Anchor Text: What You Control, What You Request, and What You Leave Alone

Anchor text is the visible, clickable text of a link. Google's documentation defines it in one line, sets one standard for good anchor text, and names four examples to avoid. This is for the founder or content owner who is wiring their own internal links and has started receiving, or asking for, links from other sites: what anchor text actually does, which anchors you control, what you can reasonably request, and where the published guidance ends and the folklore begins. The organizing idea is one sentence. Sort anchors by who writes them rather than by what they look like, and most of the confusion sorts itself out.

The short version

• Anchors on links you did not place are not yours to fix, and the correct action for them is usually nothing at all.
• The useful question is not which type an anchor belongs to. It is who wrote it, because that decides which of Google's rules applies to it.
• Google's link documentation defines anchor text in one sentence and names four things not to write: click here, read more, website, article.
• Four documents were checked for a ratio and none of them gives one: Google's link best practices page, Google's spam policies, and the two competing guides opened for this article. The percentages that circulate disagree with each other, which is what a convention looks like rather than a finding.
• Google's spam policies attach the phrase "optimized anchor text" to paid and placed links. They do not attach it to writing a descriptive anchor on your own page.

What is anchor text?

It is the part of a link a person can see and click. Google's link best practices documentation puts it in one sentence: "Anchor text (also known as link text) is the visible text of a link."

In markup, it is everything between the opening and closing tag:

<a href="/blog/technical-seo-guide">technical SEO guide</a>

Here, technical SEO guide is the anchor text. The href is the destination and nobody reads it. The title attribute, if you add one, is a tooltip and is not the anchor.

Two near-neighbors are worth separating now, because both of them muddy this term. Alt text describes an image; it becomes anchor text only when the image itself is wrapped in a link, which is covered further down. And in plain HTML, an "anchor" once meant a named position inside a page, the thing a #fragment jumps to. Same word, different sense, no relation to any of this.

Does anchor text still affect rankings, and how much?

Yes as a relevance signal, no as a dial you can turn. Google's link documentation states the general position plainly: "Google uses links as a signal when determining the relevancy of pages and to find new pages to crawl."

Anchor text does three separate jobs, and they get conflated constantly.

• It describes the destination to a machine: a short summary of a page, written by someone other than that page's owner. That is unusual and it is why the signal exists at all.
• It sets expectations for a reader: Google's own framing is that good anchor text "provides context for the link, and sets the expectation for your readers." Click-through and satisfaction are downstream of that.
• It is one documented input to the title Google shows: Google's documentation on title links lists "Text within links that point to the page" among the sources it may draw on when generating the headline in results.

Now the limit, because this is where anchor text advice usually loses its footing. Google publishes no weighting for it. Nothing quoted in this article ranks anchor text against page content, links, or anything else, and a guide that places it in an ordering is inferring from something other than Google's documentation. Treat it as one input you happen to control on your own site and mostly do not control anywhere else.

Who actually writes the anchor?

This is the question that decides what you are allowed to do, and it is not the question the type lists answer. Both of the competing guides opened for this article sort anchors by appearance: branded, exact match, naked, generic. That is a vocabulary. It does not tell you what you may do about the particular link in front of you, which is what a reader actually arrives wanting.

In our implementation work the split that matters is a different one: whether the words are ours to write, ours to suggest, or somebody else's entirely.

Where the link lives	Who writes the anchor	What you can do about it	Which rule governs it
An internal link on your own site	You, completely	Write it, vary it, put it inside a sentence that gives it context	Google's link best practices. No ratio specified
A link you pitched for: guest post, digital PR, outreach	You propose, the publisher decides	Suggest wording once. Do not press for the same phrase across placements	Link best practices. The spam policies name optimized anchors where payment is received
A paid or sponsored placement	You, but the placement is bought	Qualify the link with rel="sponsored". An optimized commercial anchor here is the textbook case	Spam policies, explicitly
A wire-distributed press release	You write it, a wire republishes it everywhere	Qualify the links with rel="nofollow" or rel="sponsored"	Spam policies name distributed press releases
An unsolicited editorial link	Whoever wrote the article	Nothing, and that is the correct answer rather than a resigned one	Nothing to comply with. You did not place it
An image link	Nobody writes text. The alt attribute stands in	Write alt text that describes where the link goes	Google's link best practices, image links

Read row one against row three and the reason for the axis becomes obvious. Both involve writing a keyword-rich phrase inside a link. One is Google's own recommendation about descriptive text; the other is named in the spam policies. What separates them is not the words. It is who paid for the placement.

