Home / AI Coding Tool Routes: Website Tracking Interest Study
AI coding tool routes: website tracking interest and checkout intent
A first-party study of 6 SessionWatcher website routes (Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, opencode): how many visitors reached each route in a fixed window, and how many of those visitors separately completed the sitewide checkout-intent goal in the same window.
Last updated 2026-08-17 · Source data verified 2026-08-17 · By Soren Starck
Read this before the data: what this study is and is not
- Not app telemetry. SessionWatcher ships with no telemetry. This study never observes in-app usage, active tool use, or installed-app behavior — only aggregate first-party website analytics from sessionwatcher.com.
- Not market share. Route visitor counts describe interest among visitors to this one website, not usage or popularity of any tool in the wider market.
- Not a purchase claim. The checkout-intent goal (
checkout_clicked) records a "buy button clicked" intent event that can fire anywhere in a visitor's journey — before or after visiting the route — on any product page. It is not a completed purchase, and it is not claimed to be caused by the route visit. - Cohorts overlap. Cohorts overlap: a visitor can appear in more than one route row. Do not sum the six routeVisitors or the six checkoutIntentVisitors counts and present them as a unique-visitor total.
Headline findings
- Codex had the largest route cohort: 1,225 visitors to
/codexin the window. - Cursor had the highest descriptive checkout-intent ratio, 16.7% — but its cohort was only 72 visitors, far smaller than the largest cohorts. A ratio computed on a small cohort is not comparable in precision to one computed on a cohort in the hundreds or thousands.
- Neither finding is a market-share, popularity, or conversion-causation ranking. Both are descriptive statements about this website's own visitor base in this window only.
The data
Each row is one tool-specific commercial route on sessionwatcher.com. "Route visitors" is the count of distinct DataFast-tracked visitors who visited that exact path during the window. "Checkout-intent visitors" is the subset of those visitors who also completed the sitewide checkout_clicked goal at any point during the same window. "Descriptive ratio" is checkout-intent visitors divided by route visitors — plain division of the counts, not a statistical estimate.
| Route | Provider | Route visitors | Checkout-intent visitors | Descriptive ratio |
|---|---|---|---|---|
/claude | Claude Code | 719 | 78 | 10.8% |
/codex | Codex | 1,225 | 79 | 6.4% |
/cursor | Cursor | 72 | 12 | 16.7% |
/copilot | GitHub Copilot | 48 | 3 | 6.3% |
/gemini | Gemini CLI | 54 | 1 | 1.9% |
/opencode | opencode | 72 | 1 | 1.4% |
Cohorts overlap: a visitor can appear in more than one route row. Do not sum the six routeVisitors or the six checkoutIntentVisitors counts and present them as a unique-visitor total.
Why only these six routes
SessionWatcher currently supports more providers than are studied here. These six routes are the exact tool-specific commercial routes that existed throughout the entire frozen analysis window on sessionwatcher.com. The remaining supported providers (Antigravity, Devin, Grok, OpenRouter, Z.ai, Kimi / Kimi Code workflows, Qwen) are not rows in this study: These registry providers did not have an equivalent tool-specific commercial route that existed throughout the entire frozen common analysis window (2026-05-25..2026-07-31) on sessionwatcher.com, so including them would compare unequal windows as if they were equal — false precision.
Methodology
For each route, two authenticated DataFast CLI queries were run against website id 699eb8beb26e6654a29d2b1f (timezone Asia/Ho_Chi_Minh), restricted to the complete fixed window 2026-05-25 through 2026-07-31 inclusive:
- A route-cohort query, filtered to visitors of the exact route path, read as "route visitors."
- The same query with an added completed-goal filter for
checkout_clicked, read as "checkout-intent visitors."
Limitations
- Aggregate first-party website analytics only — not app telemetry, not in-app usage.
- Describes SessionWatcher's own visitor base in this window, not the wider market.
- Cohorts overlap; the six rows must never be summed into a unique-visitor total.
checkout_clickedis a sitewide intent goal, not a completed purchase, and not causally attributed to the route visit.- Ratios are plain division with no statistical test, confidence interval, or claim of representativeness beyond this website's visitors in this window.
- Small cohorts (e.g. Copilot's 48, Gemini's 54, Cursor's and opencode's 72) produce ratios that move a lot per additional visitor; they are reported as-is, not smoothed or modeled.
Reproducibility
Every row can be independently reproduced with an authenticated DataFast CLI session. Example for /claude:
datafast --json visitors list --website 699eb8beb26e6654a29d2b1f --from 2026-05-25 --to 2026-07-31 --visited-page /claude --limit 1datafast --json visitors list --website 699eb8beb26e6654a29d2b1f --from 2026-05-25 --to 2026-07-31 --visited-page /claude --limit 1 --completed-goal checkout_clickedSubstitute the route path for any of the other five routes to reproduce that row. In both cases, the number reported is pagination.total from the JSON response.
Source and verification
Sole source for all six rows: DataFast (authenticated CLI), website id 699eb8beb26e6654a29d2b1f, verified 2026-08-17. Provider identity and display labels are read from SessionWatcher's canonical release registry (app v6.5.50), verified 2026-08-11 — see the full supported-provider map.
Window provenance (Git, verified 2026-08-17): the frozen window start 2026-05-25 is derived from this website repository's own commit history, not chosen arbitrarily:
/claude, /codexwere added 2025-09-25 (commit8f5fc5ddbc7c9632854c5eabf1377f3d1310dd59)./cursor, /copilot, /geminiwere added 2026-05-04 (commitde71f87363638ca915eba96e068c79d9397b02da)./opencodewas added 2026-05-24 (commit1c2fc44b972d074513f44237f7e2b1d5c255d9c9).
/opencode was the last of the six studied routes to exist, so the first complete common day on which all six existed — this study's window start — is 2026-05-25.
SessionWatcher is an independent product and is not affiliated with, endorsed by, or sponsored by Anthropic, OpenAI, Cursor, Microsoft, or Google. Product and provider names are trademarks of their respective owners.
Measurement note for this page
This is a Lane B/C authority-asset experiment (candidate c_15de27e75c3b). Its frozen success criterion is at least 2 qualifying normalized independent editorial referring hostnames, each with at least 2 distinct visitors to this exact page in DataFast, over the first 90 complete days after it becomes indexable. That clock has not started: the first complete indexable day has not yet been verified. Search engines, social networks, paid/UTM traffic, link shorteners, localhost/preview hosts, sessionwatcher.com itself, and AI assistants are excluded from this count (AI-assistant referrals are tracked separately as a diagnostic, never as a success criterion).
This is a brand-new URL with no page-level baseline. Full contract: config/ai-coding-tool-usage-gate.mjs; reviewer record: AI-CODING-TOOL-USAGE-GATE.md.