Workshop · back to the section

Why noindex isn't the answer

There's a task that looks trivial at first: have a page that doesn't come up for one particular name. The obvious tool, however, cuts away far more than it should.

The situation is this. I write an account of a finished job. The account is useful: what the problem was, what the solution became, what worked and what didn't. Anyone searching with a similar problem should read it. One thing I don't want: that when someone searches for the client's company name, my page is what comes up. The client's name is theirs, not mine — their own site should rank for it.

The obvious idea is noindex. One line in the head, and the page disappears from the results. Except that it disappears from everywhere.

What noindex actually cuts

noindex is not a filter that skips the page for certain searches. It tells the search engine: don't keep this page in your index. And what isn't in the index, nothing can cite.

What noindex affects, and what data-nosnippet affects Noindex removes indexing and with it the result, the snippet and the AI citation; data-nosnippet only forbids the snippet from being shown. noindex data-nosnippet result in the listing snippet under the result citation in an AI answer ranking for the subject
  • result in the listing noindex
  • snippet under the result noindex · data-nosnippet
  • citation in an AI answer noindex
  • ranking for the subject noindex

The upper branch is noindex: it severs all four boxes. The lower one is data-nosnippet: exactly one, the snippet.

The fourth box is the one that matters. To someone searching a professional question — how to connect booking with payment, what running a site costs — the account is a good answer. noindex destroys that possibility as well, for the sake of a problem affecting a fraction of searches.

And the third box grows more important every year. Google's answer surfaces are built on the same index the ordinary results use. What leaves the index cannot appear there as a source either — precisely on the surface where the centre of gravity of search is shifting.

What data-nosnippet does, and what it doesn't

data-nosnippet is widely misunderstood. Many use it as if it hid something. It hides nothing: the text inside is read, processed and weighed by the search engine. It forbids one single thing — that the text appear in the snippet beneath the result.

That isn't much on its own, but it isn't useless either. If a name occurs only there, it won't show in the excerpt displayed in search — the preview of the page won't be about the client. It's just that a name not being shown doesn't stop the page being findable by it.

What does work: the structure of relevance

If no tool solves it, what remains follows from how search works. An engine doesn't decide a page's subject word by word. Weight comes from positions: the title, the headings, link text, the description, structured data, the file name, the URL. If a name appears in none of these, only at a single point in the body text, then the page effectively doesn't compete for that name.

This is not a guarantee, and I won't claim it is one. For a sufficiently specific search — the name plus a word that happens to be on the page — the result can surface. What can be achieved is that the page is practically absent for those searching the name, while being fully present for those searching the subject.

I don't rely on myself to keep this rule: a separate check confirms that client names really do appear in one designated place only, and fails if one turns up anywhere else. That's the subject of the piece on the checking script.

When noindex is right after all

The tool isn't bad. Thank-you pages, internal search listings, filtered views, staging environments — none of these belong in any search, and there noindex does exactly its job.

The difference is simple. There, we don't want visibility. Here we do, just not for one word. And for that you don't switch things off; you put the weight somewhere else.

Questions on this topic

What does data-nosnippet actually do?

It forbids the text inside it from appearing in the search snippet. It does not affect indexing or ranking: the search engine still reads, processes and weighs that text. Anyone who thinks it hides something is mistaken — all it achieves is that the text won't show beneath the result.

Does noindex also remove a page from AI answers?

Google's own answer surfaces are built on the index, so anything not in the index cannot be cited by them. Noindex therefore removes a page not only from the ten blue links but also from the surface where a growing share of answers is produced.

So how can a name be pushed into the background?

By making sure the name is not what the page is about. Search engines infer a page's subject from the title, the headings, link text and structured data. If the name occurs in exactly one place, in plain body text, the page effectively won't compete for that name — while still competing fully for questions about the subject.

Is there a case where noindex is the right call?

There is: thank-you pages, internal search result listings, staging environments, anything that has no business in any search at all. The difference is that there we don't want visibility. If a page is a good answer to some question, noindex cuts that off too.

← Back to the Workshop