Picking the attribute itself is a related but separate job, and our guide to the rules for nofollow, sponsored and ugc works through those mechanics.

How should you write the anchors on your own site?

Descriptively, and then stop thinking about it. Google's guidance is short: good anchor text is "descriptive, reasonably concise, and relevant to the page that it's on and to the page it links to." The same page names four anchors to avoid: click here, read more, website, article.

One test does most of the work. Cover the rest of the sentence and read the anchor alone. If it still tells you where you would land, it is doing its job. If it could lead anywhere, rewrite it.

Placement matters as much as wording. An anchor inside a sentence inherits the meaning of everything around it, which is why a link buried in a related-posts strip carries less than the same words in a paragraph that explains why you would click.

When not to optimize: if you find yourself forcing the identical exact-match phrase into every internal link to a page, you have stopped writing for readers. Vary the phrasing the way you naturally would in prose. Nothing is lost by it, and repetition at scale is the one internal pattern worth avoiding. Equity flow, hub-and-spoke structure and orphan pages are a separate subject with their own mechanics, and our internal linking guide handles them properly rather than in a paragraph here.

What can you ask for when someone else links to you?

You can ask. You should not expect to decide, and you should not ask twice.

When you pitch a placement, offering the page's real title or a natural phrase that describes it is entirely reasonable. What is not reasonable is treating the anchor as a deliverable. The moment a phrase becomes a requirement rather than a suggestion, you have changed what the link is.

What I've seen in practice is that the damage comes from repetition rather than from the phrase. One exact-match anchor on a genuinely earned placement disappears into the page. Twenty identical ones, across unrelated publications, describe a process rather than a set of independent editorial decisions, and that is the shape the spam policies name.

There is also a practical argument that has nothing to do with policy. An editor who rewrites your suggested anchor into their own words has just given you a link they are invested in, embedded in a sentence they wrote. That is worth more than the phrase you wanted. How links get earned in the first place, and how to evaluate whether an opportunity is worth pursuing, sits in our guide to earning links.

What does Google's documentation actually say?

Less than the advice built on top of it, and the gap is the interesting part. Two pages carry the substance of it.

Link best practices (last updated 2025-12-10 UTC) gives the definition, the four bad examples, and the standard for good anchor text quoted above. It also notes, on the related question of how many links a page should have, that "There's no magical ideal number of links a given page should contain." No percentage, no split, no distribution target appears on the page.

The spam policies (last updated 2026-08-28 UTC) are where the phrase "optimized anchor text" appears, and the setting is narrower than its reputation. Google's spam policies documentation files it inside the entry on advertorials and native advertising: payment changing hands for articles carrying links that pass ranking credit, or, in Google's wording, "links with optimized anchor text" in guest posts and in releases distributed across other sites.

Look at what that is scoped to. Paid articles. Native advertising. Guest posts. Releases pushed out over a wire. Every item on the list is a placement somebody arranged, and the anchor is one attribute of the arrangement rather than the thing being policed on its own.

Then comes the half that is easiest to leave out:

"It's not a violation of our policies to have such links as long as they are qualified with a rel="nofollow" or rel="sponsored" attribute value to the <a> tag."

Quoting the prohibition without the qualification turns a policy about disclosure into a prohibition on paid placement, which it is not. This matters most in wire distribution, where one release becomes hundreds of copies of the same anchor across sites you never chose, and our guide to what press release SEO can and cannot do covers that specific case.

Is there a correct anchor text ratio?

Not in any of the four documents checked for this article. Here is exactly what was checked on 2026-09-09 and what came back.

Google's link best practices page: no percentage, ratio or distribution target. Google's spam policies: none either. The two competing guides opened for this article, Semrush's anchor text guide and Yoast's, both define the anchor types at length and neither gives a split.

Searching for where the circulating numbers come from returned more numbers rather than a source. Across the recommendations those searches surfaced, the ceiling for exact-match anchors ranged from 5% to 20%, and the recommended branded share ranged from roughly a fifth of a profile to roughly two thirds. Sets that disagree with each other by a factor of three are conventions, not findings, and a convention is a poor thing to build a link campaign around.

