Mission Growth

What Is Anchor Text? Types, Studies and What Google Reads

Anchor text is the clickable words of a link. See its types, what Google documents, where the studies disagree and how to audit yours for free.

Published 33 min read
Anchor text as a slate page card with one line holding a glass keycap, an emerald underline beneath it and a lens hovering above
On this page

Most advice about links comes down to three rules: keep exact-match anchors to a set percentage, use a mix of types, and never repeat yourself. None of the three has a causal source behind it. What Google documents is narrower and easier to apply: describe the destination, stay reasonably concise, write naturally and give image links real alt text.

You will find what the words of a link are, how Google reads them, which popular claims have primary support, and a free audit of your own anchors in Search Console.

What is anchor text?

Anchor text is the clickable wording of a link, the words that tell a reader and Google what the linked page is about. Google's link documentation defines it as "the visible text of a link", adding that "This text tells people and Google something about the page you're linking to."

In practice, take the sentence "view our pricing". Those three words are the anchor, and the page address behind them is the destination. Every other question on this topic, from types to ratios to studies, is a question about which words to put there and what difference the choice makes.

The name says which end of the link you're writing. The June 1993 HTML draft by Tim Berners-Lee and Daniel Connolly called the sections of text that form the beginning or end of a hypertext link "anchors", all made with the A tag. An A with HREF was the clickable start; an A with NAME was a spot other links could land on.

Only the starting end ever carried words for the reader to click. Jump links still work that way: in a table of contents, <a href="#types">anchor text types</a> puts the description on the link, and the heading it lands on adds none. So write table of contents entries as real anchors, not "section 3".

Anchor text does two jobs at once. For a reader it is a promise about what the click will deliver. For Google it is one of the few short descriptions of a page written by someone other than the page's own author, which is why people keep arguing about it.

That second job is as old as Google. Brin and Page's 1998 paper explains that Google passed the words of a link on to the page it points to, mostly because anchor text can help provide better quality results, and reports over 259 million anchors indexed from a crawl of 24 million pages.

What anchor text looks like in HTML

In HTML, anchor text is whatever sits between the opening and closing a tags, and the href attribute holds the destination. A complete link looks like this:

html
<a href="https://example.com/pricing">view our pricing</a>

The words "view our pricing" are the anchor text. The value inside href is where the click goes. Browsers usually underline links and color them, but that styling is a CSS default and says nothing about how Google reads the words.

The href is what turns the words into anchor text. The HTML standard says an a element with an href "represents a hyperlink (a hypertext anchor) labeled by its contents", and one without it is only "a placeholder for where a link might otherwise have been placed". Google draws the same line for crawling: it can't reliably extract URLs from a elements that have no href.

That catches menus and buttons built with a click handler instead of an address. The visitor sees the same words in both lines below, but only the second gives Google a destination to attach them to:

html
<a onclick="goTo('/pricing')">view our pricing</a>
<a href="/pricing">view our pricing</a>

Check your templates for the first pattern before you rewrite a single anchor. No wording fixes a link Google can't follow.

An image link has no words between the tags, so Google takes its anchor from the image's alt attribute:

html
<a href="https://example.com/pricing"><img src="pricing.png" alt="View our pricing"></a>

If the alt is empty as well, Google can fall back to a title attribute on the a element. With neither, Search Console lists the link as "(empty)": a link that tells Google nothing about its destination.

Creating the link takes two steps in most editors, whether the site runs on WordPress, Shopify, Wix or hand-written HTML. First select the words you want to link. Then choose insert link and paste the destination. The choice that matters happens before the second step: which words you select.

Selecting a longer or shorter stretch of the sentence changes the anchor. "Our pricing page" and "pricing" link to the same URL and give a reader, and Google, different amounts to go on. The rules for choosing come later in this guide.

Google reads the text between the a tags first, uses the alt attribute when the link is an image, and falls back to the title attribute only when the a element is empty. That order comes straight from Google's documentation, and it rarely gets spelled out.

Flow of three steps showing link text, image alt and the title attribute fallback
How Google fills in an anchor: link text, then alt for an image link, then the title attribute only when the a element is empty

For a link made of words, the words themselves are the anchor. For an image used as a link, Google states that it "uses the alt attribute of the img element as anchor text, so be sure to add descriptive alt text to your images". A linked image's alt therefore does the job the visible words would do.

