# Site audit > How the site crawl works, what the health score means, how to find and fix every issue, compare crawls, schedule them and hand fixes to an AI. ## What the site audit does **Site audit** (SEO tools in the left menu) crawls your website the way a search engine does: it follows links from the homepage, reads every page it finds and runs a set of checks — broken links, missing titles and descriptions, duplicate titles, redirects, slow pages, indexing rules, structured data, security headers and more. The result is a **health score**, a list of problems and a prioritised **Action Plan**. The crawl uses Rankfinity's own crawler. It respects your site's robots.txt and stops at the page and size limits shown before you start. Pages per crawl by plan: Free 500, Solo 1,000, Business 2,000, Pro 5,000, Agency 10,000. ## Running a crawl 1. Open **Site audit**. The domain comes from the project. 2. Click **Run site crawl**. The price is on the button: 2 credits per started 100 pages. A crawl on a schedule is included in your plan. 3. A progress bar shows pages crawled. You can leave the screen; the crawl continues. To stop a crawl, click **Stop**. You are charged only for the pages already crawled. A crawl stopped by its size or time limit is marked "Stopped early". To check one page without crawling the whole site, use the **Page audit** tab. ## The tabs **Summary · Action Plan · Checks · Pages · Page audit · Changes · Core Web Vitals · Indexing · Performance** ### Summary - **Health score** — a 0–100 circle with one sentence explaining it, and the change since the previous crawl. - Progress since the last crawl: fixed and new problems. - **Fix these first** — the most important problems in plain language: title, how many pages, and "See all … and how to fix". - **Technical details** (folded) — response codes, redirects, canonical tags, server headers and crawl cards. - **Crawl details** (folded) — crawl mode, requests, robots.txt, llms.txt, size. - **Your lookups** — past crawls with date, pages, health score, and errors / warnings / notices. Click one to open its summary. Errors are critical and high-severity problems, warnings are medium, notices are low. These are counts of **failed checks**, not of pages. ### Action Plan The problems ordered by impact and effort, so you know what to do first. Each item shows how many pages it affects and the effort to fix. ### Checks Every check, passed or failed, with its plain name and an ⓘ that explains it. Basic checks (broken links, titles, speed, mobile) are at the top; technical ones (response codes, redirects, canonical, server headers) further down. ### The issue page Click any problem — from Summary, Action Plan, Checks, Compare, Changes or a link in an email — and its page opens directly: - **What it is** and **Why it matters**. - Pages affected and pages checked. - **How to fix** and **It passes when**. - **Copy fix prompt for your AI** for just this problem. - **Export CSV**. - Every affected page with details, 50 per page (**Show 200 per page**), with a filter. - The history of this check across crawls. - **Back to ** returns where you came from. The browser's Back button works too, and the address can be bookmarked. ### Pages Every crawled page with status code, title, word count, issues and, when Search Console or Google Analytics is connected, clicks and sessions. A dash means the value was not measured, not zero. Click a page for **URL details**. ### Changes and Compare **Changes** lists what changed since the previous crawl: fixed, new and still failing, pages new or gone, status codes changed. **Compare** puts any two crawls of the same site side by side. Both are free. ### Core Web Vitals, Indexing, Performance - **Core Web Vitals** — real-user and lab speed per page, free within limits. See [Core Web Vitals](/docs/core-web-vitals). - **Indexing** — submit pages to search engines and check Google's index status. See [Indexing and IndexNow](/docs/indexing). - **Performance** — a paid lab speed test per URL. ## Copy fix prompt for your AI Click **Copy fix prompt for your AI** on the Summary or an issue page. Choose who will do the fixes: - **ChatGPT / Claude** - **Cursor / code editor** - **WordPress site** Pick the issues to include, check the **Prompt preview**, then **Copy** or **Download .md**. The prompt lists the problems in Action Plan order with the affected pages and what "fixed" means. It is free. ## Scheduling crawls Open **SEO tools › Site audit**. On the **Summary** tab, right under the **Run site crawl** button, is the **Schedule** card: - **Frequency** — Off, Daily, Weekly, Every 