GTMRun

Changelog

Changelog

What changed on gtmrun.co, newest first. The site is a preview while GTMRun is rebuilt: the check is real, the copy and prices are drafts.

Last updated 13 November 2026 · Atom feed

The cards page shows the check running, and its numbers add up

  • After you add a brand or press "Run this week's check now", the page now says it is measuring and shows the Fix cards (and the post draft) the moment they are ready — no reload needed. If a check takes longer than two minutes the page says so instead of waiting silently.
  • The brand line reads, for example, "12 of 13 measured checks passing · 1 failing · 3 informational checks failing · 4 Fix cards". Before, it said "4 failing" next to "12 of 13": the 12 and 13 counted the scored checks only, while the 4 counted every Fix card, informational checks included. Both numbers were right; the sentence now says which is which.

The week's cards, for invited brands: Fix cards from a weekly check, Write cards as pull requests

  • Sign in (invite-only during the preview) and add your website: every Monday the check runs on it, and every failing measured row becomes a Fix card with its guide. Deploy the fix and press Verify: a re-check in seconds closes the card when the row passes. Rows that flip on their own close their cards on Monday.
  • Write cards: a post drafted from your own material — your homepage and your check results — after reading what the top search results say, so the draft covers what they do not. No invented facts: when nothing first-hand supports the topic, the card asks you two or three questions instead of writing filler. Approving a draft opens a pull request with the post into your repository (you store a token for exactly that repository; without one, the file is on the card to copy). Merging is publishing; nothing is posted anywhere for you.
  • A cards page at /app/ to approve, skip and verify, and the same cards for agents through the MCP endpoint (list_brands, list_cards, approve_card, skip_card, verify_card, draft_post with your sign-in token).
  • A blog at /blog/, written the way the Write cards write: first post, "Why the GTM check has no AI score".
  • Pricing: free while in preview, one brand per invited account; prices are not decided yet and nothing is charged. A pricing page says the same.
  • The check's search-based rows now run on GTMRun's own quota, so busy minutes no longer turn them into "not measured".

A free check of your site against Google's own guidance, rebuilt from the ground up

  • Paste your address and get, in about twenty seconds, every finding with a guide: email authentication (SPF, DKIM, DMARC, MX), HTTPS and HSTS, robots.txt and sitemap, soft 404s, title and description, indexability and snippet eligibility, pricing page, call to action, analytics and email capture.
  • Three rows now come from a real browser rendering of your page (canonical URL, one H1, image alt text), one from an image audit (do your homepage images fit a phone's budget?) and one from the public link graph (does anyone link to you yet?). Each is labelled with its source.
  • Every row is pass, fix or not measured. There is no AI score any more: a number that moved between two checks of the same site a minute apart could not tell you whether a fix worked, so it is gone.
  • A "Re-check after a fix" button re-runs the checks in seconds, with no third-party calls.
  • Results have a link you can share; they stay for ninety days. No sign-up, no e-mail, no cookies.
  • Agents can run the same check: an MCP endpoint at /mcp with gtm_check, gtm_recheck and gtm_check_result, and a GitHub Action that fails a deploy when the rows you choose fail.
  • The fix guides moved onto this site under /fix/, with a page on how scoring works and one on what we deliberately do not check, quoting Google.