So what do you look at instead? Two questions, neither of which needs a target:

• Does this look like something people wrote? Real coverage produces your brand name, page titles, bare URLs, and phrases someone chose because it fit their sentence. A profile that reads as a list of commercial keywords reads that way to a person too.
• Which of these did we place? Sort your anchors by whether the link was pitched, paid for, or arrived on its own. Concentration in the first two columns is worth attention. Concentration in the third is just a popular page.

What counts as over-optimized, and what happens then?

Exact-match anchors at scale on links you placed or paid for. That is the pattern the documentation describes, and every word of it is load-bearing.

Writing internal linking guide as the anchor on a link to your internal linking guide is not over-optimization. It is the descriptive anchor Google's documentation asks for. Anchoring forty bought placements on best project management software is a different thing entirely, and the difference is not the phrase.

What actually happens is more mundane than the folklore suggests. Google's documented remedy for links in advertorials, paid articles, guest posts and distributed releases is qualification: add rel="nofollow" or rel="sponsored" and the links are not a violation. There is no published penalty threshold, no percentage that triggers anything, and no documented consequence tied to a ratio.

The reflex worth resisting: a backlink tool reports a high share of exact-match anchors and the next instinct is to reach for the disavow file. Anchor distribution is an input to third-party toxicity scores, not to Google's stated test for whether you should file anything. Our guide to when to use the disavow tool sets out the two conditions Google actually publishes, and a ratio is not one of them.

The named anchor types, and what they are actually for

They are a vocabulary for describing a profile you already have, not a plan for building one. Useful for saying "most of these are branded" in a sentence somebody understands. Not useful as targets.

Type	What it looks like	What it is good for
Branded	Your company or product name	The default output of real coverage. Nobody has to plan for it
Exact match	The precise phrase the page targets	Fine occasionally and on your own site. The type that gets counted when a placement is bought
Partial match	The phrase wrapped in surrounding words	What descriptive writing produces without trying
Naked URL	The address itself, pasted in	Common in forums, footnotes and citations. Says nothing about the page
Generic	click here, read more, website, article	The four Google names as examples of what not to write
Page title	The destination's actual headline	What editors reach for. Descriptive by construction
Image	No text. The alt attribute stands in	Logos, badges and banners. Only as good as the alt text

One caution about the lists themselves. Semrush's guide names nine types and Yoast's names eight, and the two sets are not the same items: Semrush counts compound and image anchors, Yoast counts site name separately from brand. Both were opened on 2026-09-09. A count like that is a consequence of where an author drew the boundary, so a guide that promises "the 9 types" is describing its own taxonomy, not a standard.

What about image links, buttons and empty anchors?

On an image link, the alt attribute does the job. Google's documentation is direct about it: "For images as links, Google uses the alt attribute of the img element as anchor text, so be sure to add descriptive alt text to your images."

That changes how you write alt text on a linked image. On a decorative photo, alt describes the picture. On a logo that links to your homepage or a badge that links to a case study, it is standing in as the anchor, so describe the destination rather than the artwork.

Buttons and icon links are the quiet gap. An <a> wrapped around an icon with no text and no alt attribute contains no anchor text at all. Nothing is broken and nothing is penalized. There is simply no description of the destination in the link, which is worth knowing before you build a navigation entirely out of icons.

How do you audit the anchors already pointing at you?

Search Console, once, and then leave it alone. The Links report has a "Top linking text" section, described in Google's Links report documentation as showing "which link text is used in backlinks to your site. Covers only links from outside your property." It is free, it is Google's own view, and it is the first place to look.

Read the limits before you read the data. Google states that the report "isn't a comprehensive list of every link on your site" and that it "shows a sample," and that tables are capped at 1,000 rows so larger sites see a truncated view. A sample cannot support a percentage. Anyone computing a precise distribution from this report is computing a precise distribution of a sample, which is a different number wearing the same clothes.

What the report is genuinely good for is the shape of the thing. Scan the top twenty entries and ask the two questions from the ratio section. If the answers are boring, you are done. If a commercial phrase you paid to place is sitting at the top, you have learned something specific and actionable, which is more than a score would have told you.

The habit worth building is small: decide the anchor when you place the link, not in an audit six months later. Anchors written in the moment, by someone who knows why the link is there, come out descriptive without any effort. Anchors reverse-engineered from a target distribution tend to read like it. And stop treating the anchors you did not write as a problem to be solved, because they are the evidence that other people found your work worth pointing at.

This week, open the Links report in Search Console and read your top linking text once. Ten minutes, no spreadsheet. If your site needs the wider fix rather than the read, our SEO growth engine covers the technical and on-page work this sits inside.