2 weeks, Monthly. Every plan offers every option; your plan's limit decides the shortest interval you can save, and a greyed-out option says what the limit is. - **Day** — for example Wednesday, or Any day (not shown for Daily). Monthly with a day means the first such day of the month. - **Time** — a full hour in the project owner's time zone (Account › Profile), or Any time. - **Pages** — how many pages each scheduled crawl reads, up to your plan's limit per crawl (Free 500, Solo 1,000, Business 2,000, Pro 5,000, Agency 10,000). Click **Save**. The card shows the next crawl ("Wed, Sep 30 · 09:00") and the cost: scheduled crawls are included in your plan and cost no credits; a manual crawl (Run site crawl) costs 2 credits per started 100 pages. A scheduled crawl always covers the whole site from the project's domain. A crawl finished today counts as today's crawl; a failed crawl is retried the next hour. The same setting is also in **Settings › Project › Advanced settings › Automation › Crawl the site**. **From Claude or ChatGPT (MCP):** with the Rankfinity connector, ask for example "run the site audit every Wednesday at 9 for 200 pages". The AI reads the schedule with `site_audit_schedule` and changes it with `set_site_audit_schedule` — same options, same plan limits and the same prices as the card. Changing it needs **Allow changes** on the connection (Account › MCP). See [Connect Claude or ChatGPT (MCP)](/docs/mcp). ## The crawl report email After every finished crawl (manual or scheduled), everyone who turned it on gets an email "New crawl for ": pages analysed, site health and its change, errors / warnings / notices, and **What's new** — each check with its count now and the change (red +N, green −N) with a **View** link straight to the issue page. Turn it on or off per project in the project menu **⋮** (next to the project name, or on the project card in All projects) **› Email me › Crawl report** — or ask the MCP connector (`set_site_audit_schedule` with email on). It is off by default and personal: turning it on does not change it for the rest of the team. For all projects at once, use your email preferences. See [Alerts and emails](/docs/alerts-emails). ## Download sheet (XLSX) A shared client report with a site audit section has a **Download sheet** button. It gives a spreadsheet of the crawl with two sheets: **Issues** (one row per address and problem, with severity, area, what was found, how to fix it and the check id) and **Pages** (status, title and the other columns the table sorts by). Use it when a PDF of two thousand pages is too long to read. It is free, like Download PDF. Your AI client can give you the same file for the latest crawl: ask for the site audit as a spreadsheet, and it returns a download link that works for a few minutes (see [Connect Claude or ChatGPT](mcp)). ## Common questions **Why did my health score drop when I changed nothing?** A new check may have been added (for example duplicate titles), or the crawl found pages it did not reach before. The Changes tab shows exactly what is new. **A redirected address is listed as an issue. Is that a problem?** Redirects are normal; we list them for completeness. Chains and redirect loops are the problems to fix. Redirected addresses are not counted as separate pages. **Why is a check "not measured"?** We could not test it — for example the security check could not reach the server. That is not a fail. **Does the crawl see JavaScript-only content?** Pages that show text only after scripts run may look nearly empty. If important pages appear empty, make sure their main content is in the HTML. **Can a client run a crawl?** Clients can view results. Running crawls and changing the schedule is for the agency. ## Fix it in WordPress with the agent If the site runs on WordPress, Site audit can hand the fixes to **Agent to WordPress**: **Fix with Agent to WordPress** on an issue or a page, or **Fix all with Agent to WordPress** for every fixable issue. You see the price first, approve each change, and a re-check of only the changed pages marks the issue **Fixed by the agent · verified** or **Still failing**. The bar at the top says whether WordPress is connected. With the Rankfinity Connector plugin 1.3.0 the agent also fixes broken (404) and mistyped addresses with a redirect, canonical problems and noindex settings. On plans with the AI assistant, every issue — also the ones the agent cannot fix — has **Ask the AI assistant**, which opens the assistant with the issue already written. See [Agent to WordPress](wordpress-agent) for the full loop.