As a fallback, Google says it can use the title attribute as the anchor if the a element is for some reason empty. Treat that as a safety net. Google documents no such role for a title attribute on a link that already has words, and an empty link with a title is still a poor experience for a reader.

An empty link has no visible words and no image alt. Search Console shows such links as "(empty)" in its link report. Zyppy's 2022 study found that over 6% of internal links in its dataset had no anchor text, and that their existence made no difference statistically to traffic. That is a correlation in one dataset, and it does not make empty links a good habit.

Many sites insert links with JavaScript. Google's guidance is direct: if you are using JavaScript to insert anchor text, use the URL Inspection Tool to make sure it is present in the rendered HTML. The raw source of your page can hold an empty a tag while the browser shows words, and Google only reliably reads what survives rendering.

Types of anchor text

Eight types of anchor text cover almost every link you will meet: exact match, partial match, branded, brand plus keyword, naked URL, generic, image and article title. The table gives each type an example, when it fits and what the evidence says, and it labels no type as good or bad on its own.

TypeExampleWhen it fitsEvidence
Exact matchanchor text (for a page targeting "anchor text")A body link where the sentence already says the keywordAhrefs' 2019 study found a weak correlation of 0.1436 on average between exact-match share and rankings; Zyppy found pages with at least one exact match anchor from internal links had at least five times more traffic than pages without, a correlation
Partial matchdifferent types of anchor textThe default for body links, because it reads naturally and still describes the pageGoogle asks for anchors that are descriptive and natural, and warns against cramming keywords
BrandedMoz, YoastLinks you do not control and anywhere naming the source is the honest labelGoogle's spam clause targets optimized anchors in guest posts and press releases
Brand plus keywordMoz anchor text guideA resource link that names its source and its topicNo study row; judge it by wording, since it is partial match plus a brand
Naked URLhttps://moz.comFootnotes, citations and references where the address itself is the pointZyppy: pages with URL anchors from internal links saw almost 50% more traffic, and they were under 1% of links (correlation)
Genericclick here, read moreRarely; a rewrite almost always existsWCAG 2.4.4 asks for link purpose from the text alone or its programmatic context, and screen-reader users rank links that do not make sense third among problems
Imagea linked logo with alt text "Acme Plumbing"Logos and icon linksGoogle uses the alt attribute; Search Console shows image links as "(empty)"
Article titleWhat Is Anchor Text? A GuideCiting another page by its titleExact or partial match by wording, so it carries that type's evidence

Download CSV (CC BY 4.0)

A naked URL link is a link whose clickable text is the web address itself, such as https://moz.com.

A few notes on the rows. Exact match anchor text is a link whose words equal the target keyword, and the two Zyppy and Ahrefs numbers in that row measure different things, as the studies section explains. Article-title anchors are not a separate ranking category: a title that contains your keyword exactly is exact match, and a title that contains it among other words is partial match.

If you read anchor text types lists elsewhere, expect them to run shorter or longer, depending on whether brand plus keyword, image and article title get their own line. The split above keeps them apart because each answers a different question when you are choosing an anchor.

A hyperlink is the whole element, the a tag with its href destination and its text, while anchor text is only the clickable words inside it. People use the two terms interchangeably in conversation, and that is mostly harmless, but the distinction explains several edge cases.

html
<a href="https://example.com/pricing" title="Plans">view our pricing</a>

The hyperlink in that line is everything shown, including the href and the title attribute. The anchor is "view our pricing". A naked URL anchor is a hyperlink whose visible words happen to be an address. A rich anchor, meaning a descriptive phrase with context built into it, is simply the same element with more words than a bare keyword.

The title="Plans" in that example isn't the anchor either. Google reads "view our pricing" and uses the title attribute only as a fallback when the a element itself is empty, so a title is no place to put a keyword the visible words leave out.

Every anchor text belongs to a hyperlink, but not every hyperlink has clickable words. The HTML standard says a link element with rel="canonical" in the page head "creates a hyperlink" to the preferred URL. That hyperlink has no visible words. It can tell Google which URL you prefer, but it describes nothing, so a canonical never stands in for a descriptive body link to the same page.

