Audit a site's structure, then plan the new one

Drop the site's sitemap.xml, a crawl export or a list of URLs. Your browser draws the current structure and finds what is wrong with it - duplicate spellings, messy URLs, pages buried too deep, sections with no hub, orphans - free, before you sign in, and nothing is uploaded. Then a paid run writes the architecture plan: the page hierarchy, the 301 redirect map, the navigation and the internal links.

Each example has a saved model run, so you can see the whole page for free.

Site name

Site type

The site's URLs

Everything is read in your browser. For a sitemap index, drop its child sitemaps together. A crawl export with an Inlinks column also finds orphaned pages.

Drop a sitemap .xml or .xml.gz (several child sitemaps at once are fine), a crawl .csv or .txt, or a set .json saved from this page, or

Key pages (paths, one per line)

Current navigation (optional)

Top goals for the site (optional)

Who it is for (optional)

Prescan (free, in your browser, no AI)

Current structure

Top-level sections

Section	Pages	Hub page	Sub-folders

Flags

Free: no account, no run.

Anything to ask (optional)

Paste the URLs to price the run.

Written by AI from your URL inventory and the browser's checks. Review before changing URLs on a live site.

findings

planned nodes

redirects

old URLs placed

disagreements

Findings

Ref	Severity	Finding	Fix

Proposed page hierarchy

URL map

Page	URL	Parent	Nav	Priority	Status

301 redirect map

From	To	Why

Export as

Test redirects on staging first. Host (www), protocol (https) and trailing-slash fixes are single server rules and are listed in the next steps, not here.

Header navigation

Footer

Internal linking

Hubs and spokes

Cross-section links

From	To	Anchor text

Orphaned pages

Next steps

Prescan flags, answered

Reconciliation with your inventory

Every flag needed an answer and every high or medium flag a finding; the status could not be looser than the flags left standing; proposed URLs had to obey the plan's own URL rules and mirror their parents; pages not marked new had to exist; every redirect had to start from a URL you gave and land in the plan, with no chains or loops; and every old URL sent had to be kept or redirected.

Raw model reply

The reply did not parse as the structured result, so it is shown as it arrived.

Your recent plans

What this does, and what it does not

The prescan reads the URLs you give it and nothing else - it does not crawl the site. It checks the conventions from the source skill: readable, lowercase, hyphenated paths with no file extensions, query strings, dates or ids; one trailing-slash policy; paths that mirror the hierarchy; no page buried deeper than is usual for the site type; a hub page for every folder; one parent per idea. With a crawl export's Inlinks column it also finds pages nothing links to. From the sitemap protocol it checks the 50,000-URL limit per file and notes that Google ignores <priority> and <changefreq> (sources: sitemaps.org protocol; Google Search Central, "Build and submit a sitemap", checked 2026-09-26). The 3-click rule, 4-7 header items and typical depths are house guidelines, not search-engine rules.

The plan is checked against your inventory: it may propose new pages, but it may not claim a page exists when it is not in your list, redirect a URL you never gave, leave an old URL with nowhere to go, or chain redirects. Nothing on your site changes. Derived from the agent skill @coreyhaines31/site-architecture (coreyhaines31/marketingskills, MIT). The example sites are fictional.