Skip to content
How it works

How SourceWatch tracks AI search.

Connect your sites once. SourceWatch then queries ChatGPT, Perplexity, Gemini and Claude every day, captures the real AI crawlers and visitors hitting your pages, and pipes it all into Claude Code so you can act. Here is the whole pipeline, end to end.

Three steps

From connected to cited, in one loop.

No tags to manage, no spreadsheets to babysit. Connect once and the tracking runs itself.

  1. Connect your sites

    Drop in a Cloudflare Worker or a one-line snippet. Every site you own reports into one dashboard, tagged automatically — no code per page.

    • Add a Cloudflare Worker for zero per-site code, or paste a one-line Next middleware / server snippet.
    • Each site gets its own ingest key, and every row is keyed by property in one shared dashboard.
    • Set your business info, brands and the prompts you want tracked — or generate prompts automatically.
  2. Track the engines

    SourceWatch queries ChatGPT, Perplexity, Gemini and Claude with your prompts every day, and captures which AI crawlers and AI visitors hit your pages.

    • A daily cron run asks ChatGPT, Perplexity, Gemini and Claude your prompts across regions.
    • Each answer is scored for mention rate, self-citation, share of voice, sentiment and brand rank.
    • Captured traffic shows which AI crawlers read your pages and which real visitors arrive from AI.
  3. Act on it in Claude Code

    The MCP server pipes the data into Claude Code, so you can turn citation gaps and the real queries into optimized content — in the same loop.

    • The MCP server pipes live visibility and traffic data straight into Claude Code.
    • Run a page-level GEO/AEO audit, then generate answer-first content from the real captured queries.
    • Close citation gaps in the same loop — no dashboard round-trips, no copy-paste.
The pipeline

Two tracks, one dashboard.

Traffic and visibility run on separate rails, then meet in Claude Code. Here is the path your data takes end to end.

First-party data

Traffic track

Measured straight from your own traffic — never inferred. One install covers an entire site.

Your site

Cloudflare Worker or a one-line middleware snippet. Zero per-page code.

/track ingest

Every request streams to your per-site ingest key.

Classify

Sort AI crawlers from real AI referrals, per property.

Verify IP

Checked against vendor IP ranges — spoofed bots rejected.

Daily auto-run

Visibility track

SourceWatch queries every engine on a schedule and scores how the answer represents you.

Daily cron

Auto-runs every day, across regions you choose.

Query 4 engines

ChatGPT, Perplexity, Gemini and Claude, with your prompts.

Read answers

Capture the real queries the model ran and what it cited.

Score

Mentions, self-citation, share of voice, sentiment, gaps.

Both tracks feed Claude Code

The SourceWatch MCP server hands live visibility and traffic data to Claude Code. Audit any page for AI-readiness, then generate answer-first content from the real captured queries — closing citation gaps in the same loop.

Why the data holds up

Verified and first-party, not guessed.

  • Every bot hit is IP-verified

    Crawler traffic is checked against each vendor's published IP ranges, so a spoofed GPTBot can't inflate your numbers.

  • Traffic is your own, not modeled

    Most tools only run synthetic prompts and infer the rest. We do that for visibility — and pair it with the real AI traffic hitting your pages.

  • You see the actual queries

    Instead of a single score, you get the real searches the model ran before it answered — the biggest factor in who gets cited.

See where AI sends your customers — before your competitors do.

Connect your first site and watch SourceWatch score your AI visibility in minutes.