Why anchor text matters for SEO

Google treats the words of a link as information about the linked page, and its documentation says so directly: the link text tells people and Google something about the page you are linking to. Anchor text SEO advice builds on that sentence, and the part Google documents is smaller than the advice suggests.

What Google documents is a description of good anchors: descriptive, reasonably concise, and relevant to the page they sit on and the page they link to. That passes context about the relevance of the destination from the linking page to Google, and it is all the documentation claims.

How much weight those words carry in ranking is the contested part. Three studies give three answers because they measure different outcomes: Ahrefs in 2019 found anchor text played a rather insignificant role in ranking, Zyppy in 2022 found more anchor variations correlated with more traffic, and Zyppy in 2024 found a negative correlation after Google updates. The next sections separate them.

Why anchor text matters for readers

Screen-reader users depend on link text more than sighted readers do, because assistive technology can list the links on a page and read them out of context. WCAG 2.4.4, Link Purpose (In Context), is a Level A requirement.

Under it, a reader must be able to tell what each link does from its wording alone, or from that wording plus its programmatically determined context, except where the purpose would be ambiguous to users in general.

The W3C's explanation notes that assistive technology has the ability to provide users with a list of links on the page. In that list "click here" appears again and again with nothing around it, so link purpose in context fails the moment the text is generic and the surrounding sentence is not programmatically attached to it.

The WebAIM screen reader survey gives a sense of how users feel about it. The 2024 edition, with 1,539 valid responses, asked respondents to pick their most, second most and third most problematic items, and links or buttons that do not make sense ranked third. The item's wording includes buttons, and the 2026 edition did not repeat the question, so treat it as that edition's finding.

The 2026 edition, with 1,780 valid responses, adds context on how much the links list is used. Of respondents, 67.8% start a long page by navigating through its headings and 6.7% start with the links list. Dividing 67.8 by 6.7 gives about 10, so there are about ten headings-first users for each links-first user.

Grouped bars comparing headings-first and links-first shares in the 2024 and 2026 WebAIM surveys
Share of screen-reader users who start a long page with headings or with the links list, 2024 and 2026

Read that ratio as a weighting aid. Clear headings reach far more of these users first, which is why a well-built heading structure (see our guide to header tags seo) comes before polishing every anchor. Descriptive anchors cost nothing and serve the smaller group fully.

That smaller group is steady rather than growing. Links first was 7.1% in the 2021 edition (1,568 responses), 4.8% in 2024 and 6.7% in 2026, and the users who open the list get only your link text.

They also help the least experienced users most. In the 2026 survey, 72.1% of advanced screen-reader users use headings to find information, against 41.4% of beginners, and beginners are 2.7 times as likely to use the Find feature. A beginner who jumps by search or by the links list meets your link words without the page structure around them.

What studies say about exact-match and variety

Three studies point in different directions because they measure different things: Ahrefs' 2019 rank correlation for exact-match anchors, Zyppy's 2022 internal-link traffic study and Zyppy's 2024 post-update study. First, the claim that circulates most widely, from Moz:

Using the same anchor text across multiple links can trigger Google's spam detection algorithms and should be avoided.

Moz gives no source for that sentence on the page. In the same guide it cites the Zyppy 23 million link research as saying a larger number of anchor variations "may lead to increased" Google search traffic, so one guide leans on a primary study for traffic and on nothing for spam.

Timeline from the 2016 Penguin change to Zyppy’s 2024 study with each finding marked
Four data points on anchor text between 2016 and 2024 and which way each one points
StudySampleOutcome measuredFindingCaveat
Ahrefs, 2019Top 20 results for 19,840 keywords, 384,614 pagesRankingsExact-match share correlated weakly (0.1436 on average); anchor text plays a rather insignificant roleCorrelation across a sample, with flaws Ahrefs lists
Zyppy, 202223 million internal links, 1,800 websites, about 520,000 URLsGoogle traffic per URLMore anchor variations highly correlated with more trafficMostly a correlation study
Zyppy, 202450-site case studyTraffic change per site after Google updatesMore anchor variations per page correlated negatively (-0.337 internal, -0.352 external)Very small sample size; causation unknown

Download CSV (CC BY 4.0)

Ahrefs 2019, rank correlation

Ahrefs examined the top 20 results for each of 19,840 keywords, 384,614 pages in all, and compared the share of each anchor type with the page's ranking. For exact match the correlation was weak: 0.1436 on average and 0.1869 at the median, and Ahrefs read both as weak.

