Anchor text: what it tells Google about you

Anchor text is the visible wording of a link: the part a reader clicks. It is the shortest description anyone will ever write about your page, and both humans and search engines treat it as a summary of what waits on the other side. A page linked forty times by "click here" has forty votes and no explanation.

The five types of anchor, and what each one signals

The column that matters is not the row-by-row risk: it is the distribution. Each of these types is perfectly normal in isolation. Their proportion is what tells whether your profile assembled itself or was manufactured.

Two ways to get it wrong

Too vague. "Click here", "read more", "this article". The link works, it may even send good traffic, but it describes nothing. It is not dangerous, it is simply waste: somebody cited you and did not say why.

Too optimised. The same exact keyword, repeated across every link you have obtained. This is the expensive mistake, because it is the one that looks deliberate. Nobody writing naturally links to a plumber four hundred times with the phrase "emergency plumber London". The pattern does not look like a site people appreciate; it looks like a campaign.

The useful middle is simply how people actually write: a phrase that makes sense inside the sentence.

What a natural profile looks like

Left alone, a link profile drifts towards a mix nobody planned: the brand name because that is how people refer to you, the bare URL because some sites paste links as they are, descriptive phrases because that is what a writer types without thinking, and a minority of exact keywords from the few people who thought about SEO.

You will find articles quoting precise percentages for each. Treat them with suspicion: no such threshold has ever been published, and a profile that hits the "correct" ratios exactly is itself a pattern. The only solid reference point is comparative: look at three competitors who have been established in your market for years, and notice that their most frequent anchor is almost always their own name.

Taking inventory, once a year

The exercise takes an hour and is done calmly, not in reaction to a drop.

1. Export the list of links you know about, with their exact anchors.
2. Group identical anchors together and count them.
3. Sort from most frequent to least.
4. Look only at the first five rows.

If those five rows are your brand, your domain name and two descriptive phrases, there is nothing to do. If the first row is a phrase you would like to rank for, you already know where it came from, and the rest of this article is about you.

For one specific row you have doubts about, checking a single link is enough: reading a link's anchor and its rel gives you the two things that matter, without opening the page or reading its source.

When you have a say

In outreach you can often propose the wording. Suggest something descriptive that reads well inside the sentence: it is usually accepted without discussion, precisely because it does not sound like a request from an SEO.

A good proposal looks like "their guide to migration redirects", not like "301 redirect migration". The first is a phrase, the second is a search query. The moment you start negotiating for an exact-match phrase, you have changed the nature of the link, and the site owner knows it as well as you do.

On a link you do not control at all, one quiet lever remains: write a good page title. Many writers simply copy the title as their anchor text, so a title written as a sentence rather than as a list of keywords quietly improves every anchor you will ever be given. It is the cheapest intervention on this entire subject, and the only one that works on links you never hear about.

The anchors you control entirely: your own

People regularly forget that half the anchors describing your pages are written by you, inside your own articles. Nobody negotiates them, nobody rewrites them, and search engines treat them just as seriously.

Three rules are enough:

• Describe, do not announce. "our migration check-list" beats "read the article" and costs the same number of characters.
• Vary without forcing it. If twenty articles link to the same page with exactly the same wording, you have just manufactured at home the pattern you are watching for elsewhere.
• Never build anchors from a template. Interpolated anchors produce sentences nobody proofread, of the "learn more about Page 42" variety.

This is also the one place where fixing is instant: an internal anchor is rewritten in ten seconds, without emailing anyone and without negotiating anything. Starting there before worrying about other people's anchors is simply the rational order.

What anchor text does not do

It does not compensate for a bad link. Perfect wording on a page nobody reads is still perfect wording on a page nobody reads. It does not make a page rank when the page does not answer the query. And it is not a setting: you do not "fix" an anchor profile in an afternoon, you let it rebalance over months.

The only genuinely profitable move on this subject is to record the anchor on every check rather than guess at it once a year.

What to do about an over-optimised profile

The answer is almost never to remove links. You dilute: you keep earning ordinary links, with brand and descriptive anchors, until the top row stops being a commercial phrase. It is slow, and it is the only method that breaks nothing.

What you should not do is open a disavow file because a table worried you. Disavowing solves one narrow problem, and an unbalanced anchor profile is not it: you would lose legitimate links in exchange for nothing.

And if some of those over-optimised links were paid for, the question is no longer the anchor. It is whether they are declared, and that is what the rel attribute is for.

Anchors change without telling you

This is the part most people never check. An editor rewrites a paragraph. A redesign moves the link. A translation replaces your wording with a rough equivalent. The link survives, it is still followed, and its anchor, the description that gave it meaning, is now something else.

That loss is real and almost always reversible: a polite note pointing out that a rewrite made the passage less clear works, as long as the article is recent. You have to hear about it first: it is one of the six changes monitoring records, and the only one you will not spot by glancing at the page.

The most brutal case remains your own redesign: if your URLs change, every anchor on the web suddenly points at pages that no longer answer, whatever their wording. The check-list exists, and it is read before the build, not after.

Worth reading next: what makes a backlink good, of which the anchor is only one of six criteria.