBot— You.com engine -
DuckAssistBot— DuckDuckGo Assist -
Amazonbot— Amazon training (Alexa+) -
PetalBot— Huawei AI engine -
DeepSeekBot— DeepSeek crawler (China) -
Diffbot— structured extraction (used by multiple LLMs)
Four views. No estimated number.
What the product’s AI page shows. Everything comes from your real traffic, nothing is extrapolated.
Day after day
over the chosen periodBot hits and visits from assistants, day by day. Never added together.
The assistants that send you visitors
at leastChatGPT, Perplexity, Claude, Gemini: how many visits each, when the app passes on where they came from. A minimum, shown as such.
AI bots
family + purposeEvery bot under the name it declares, with the use its vendor announces: collecting for training, indexing for a search engine, or reading to answer.
Your pages, as AIs see them
page by pageWhich pages each bot reads, and how many times.
All of this is open from day one of your 14-day free trial.
Two sensors. No cookies.
- Visitors from an AI chat: your tag is enough. When the app passes on where the click came from, Snorklee recognizes the assistant and counts the visit separately. Many apps do not: those visits stay in “Direct”, and the number shown is a minimum.
- AI bots: captured on your server. GPTBot or ClaudeBot read your page without ever rendering it: a plain tag never sees them. Add the small server snippet — ready-made code and a step-by-step guide, without even creating an account.
- Every bot under the name it declares. Anyone can claim to be GPTBot: a user-agent can be forged. Snorklee counts what the bot declares, and never presents it as a proven identity.
- It all lands on one screen. Your visitors, your Google and Bing keywords, and your AI bots. No cookie is set, for either of them.
Who these bots actually are.
We are often asked whether GPTBot and ChatGPT-User are the same bot. They are not. The first collects pages to train a model; the second opens your page while an AI answers someone. Blocking both under one policy cuts you off from traffic you wanted to keep. So here is, bot by bot, what each one does according to its vendor.
| Bot | Vendor | What it does |
|---|---|---|
GPTBot |
OpenAI | Collects to train a model |
OAI-SearchBot |
OpenAI | Indexes for search |
ChatGPT-User |
OpenAI | Reads the page live to answer |
ClaudeBot |
Anthropic | Collects to train a model |
Claude-User |
Anthropic | Reads the page live to answer |
PerplexityBot |
Perplexity | Indexes for search |
Perplexity-User |
Perplexity | Reads the page live to answer |
Google-Extended |
Collects to train a model | |
Applebot-Extended |
Apple | Collects to train a model |
Meta-ExternalAgent |
Meta | Collects to train a model |
Bytespider |
ByteDance | Collects to train a model |
CCBot |
Common Crawl | Collects to train a model |
Amazonbot |
Amazon | Undetermined purpose |
mistralai-User |
Mistral | Reads the page live to answer |
DuckAssistBot |
DuckDuckGo | Reads the page live to answer |
PetalBot |
Huawei | Collects to train a model |
A number is worthless if you do not know what it proves. Our three counters do not measure the same thing: we never add them up. Here is what each one establishes, and above all what it does not.
| What we counted | How it is captured | What it proves | What it does not prove |
|---|---|---|---|
| An AI bot read the page | Bot user-agent, read server-side | A request declaring itself as this bot fetched this page | Does not prove the page will be cited, nor that it really is this bot |
| An AI opened the page to answer | Bot whose stated purpose is “assistant” | An AI opened the page during a conversation | Does not say what the AI did with it |
| A visitor came from an AI chat | Visit referrer (chat.openai.com, perplexity.ai…) | A human came from an AI answer | Misses arrivals without a referrer: a minimum |
Frequently asked questions.
How much does it cost?
How do I block GPTBot on my site?
Which AIs are recognized?
Do you send questions to the AIs?
Do I need a cookie banner?
Do you guarantee ChatGPT will cite me?
Do I need to install anything?
See which AIs read your site.
€4 per month ex. VAT. 100,000 pageviews and 10 sites. 14-day free trial, no credit card.