The other measures were weaker still. Phrase-match anchors averaged 0.1057 (0.1393 at the median), which Ahrefs called very weak. An exact-match keyword in the words around the link averaged 0.0640, which Ahrefs reads as no notable effect. The occurrence of all query words in that surrounding text averaged 0.0205, almost zero.

Random anchors, the "click here" kind, averaged 0.0161 and 0.0130 at the median: effectively no correlation. So the case against generic anchors rests on clarity and accessibility, which the readers section covers.

Ahrefs' own conclusion, "any potential flaws aside", was that anchors played a rather insignificant role in 2019 rankings. That is a statement about correlation with rank across many keywords, and it leaves open how much any single link matters.

Two limits shape how far those numbers travel. The keywords had 2K-5K monthly searches and 2-4 words each, so head terms sit outside the sample.

A second test balanced 16K pages, 800 per ranking position, cut down from 27,156. At every position 20-25% of links carried at least one query word in the text around them, and the correlation was -0.0701. You have nothing to tune in the surrounding words either.

Zyppy analyzed 23 million internal links across 1,800 websites, about 520,000 individual URLs, and measured Google traffic per URL. Pages whose internal links used many different anchor variations were highly correlated with more Google search traffic, and pages with at least one exact match anchor had at least five times more traffic than pages without.

Two side findings matter for how you read the headline. Pages with URL anchors from internal links saw almost 50% more traffic, yet those links were under 1% of the dataset. And the study states its own limit: this is mostly a correlation study, and correlation is not causation, though Zyppy adds it can be a very strong hint.

The link count had a ceiling of its own. URLs with 0-4 internal links averaged 2 clicks from Google Search, URLs with 40-44 averaged 4 times that, and past about 45-50 internal links the effect reversed.

Variation counts thin out too. Only a tiny share of URLs had 25 or more anchor variations, so the curve above that point rests on few pages.

Zyppy 2024, variety and post-update traffic

The same author's 2024 study asks a different question. It is a 50-site case study that used metrics from Ahrefs and measured traffic change after Google updates for each site. Losing sites were far likelier to have more anchor variations per page than winning sites, with a correlation of -0.337 for internal variations and -0.352 for external ones.

Raw link counts were not significantly correlated, neither internally nor externally. The study says the sample is very small and that causation is unknown: "We don't know why sites with better-optimized anchors would fare worse."

Two columns contrasting the 2022 internal-link study and the 2024 post-update study
Zyppy 2022 and Zyppy 2024 side by side: sample, outcome measured and direction

The two Zyppy studies do not contradict each other. One measures URL-level Google traffic across a large link sample; the other measures site-level traffic change after Google updates across 50 sites. Different outcomes can move in different directions without either being wrong. What none of the three gives you is causation, and none gives you a ratio.

Dot plot of six correlation coefficients from two studies with different outcome measures
Correlation coefficients from Ahrefs (rank) and Zyppy (traffic change); different outcomes, so compare direction only

The practical reading: no study supports engineering a mix of anchor types, and the strongest negative number on the page comes from the smallest sample.

Is there an ideal anchor text ratio?

No primary source publishes an ideal anchor text ratio, and the figure that circulates for exact match is credited to no named source on the page that states it, only a screenshot from an unnamed blog. Here is the sentence from Ahrefs' study page:

Most SEOs now seem to recommend using exact-match anchor text sparingly

The sentence continues with a range of about 1% to 5%. The page presents it as what SEOs recommend, not as a result of its own data, and the same page's study found only weak correlations. If you see an anchor text ratio quoted as a rule, no published ratio stands behind it, only repetition.

What does Google document? Good anchor text is descriptive and relevant, and the advice on keywords is to write as naturally as possible and resist the urge to cram in every keyword related to the page. Google also notes that keyword stuffing is a violation of its spam policies. None of that has a percentage in it, and the correlation studies cannot supply one because correlation is not causation.

