Link Signals

Whether the Linking Page Is Indexed, and Why Your Tool Won't Say

A row in your backlink tool means one thing: that vendor’s crawler fetched a page and found an anchor pointing at your URL. It does not mean Google fetched that page, kept it, or would show it to anyone. Those are separate facts, held by a company that doesn’t expose them for pages you don’t own.

This gap is the quiet source of a lot of over-claiming. A report says “we earned 14 new links this month” when what happened is that one crawler saw 14 anchors, some of which live on pages no search engine retains.

Two indexes, two different questions

Your link tool’s index is a crawl kept under a retention policy — the machinery described in what a link index actually contains. Its job is to record link relationships, so it has every incentive to keep a page’s outbound links even if the page is thin, duplicated, syndicated, or excluded from search.

A search index is selective by design. Pages get discovered and then not kept; kept and then dropped; kept but never surfaced for anything. None of that selection is visible from a link tool, because the link tool isn’t asking that question.

So “in my backlink tool” and “in Google’s index” are not degrees of the same measurement. They are two independent observations, and only one of them is available to you at scale.

What you can actually check, in order of strength

Strong, for your own property only. Search Console’s URL Inspection reports Google’s own index status — but only for properties you have verified. That covers your target pages, never the third-party page hosting the link. This is the same boundary that limits the Search Console links report: Google will tell you about you.

Moderate: fetch the page yourself. Request the referring URL, unauthenticated, and record what comes back. A status code, a final URL after redirects, whether your anchor is in the returned HTML, and any noindex in a meta tag or X-Robots-Tag header. That is four checkable facts about the page’s current state, none of which your vendor’s row includes.

Weak: searching for the page. Querying a distinctive phrase from the page, or a site: restriction, is not an index audit. It’s an ordinary search with ordinary ranking, filtering, and deduplication behaviour, and the result count is an estimate. Absence from a query result is compatible with the page being indexed and simply not competitive for that phrase. Treat a hit as suggestive and a miss as nearly uninformative.

Effectively unavailable. Whether Google indexed the page yesterday, whether it dropped it, whether it treats it as a duplicate of another URL, and what it does with outbound links on pages it declines to keep. As of this writing there is no public interface that answers any of these for a page you don’t control. The cached-copy operator that people used as a proxy has been retired.

The distinction that survives all of this

You can establish, cheaply and defensibly, that the link exists and is fetchable. You cannot establish that it is in a search index. Both statements can go in a report; only the first one can go in unhedged.

That’s a narrower claim than most link reporting makes, and it’s the one that holds up when someone checks.

Where the difference bites in practice

Syndicated and scraped copies. One article republished across a network produces many rows in a link index. A search engine’s handling of near-duplicate pages is its own business, and you have no visibility into which copy, if any, it retains. Counting all of them as separate wins inflates a report on the least verifiable axis available.

Pages behind a directive. A noindex page is fetchable, so it can sit in a link index indefinitely while being excluded from search by the publisher’s explicit instruction. That directive is visible to you in one HTTP request, and it’s not in your export.

Pages that answer differently to different agents. A site can return 200 to a commercial crawler and 403 to others, or serve different markup entirely. Fetching the page yourself is a second observation from a second vantage point, which is exactly why it’s worth doing.

Long-tail directories and profile pages. Frequently fetchable, frequently uninteresting to a search index, and they inflate a referring-domain count that then gets reported as reach. If a profile is dominated by them, that’s a finding — see what a toxic-link score is actually measuring for why a vendor score won’t reliably name it either.

Reporting practice

Add two columns to any link list you’re prepared to defend:

On a large profile, sample rather than check everything: pull twenty or thirty rows, weighted toward the ones you’re about to claim credit for. The failure it catches is not subtle — it’s the difference between “14 new citations” and “14 rows, of which 9 verified live on fetchable pages, 2 returned 404, 3 carry noindex.”

Then write the unknown down explicitly, once, and reuse it: index status of third-party referring pages is not observable and is not claimed here. That sentence belongs in the unknowns section described in building a link report you can defend.

What nobody outside Google knows

Whether any specific third-party page is in the index today. How duplicate clustering resolves across syndicated copies. What happens to link signals on pages that are crawled but not retained, or retained but never served. Whether any of that has changed recently.

The useful conclusion isn’t pessimistic, it’s just precise. A link tool answers “did a crawler see an anchor,” which is a real observation and a good one. It does not answer “is this link part of the web a search engine reads,” and no amount of confidence in the export closes that gap. Fetching the page yourself closes part of it, in about ten seconds per row, and that’s the cheapest upgrade available to a link report — related to, but distinct from, the attribution problem in how to tell whether a link did anything.

metricsdata-quality