Backlink health checks

The checklist

A health check takes one source page and one target URL and answers six yes-or-no questions. A link that passes all six is labelled healthy. The order matters: later checks only make sense once earlier ones pass, and a monitor typically stops at the first condition that rules the link out.

The label set follows the outcome scheme described on the outcome codes page. Different tools name these states differently, but the underlying questions are the same.

1. The link is present

The monitor parses the HTML of the source page and looks for an anchor whose href, once resolved against the page URL, equals the target. Matching should tolerate trivial differences such as a trailing slash or letter case in the host, but not a different path. A link that now points to a different page on your site, or to a competitor, fails this check just as a deleted link does. Causes are catalogued under why a backlink goes missing.

Links injected by JavaScript after page load are a known edge case. A check that reads only the server HTML may miss them, so a link that appears in the browser but not in the check deserves a manual look.

2. The rel value is the one agreed

Google's documentation describes three qualifying values for outbound links: sponsored for paid placements, ugc for user-generated content and nofollow for other cases where a site does not want to be associated with the linked page. Values can be combined, for example rel="ugc nofollow". Google states that links carrying these attributes will generally not be followed.

The check therefore compares the observed value with the expected one rather than treating any qualifier as a failure. Placements bought with disclosure should carry sponsored, and that is the healthy state for them. The rel values page explains each value in detail.

3. Googlebot may crawl the source page

Robots.txt controls which URLs a crawler may fetch. If the source site disallows Googlebot from the linking page, Google will not crawl the page content, so the link cannot be read from a new fetch. Google notes that a disallowed URL can still appear in results without a description when other pages link to it, which is why "the page shows up in Google" does not prove the page is crawlable. See robots.txt and backlinks.

4. No noindex rule

A noindex rule can be set in a <meta name="robots"> tag, a <meta name="googlebot"> tag or an X-Robots-Tag HTTP header. Google treats the meta tag and the header as equivalent, and once Googlebot extracts either, it drops the page from search results. The check reads both the HTML head and the response headers, because a header rule is invisible in the page source. More on this in noindex on source pages.

5. The canonical points to the page itself

A rel="canonical" element tells Google which URL the site prefers among duplicates. Google treats it as a strong signal rather than a command. If the source page declares a different URL as canonical, Google may index that other URL instead and consolidate signals there, and the link may not exist on the canonical version. Canonical tags on source pages covers the common patterns, such as paginated archives and syndicated copies.

6. The source page is indexed

The last check asks whether Google has the source URL in its index. A link on a page that Google has never indexed is weak evidence of value, since Google may have no current record of the page. This is the condition a backlink indexer is designed to address, and source page indexation discusses how to test and improve it.

Inconclusive results

Some runs end without an answer: the server times out, a firewall serves a challenge page, or the HTML cannot be parsed. These results describe the check, not the link. A sensible policy is to re-run them before contacting anyone, and to treat repeated anti-bot blocks as a reason to verify by hand.

Running the checklist

Checking six conditions by hand takes a few minutes per link and is error-prone for header rules and canonicals. A backlink monitoring tool runs them on a schedule. The IndexChex monitor, from the publisher of this handbook, records all six conditions per run; other tools vary, so compare their documented checks with this list.