Docs · Updated 2026-09-11
Guide: a pre-launch SEO check
The expensive launch mistakes are small: a leftover noindex, a canonical pointing at staging, a robots.txt rule copied from the old site. Here is how to have your assistant catch them.
Why a checklist, and why an assistant
Most launch-day SEO problems are not strategy problems. They are configuration left over from staging: the CMS's “discourage search engines” switch, a Disallow: / in robots.txt, or canonical tags generated from the staging hostname. They are easy to fix and easy to miss because the page looks perfect in a browser.
Running the same checks by hand on every important page takes long enough that people skip it. An assistant with Site Audit can run them in one request, and — unlike a static report — it can explain each finding in the context of your site and draft the fix.
1. Choose the pages that matter
Don't audit everything. Pick the pages that carry the launch:
- the homepage;
- one example of each template (a product page, a blog post, a category page);
- the pages you will link from announcements and ads.
2. Run the check
Paste a prompt like this, with your own URLs:
Run a pre-launch check on these pages:
https://example.com/
https://example.com/pricing
https://example.com/blog/launch-announcement
For each page call audit_page, check_crawl_rules (Googlebot) and check_social_preview.
Run check_links with scope "internal" on the homepage.
Then give me:
1. Blockers: anything that stops indexing or sharing (noindex, robots.txt blocks, canonical to another host, missing or broken preview image).
2. Broken internal links.
3. Everything else, grouped by fix, with the exact tag or header to change.In clients that support MCP prompts, the built-in launch_check prompt does the same for one URL.
3. Read the results in the right order
Blockers first
- Page is marked noindex — check both the meta tag and the
X-Robots-Tagheader; some hosts add the header to preview deployments. - robots.txt blocks Googlebot — the report gives the matching line number.
- Canonical points to a different URL — fine when intended (for example, filtered product lists), a serious problem when it points at a staging domain.
- Preview image returns an error — every share of the page will have no image.
Then the quality issues
Titles that are too long, missing meta descriptions, several h1s and images without alt text. Worth fixing, but they will not stop a launch.
4. Re-run after fixing
Ask the assistant to re-run only the failing checks and compare. The heuristic score makes “before and after” easy to read, but judge by the findings, not the number. Social platforms cache previews, so use their own refresh tools after fixing Open Graph tags (see fixing link previews).
What this check does not cover
- Rankings, keywords or backlinks — use Search Console or an SEO suite for those.
- Content rendered only by JavaScript — Site Audit reads the served HTML.
- Whole-site crawling — it checks the pages you name.
- Core Web Vitals — the fetch time reported is one request from our server, not real-user performance.