AI traffic — declared robots and visitors sent
The AI page shows AI robot hits (per robot, per page, per day, as they declare themselves) and human visitors arriving from an AI assistant — counted as a minimum, never mixed with your usual visitors.
Snorklee measures your visits first, simply and without cookies. On the AI side, it counts what leaves a trace on your site: AI robots that declare themselves, and visitors sent by an AI assistant — as a minimum. An honest measurement looks at what happens; it does not guess.
Frédéric Gaveau
Creator and developer of Snorklee
“I built Snorklee for my own websites. I wanted to know how many people read them without forcing a cookie banner on them, and to finally see what AI robots were doing there. I use it every day on sospreuve.fr and whereismybeach.com. I build it on my own, in France, and I'm the one who answers you.”
People no longer “google”, they ask ChatGPT, Claude, Perplexity, Gemini or Mistral — and trust the answer. Meanwhile, on your site, AI robots crawl your pages, fetch them to build those answers, and send you visitors. And nothing, anywhere, shows you.
Yesterday's tools look at yesterday's web. AI bots don't run JavaScript, so a classic analytics tool doesn't see them. And most AI assistants strip where their visitors come from, so those visits land in “Direct”. We don't pretend to fill that gap: we count what can be seen, and say what can't.
On your site, AIs leave two traces that can be counted. Hits from their robots, which declare themselves through their user-agent — visible with a server integration (WordPress plugin or server beacon). And the visitors an AI assistant sends you — a minimum, because most AI apps strip the referrer. Snorklee counts both traces on your own traffic, never guessed.
To measure that path you have to be the site's own measurement tool — not a robot firing thousands of fake questions at the AIs to guess a “visibility”. AI visibility tools guess from the outside. Classic measurement tools can't tell the AIs apart. Snorklee watches from the only reliable place: your own traffic. Measured without getting in the way, never guessed.
Snorklee does two things. The first, the core: simple visit statistics without cookies, on par with Plausible or Simple Analytics, for less. The second: an AI page showing which AI robots read your site and how many visitors AI assistants send you — without inflating anything.
Snorklee counts AI robot hits as the robots declare themselves (GPTBot, ClaudeBot, PerplexityBot…), per robot, per page and per day — with a server integration. And the visitors sent by an AI assistant, shown as a minimum: most AI apps don't pass on where the visit came from.
The same tool measures your real visits: where people come from, which pages they read, what makes them stay. No cookie, no banner, no tracking anyone — and humans arriving from an AI chat are counted separately.
These refusals aren't promises: they're already in the architecture, and verifiable.
No cookie, no browser fingerprint, no identifier that follows you from one site to the next. The marker that tells two visits apart is recomputed every day and never kept; the IP address is never stored. And as soon as a table crosses two pieces of information, we hide numbers that are too small so it can never point to a person — site owner included.
Not a box ticked in small print: a design outside the scope of consent. No consent-bound tracker, anonymised statistics. The banner isn't dodged — it's simply moot.
A 100 % EU stack: Clever Cloud (Paris, ISO 27001:2022), DB-IP (France), Brevo (France), AI operated in France, zero external CDN. Sovereignty isn't a marketing claim: it's a design constraint we impose on ourselves.
We observe what GPTBot, ClaudeBot or PerplexityBot do on your site — your audience data never touches ChatGPT, Gemini or any American AI.
90 days for the detailed data, 25 months for the totals, nothing archived elsewhere. CSV/JSON export and erasure (GDPR art. 17) in one click. Your data doesn't outlive its usefulness.
Thirty years of audience measurement in nine milestones. Pioneers, ad-tech monopoly, GDPR earthquake, waves of fines, AI breakthrough — and the gap that was left to fill.
Webhits, then Analog (1995), invent reading Apache log files. Web analytics is born server-side: one number, one page, zero JavaScript. Privacy is implicit — there's simply no one to track.
Google Analytics launches. Free, powerful, tagged everywhere. The implicit deal: you measure your site, your data flies to Californian servers. For a billion websites, it became the default for fifteen years.
Adopted in 2016, enforceable on 25 May 2018, the GDPR requires transparency, minimisation, data-subject rights and strict safeguards for transfers outside the EU. Cookie banners flood the web.
GDPR opens a market: Fathom Analytics (2018, Canada / UK), Simple Analytics (October 2018, Netherlands) and Plausible Analytics (April 2019, Estonia, open source) all answer the same question: can you measure an audience without cookies, without cross-site IDs, without reselling data? The first generation of concrete answers.
On 16 July 2020, the Court of Justice of the European Union hands down ruling C-311/18 (Schrems II) and strikes down the Privacy Shield. The operational takeaway is simple: international data transfers must be documented, contractually governed, and checked case by case. Its successor, the Data Privacy Framework (July 2023), remains closely watched.
Austria fires the opening shot in January 2022 (DSB). France's CNIL follows on 10 February. Then Italy (Garante), Denmark, Norway, Finland, the Netherlands. The shared verdict: GA exposes European visitors to US surveillance. Public sites must migrate urgently.
Meta is hit with €1.2 billion in May for illegal EU → US transfers (Ireland's DPC). TikTok gets €345 million for handling minors' data. Criteo is fined €40 million for ad targeting without consent (France's CNIL). Cumulative GDPR fines to date: over €4 billion.
AI assistants become a traffic source in their own right: they cite pages, send crawl bots, and refer visitors from their chat windows. Classic analytics tools count them as humans — or ignore them as noise. Marketers fly blind.
Snorklee takes the empty seat: simple cookieless analytics hosted in Paris, cheaper than the usual references — plus an honest AI page that counts declared AI robots and the visitors AI assistants send, as a minimum. No personal data, humans and robots counted separately.
Snorklee is an independent product, built in France. It lives on its subscriptions — not on your data, not on advertising, not on resale.
The price is published: one plan, Snorklee One, from €4/month ex. VAT — everything included, up to 10 sites. 14-day free trial, cancellable online at any time. No “talk to sales”, no hidden pricing, no dark patterns.
And nothing here is a roadmap promise: everything above is already in the product, verifiable from the dashboard.
Three screens in the dashboard: statistics, the AI page, settings.
The AI page shows AI robot hits (per robot, per page, per day, as they declare themselves) and human visitors arriving from an AI assistant — counted as a minimum, never mixed with your usual visitors.
In the settings, installation section: drop in the Snorklee code, check it works, per-platform guides, WordPress plugin, and capturing AI robots on the server, CDN or platform side. A clean install in a few minutes.
In the settings, privacy section: Status & documents, Electronic DPA signature, Retention periods, Export & deletion, data protection contact. A verifiable working base, not legal advice.
Fourteen days to try it — free trial, no card, cancellable online, sovereign analytics included.
€4 per month ex. VAT — 100,000 pageviews, 10 sites, everything included.