seokit

Audit a site for search (Google, Bing) and answer engines (ChatGPT, Perplexity, Claude, Gemini, Copilot) in one pass, because the evidence overlaps: the same raw HTML answers "can Googlebot index this" and "can OAI-SearchBot read this". Every finding quotes what was observed, every check the tools could not see says so instead of failing, and the report ends with a long-term plan traced to its own findings.

Most runs target a local server from inside the site's repository, before a change ships, so a finding lands next to the file that fixes it. A preview or production URL is the secondary path. The environment is read from the URL, never asked.

When this fires

Pick the mode, then read its file and follow it:

• audit, the default. One page closely, plus the site files that govern it (robots.txt, sitemap, llms.txt, host redirects). Up to five URLs in one run, for cross-page checks such as duplicate titles. Mode audit → read modes/audit.md, then follow it.
• crawl. A whole site broadly, from its sitemap and its links, to find template bugs and structural gaps one page cannot show. Fires on "whole site", "crawl", "all pages", or a sitemap URL. Mode crawl → read modes/crawl.md, then follow it.

A re-audit ("did my fixes land") is audit or crawl again; the report's delta section covers it. With no URL ("check SEO before I deploy"), audit finds the running local server first.

Guards