# Did your fix reach AI crawlers? Free delivery check · EarnCitations

> Free: after you change a page, check each step AI crawlers need: live, in the HTML without JavaScript, allowed, indexable, in the sitemap, linked, in llms.txt.

Source: https://earncitations.com/delivery.html

> Every page on this site: https://earncitations.com/llms.txt (the whole site as one file: https://earncitations.com/llms-full.txt)

Free tool updated 2026-11-22

## You shipped the fix. Can AI crawlers get it?

Before you read AI answers for a change, check the change arrived. Each step a crawler needs, checked on its own, so a failure has a name.

### The seven steps

- Live. The URL answers 200 (redirects are followed and named), and isn't a sign-in page: a redirect to /login or an auth host, or a bare password form, fails, because crawlers get the form, not your page.

- In the HTML. Your sentence is in the page text without JavaScript, not only in a script blob. When it's missing and your CDN answered from its cache, we say so: purge the cache and check again. If the page advertises a Markdown copy (<link rel="alternate" type="text/markdown"> or a Link header), we read that too: a copy that still has the old text, or doesn't load, is a warning, because agents that read the copy instead of the page get it.

- Allowed. robots.txt, for the final URL's host, lets every AI search crawler and user-triggered fetcher we track fetch the path. Blocked training crawlers are noted but don't fail the step: they don't decide what an assistant's search can fetch and cite.

- Indexable. No noindex (meta or header), and the canonical points to this URL.

- In the sitemap. Listed, under the URL that answers 200, with a lastmod on or after the day you shipped, and not the one date nearly every URL in that file shares.

- Linked. The homepage links to it in its HTML. A warning, not a failure: a hub page or footer link works too.

- In llms.txt. If you have one, it links to the page or its Markdown copy (/pricing.md). For a page in a folder, a folder-level file such as /docs/llms.txt counts, and we say which file lists it. Optional.

The same check is earncitations delivery <url>, the MCP tool check_delivery and GET /api/delivery ([docs](https://earncitations.com/developers.html#api)). To see everything a crawler reads on the page, use [what AI sees](https://earncitations.com/view.html); once it's delivered, check what answers say with the [answer check](https://earncitations.com/claims.html).

### Questions

#### Why check delivery before looking at answers?

Because an answer can only change after a crawler has fetched the new version of your page, and a fix can fail to arrive in ways nobody notices: the new text is added by JavaScript, a staging noindex shipped with it, the canonical points at the old URL, robots.txt blocks a search crawler, or the page is in no sitemap and linked from nowhere. Reading AI answers a week later can't tell "the fix didn't work" from "crawlers never got it". This check can, in a minute. Jam (spreadjam.com) puts it as a rule, verify delivery before interpreting answers, and we agree.

#### What counts as "in the HTML"?

The sentence you give appears in the page text our crawler gets without running JavaScript (case, spacing, curly quotes and dash styles don't matter). If it only appears inside a script or a framework's data blob, that's a warning: whether a crawler reads those varies, and the visible text is the safe place for it. If it's missing and you can see it in a browser, it's added after load, and crawlers that don't run scripts never get it.

#### Why does the sitemap lastmod matter?

It tells crawlers a page changed. [Google](https://developers.google.com/search/docs/crawling-indexing/sitemaps/build-sitemap) says it uses lastmod only when it is consistently and verifiably accurate, meaning the date of the last significant change to the page. [Bing](https://blogs.bing.com/webmaster/february-2023/The-Importance-of-Setting-the-lastmod-Tag-in-Your-Sitemap) says it uses lastmod to schedule crawling, and may disregard dates that are always set to today. So update it when the content changes, and only then. Give the date your change shipped and we check the sitemap says the page changed on or after it. We also warn when one date covers at least 90% of a sitemap file with 20 or more lastmod values and your page carries it: that usually means the sitemap stamps the date it was built, not when each page changed.

#### Does a pass mean AI will cite the page?

No. It means nothing stops crawlers from fetching and indexing the new version. Whether engines then use it is a separate question: re-ask the questions the fix targets after crawlers have revisited (how long that takes varies; assistants that fetch a page live when asked can see it sooner). Ask them in the assistants you care about, paste each answer into the [answer check](https://earncitations.com/claims.html) and save it to the [answer log](https://earncitations.com/track.html), whose [before and after a change](https://earncitations.com/track.html#change) compares the answers on either side of the day the fix went live.

#### Can I run it after every deploy?

Yes: earncitations delivery https://you.com/pricing --expect "Pro costs $29" --since 2026-11-20 exits with code 2 when a step fails, so it can gate a deploy pipeline. For a deploy that changed several pages, list them: earncitations delivery --from pages.txt --since 2026-11-20, one page a line with a tab (or spaces) and the sentence it should now contain (at most 50). Pages are checked one after another, robots.txt, the sitemap, the homepage and llms.txt are fetched once for the run, and a one-line-a-page summary comes last. A check is also a link: this page reruns whatever is in its address.

#### What do you fetch?

The page, robots.txt, your sitemap (eight files at most, children whose names match the page first), the homepage and llms.txt, about thirteen requests in all, with our identified user agent ([EarnCitationsAudit](https://earncitations.com/bot.html)) and without running JavaScript. Nothing is stored.
