Everything people ask before they connect

Arrivl is growth analytics for AI agent traffic. It shows you which AI agents — ChatGPT, Claude, Perplexity, and a hundred more — visit your site, what they read, whether they send you real customers, and what to fix so you show up in their answers.

AI agents are how more people discover products every day. When someone asks ChatGPT or Claude for a recommendation, those tools visit websites to decide what to say. If you can't see them, you have no idea how your site shows up in the answers your future customers are getting.

Google Analytics tracks people through a JavaScript beacon in the browser. AI agents don't browse that way — most never run your JavaScript — so GA filters the whole channel out as noise. Arrivl captures agent traffic server-side, where it's actually visible. The two tools cover different audiences; use both.

Several different things, and the difference matters: training crawlers (like GPTBot) that read pages to feed future models, search indexers (like OAI-SearchBot) that decide what AI search can cite, live fetchers (like ChatGPT-User) that visit right now because a real person asked something, and agentic browsers (like Claude for Chrome or ChatGPT Atlas) that act on behalf of a user. Arrivl names each one individually instead of lumping them into “bot traffic.”

How well your site reads to a machine: whether your key content is in the server-rendered HTML, whether structured data spells out your facts, whether your robots and llms.txt signals welcome the agents you want, and whether pages respond fast and unblocked. Arrivl's audit inspects the signals that shape this and scores each one with a recommended fix.

Arrivl identifies agents by their user agent strings, IP ranges, and request patterns. Major AI companies publish official identifiers for their bots, and we maintain an updated list as new ones launch. For agentic browsers like Claude for Chrome and ChatGPT Atlas, we use behavioral signals to separate them from humans.

For agents with official identifiers, detection is close to 100%. For newer or less transparent agents, we'd rather flag traffic as "unidentified AI" than mislabel it. You'll always know what we're confident about and what we're still classifying.

104 agents and counting, across OpenAI, Anthropic, Google, Perplexity, xAI, Mistral, Meta, Apple, Amazon, DeepSeek, Cohere, HuggingFace, ByteDance, Common Crawl, AI2, and more — plus the data-infrastructure crawlers (Firecrawl, Diffbot, Apify, Bright Data) that feed AI products. Each is tracked individually, because each behaves differently.

Yes. Scrapers fake ChatGPT or Claude user agent strings to slip past filters. Arrivl checks every request against the real agent's published fingerprint — official IP ranges and request patterns — so spoofed traffic stays out of your AI numbers instead of flattering them.

Intent. A training crawler reading your blog matters over months — it shapes what future models know about you. A live fetch means a real person, in a conversation, right now, triggered a visit to your page — that's the closest thing this channel has to purchase intent. Arrivl labels each visit's intent (research, indexing, training) so you can tell them apart.

When a person lands on your site from an AI tool that passes a referrer — chatgpt.com, claude.ai, perplexity.ai, gemini.google.com — Arrivl ties that human visit back to the AI that sent it. Honest caveat: some AI surfaces don't pass referrers at all, so this number is a floor, not a ceiling. What you see is real; there's more we can't attribute.

A few minutes. Connect through Cloudflare with zero code, add a few DNS settings on Shopify, or drop one server-side snippet anywhere else — and data starts flowing immediately.

It depends on your platform, and either way the setup is about five minutes. On Cloudflare you just authorize — no code at all. On Shopify you add a few DNS settings — no app, no theme changes. On other platforms, a small backend snippet wires it in. If you'd rather not do it alone, we'll set it up with you on a quick 1:1 call.

Enter your domain, authorize Arrivl, done. We deploy the tracking on your own Cloudflare account through the official OAuth flow — no code, no snippet, nothing to paste, and you can disconnect it from your Cloudflare dashboard any time.

Yes — and without an app or theme changes. A few DNS settings connect your store, it works on every Shopify plan, and we're happy to do it with you on a call. One known blind spot: Shopify doesn't allow tracking on checkout pages, so agent visits stop being visible at checkout.

Yes — we've packaged the whole install into a prompt. Give it to Codex, Claude Code, Cursor, or any AI code editor, and it handles the setup from there.

Because agents don't run JavaScript. A browser beacon only fires when a real browser executes it — which is exactly what most crawlers and fetchers never do. Tracking agents requires seeing the raw request server-side: at the edge (a Cloudflare Worker), in your backend, or from server logs. That's why Arrivl installs where the requests actually arrive.

Often, yes. If your site has its own server or backend, there is usually a way. Tell us what you use and we will check.

No. Arrivl watches your traffic quietly in the background. It never gets in the way of a page loading.

Only the traffic hitting your site: which agents and people visit, which pages they read, and when. Arrivl never sees your customer data, orders, account info, passwords, or anything inside a checkout. It cannot reach into your platform admin or your backend database.

Your existing tools like GA and Amplitude stay the system of record for human traffic — Arrivl doesn't replace them. The one place humans appear in Arrivl is attribution: when a visitor arrives from an AI tool's link, we count them as an AI-referred human so you can see which agents send real people.

Yes, and you should. Google Analytics tracks the people browsing your site. AI agents do not browse the way people do, so they never show up there. Arrivl is the part that does. Use both.

27 check points across the signals that shape how AI reads a site — server-rendered content, structured data, robots and llms.txt signals, heading structure, response health, and more. Each check is inspected, scored, and paired with a recommended fix. You can run it on your own site or a competitor's.

Arrivl is free during beta. Sign up now and you'll get full access while we build the product with early users.

We'll introduce paid plans at some point — that's no secret. Early users will hear from us well before anything changes, and being in early is how you get a say in what the plans look like.

Yes — each client site is its own project with its own dashboard, API key, and history, all under one account. There's a dedicated page on how agencies use Arrivl at arrivl.ai/agency.