Crawl, check,
fix what matters.

No. A URL the newer crawl did not reach is marked not reached, not deleted, and Check these URLs now asks for each one to see which really answer 404 or 410. When the crawls differ in size, counts are compared on the URLs both reached, and compare_crawls tells an agent the same. To compare like for like, use Re-crawl + find new.

It requests URLs from the website you choose, follows the discovery method and scope you set, records each response, and extracts page and link data for the reports. It does not change pages on the site.

It checks the ground floor: whether each AI crawler may fetch the site under robots.txt, whether llms.txt exists, and whether pages split into passages an assistant can quote on their own. Semantic search and the passage view show which pages answer a question by meaning. It does not measure whether assistants actually cite or mention you.

Spider follows links it discovers in page HTML. Spider + sitemap also reads XML sitemap files and records which discovered pages are listed there. Sitemap-only starts from sitemap URLs; List mode checks only URLs you provide.

In Auto mode, pages with fewer than 100 words in their original HTML are opened in a bundled browser and checked again after JavaScript runs. You can also render all pages or switch rendering off. This is a crawl diagnostic, not a guarantee that every site's bot defences or browser behaviour will match Google.

The app splits page text into passages, turns them into vectors with a local embedding model, then compares those vectors to group related pages and passages. Cluster labels are based on distinctive terms. The model does not write recommendations or page copy.

Two ways. From the crawl, it flags unusually similar same-language pages by meaning and shows the evidence side by side — a content-overlap lead to investigate. With Search Console connected, it lists the queries for which Google actually showed two or more of your pages, each with a real share of the impressions, the page with the most clicks first. The first needs no Google data; the second is what search really did.

Sign in to the app with Google, open Search Console on the rail and press Connect Search Console. Only that same Google account can be connected, and only with read-only access to Search Console. getlooploop.com keeps the access grant, encrypted; the reports travel from Google to your computer and are saved there, so the last one is shown, with its date, offline or signed out. Pressing Disconnect or signing out of the app deletes the grant, not the reports.

Crawl records and saved crawl history are stored in a local SQLite database. Embeddings are calculated on your computer after the model is downloaded. The app still sends ordinary web requests to the site you crawl; optional external-link checks request those URLs too, and a domain record asks public RDAP, WHOIS and DNS services. At start it reads one public version file from getlooploop.com to tell you about updates, and it sends anonymous usage statistics: a random install ID the app makes up, the app version, the Windows or macOS version, the interface language, and when the app opened and closed. Our host adds the country it sees for the connection, never the address. Never a URL, a domain or anything from a crawl. Both can be switched off in Settings. With Search Console connected, the app reads Google's API directly; getlooploop.com keeps only the encrypted access grant.

Without an account a crawl reads up to 5,000 URLs; with a free account, 50,000; paid levels go up to 1,000,000, the app's own ceiling. That is how it works now: we are still researching what people need, so the limits will change. The sitemap can contain more URLs than the crawl limit; the crawl cap limits fetched pages, not the number of sitemap entries read for membership.

While the app is open it checks for a new version every half hour. On Windows it downloads the update in the background, checks its SHA-256 against the published file, and shows a red restart button in the title bar; the update installs when you restart, with crawls and settings kept. On a Mac it tells you about the new version and opens the download page. Checking for updates can be switched off in Settings.