WebmasterID logoWebmasterID
AI visibility analytics

See which AI crawlers and AI assistants actually reach your site

WebmasterID is AI visibility analytics for modern websites. Server-side detection of GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and the AI-assistant referrals that follow. Deterministic categorisation, privacy-first event store, operator-focused dashboard.

Use cases

What AI visibility analytics is for

Operator-grade visibility into the AI side of your traffic. The goal is not to label every visitor with an AI tag; the goal is to give you a clean coverage signal for the AI surfaces that matter.

Confirm GPTBot or ClaudeBot reached a page

Per-page crawl timeline for known AI crawlers. No guessing — the event is recorded server-side or it is not.

Track Google-Extended coverage

Separate Google-Extended from Googlebot so AI training opt-out is observable, not just declared in robots.txt.

See AI-assistant referrals after a crawl

When a human follows a link from ChatGPT, Claude, Perplexity, or Gemini, that referral is captured and joined back to the crawl history.

Compare AI vs search-engine coverage

Side-by-side counters for AI crawlers and traditional search crawlers. Useful for editorial teams investing in AI-search readiness.

Detect a new AI crawler

When a never-before-seen AI user-agent appears, it is logged and visible in the Bot Intelligence view. Uncategorised stays uncategorised — never invented.

Export the AI visibility view

CSV or NDJSON export of the filtered AI-traffic slice. Goes into reports, warehouses, or a one-off investigation.

How WebmasterID helps

A small, deterministic pipeline for an AI-search-era signal

The product does one thing well: it records the signal honestly. The dashboard is the read surface; the Event Explorer is the investigation surface; the Agent + MCP layer is the AI-assisted reading surface.

  1. 1. Tracker + ingest. A small browser tracker emits page-view events. The ingest API classifies the request server-side: AI crawler, search-engine crawler, automation, or human.
  2. 2. Deterministic categories. Categorisation is rule-based against a maintained signature list. Uncategorised user-agents land in 'other' and are not speculatively labelled as AI.
  3. 3. Referrer normalisation. AI-assistant referrals (ChatGPT, Claude, Perplexity, Gemini, AI Overviews) are normalised at write time, so the AI visibility view stays clean even as referrer formats change.
  4. 4. Dashboard + Event Explorer. Read the rollup in the dashboard. Drill into the underlying events in the Event Explorer. Export what you need.
Feature

The AI visibility surface, in plain terms

No surveillance metrics, no probabilistic AI labels, no fake confidence scores. Just the signal.

AI crawler timeline

Per-bot crawl timeline (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, anthropic-ai, OAI-SearchBot, and others) with timestamps and frequency.

AI assistant referrals

Human visits that arrived from ChatGPT, Claude, Perplexity, Gemini, or AI Overviews. Filterable by date, country, source, and pathname.

Per-page coverage

Which of your pages have been seen by which AI surfaces, and which have not. Useful for editorial prioritisation.

New-bot detection

Surface AI crawlers never seen before on this site. Helpful when a new AI assistant launches its first crawl wave.

Privacy & trust

Privacy-first by architecture, not by toggle

The product does not need to know who your visitors are. It needs to know what AI surfaces reached your pages. That is a much smaller surface area.

  • No third-party cookies

    The tracker never sets cookies. The product has no concept of a single user across sites.

  • No fingerprinting

    No canvas, audio, fonts, or device-entropy signals are read. The tracker is small and the request payload is short.

  • IP anonymisation at the edge

    IPv4 last octet zeroed; IPv6 truncated to /48. Raw IPs never land in storage.

  • DNT / GPC respected

    Browsers signalling Do Not Track or Global Privacy Control send nothing. Enforced client-side and at ingest.

For privacy questions, write to info@helperg.com. See also /privacy-first-analytics and /privacy-policy.

FAQ

AI visibility analytics, answered

What does AI visibility analytics actually measure?
Two things, recorded server-side: (1) which AI crawlers fetched which pages, and (2) which human visits arrived with a referrer that identifies an AI assistant (ChatGPT, Claude, Perplexity, Gemini, AI Overviews, and similar). WebmasterID stores both signals deterministically — no probabilistic AI labelling.
How does WebmasterID identify an AI crawler?
By matching the request user-agent against a maintained list of known AI crawler signatures. Categories include AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, anthropic-ai, OAI-SearchBot, and similar), search-engine crawlers (Googlebot, Bingbot, DuckDuckBot), and automation. Uncategorised user-agents land in 'other' and are reported as such.
Is this useful if my site has low traffic?
Yes. AI visibility analytics is a coverage signal, not a volume signal. A small site can still confirm 'GPTBot fetched my pricing page this week' or 'Claude sent a human to this article' — which is exactly the information teams use to decide where to invest in AI-search-era content.
Do you train AI models on my analytics data?
No. WebmasterID is privacy-first. Event data belongs to the workspace that recorded it. We do not train shared models on it, we do not sell it, and we do not share it with third parties. The Claude/MCP integration is read-only and operator-controlled.
How is this different from server log analysis?
Server log analysis is powerful but requires log access, parsing, and ongoing pipeline work. WebmasterID gives you the same coverage signal as a managed product: a small tracker for the human side, server-side detection for the bot side, a deterministic categoriser, and a dashboard you can read in five minutes.
Can I export the AI visibility data?
Yes. The Event Explorer supports CSV and NDJSON export of the filtered event set. Operator approval is required; cross-workspace exports are impossible by construction.