A ratio also aims at a small target. Across the Ahrefs blog, only 22% of a post's traffic comes from its main target keyword on average; in the page's worked example it was 1,076 of 6,200 visits, about 17%.

Ahrefs' study of 3 million random queries found that the top 10 pages also rank for 400 to 1,300 other queries each. An exact-match quota tunes your anchors for one of those queries and ignores the rest.

You can still check yourself without a target number. Two checks come from the evidence:

  • Concentration: does one non-branded phrase account for most of the anchors pointing at a page? That is the pattern the spam fields in the next section describe, in the leaked documentation.
  • Empty anchors: are links on your own site carrying no text at all? Zyppy found them in over 6% of links in its dataset, and Google's documented fallback is only a title attribute.

Both checks read straight from a report.

How Google's spam systems look at anchors

Google's leaked anchor-spam documentation records how many spam phrases appear in a document's anchors among unique domains, how many days 80% of those phrases took to be discovered and the spam phrases' share of all the document's anchors. It is leaked documentation, which describes fields in a data model, and it does not confirm that Google uses them today or how.

What the leaked fields measure

The documentation for the IndexingDocjoinerAnchorSpamInfo module says its fields record "signals used in anchor spam classification". Three fields matter here, and the descriptions are short enough to quote almost whole.

Stack of three measures: spam phrases across domains, days for 80 percent, spam share of anchors
The three quantities the leaked anchor-spam fields record

The first, phraseCount, records how many spam phrases are found in the anchors among unique domains. The second, phraseDays, records over how many days 80% of these phrases were discovered. The third, phraseFraq, is the spam phrases' fraction of all anchors of the document. No threshold appears in any of the field descriptions.

Read together, the three describe spread across domains, speed of appearance and concentration on the page. That is a different shape from "too many exact-match anchors": the signal in the documentation is many sources producing the same phrases quickly. A site gaining links gradually from different authors would not obviously fit that pattern.

What Google's spam policy says

The public rule is shorter. Google's spam policies list, among link spam, "links with optimized anchor text in articles, guest posts, or press releases distributed on other sites".

SourceWhat it saysStatus
Google spam policiesLinks with optimized anchor text in articles, guest posts or press releases distributed on other sites are link spamPublic policy
Leaked anchor-spam fieldsRecords spam phrases across unique domains, days for 80% of them to appear and their fraction of all anchorsLeaked documentation, live use unconfirmed

Google has sized its link spam updates in public, and the numbers point at manipulation at scale. Penguin's April 2012 launch changed about 3.1% of English queries to a degree a regular user might notice, against about 12% for the first Panda change.

The effect was larger in heavily spammed languages: 5% of Polish queries changed. When Penguin joined the core algorithm in 2016, Google called it one of more than 200 signals it uses to rank.

SpamBrain raised the scale again. In its report on 2022 Google said it detected 50 times more link spam sites than in the previous link spam update, and that more than 99% of visits from Search were spam-free.

Google projected about two weeks for the December 2022 link spam update, but its Search Status Dashboard shows the rollout ran from December 14, 2022 until January 12, 2023. So give an anchor clean-up, or a ranking drop, until the rollout has finished before you judge it.

Since 2016, Penguin has devalued spam by adjusting ranking based on spam signals, rather than affecting ranking of the whole site. For what recovery looks like when a site is hit, see our guide to google penalty recovery. Our off page seo guide covers the rest of the link-spam clauses and what to do about them.

How to write good anchor text

Good anchor text is descriptive, reasonably concise and relevant to both the page it sits on and the page it links to, according to Google. Five rules follow from Google's documentation, and twelve rewrites show them applied to sentences you may recognize from your own site.

Two markup snippets side by side: a link whose anchor reads click here, and the same link rewritten with a descriptive anchor
One generic anchor rewritten as a descriptive one, in markup

Five rules

  1. Describe the destination. Google calls good anchor text descriptive, and a reader should know what the click delivers before making it.
  2. Stay reasonably concise. Google's wording is "reasonably concise", which rules out whole sentences as anchors.
  3. Write for both pages. The text should be relevant to the page it is on and to the page it links to.
  4. Write naturally. Google says to resist the urge to cram every keyword related to the page, and that keyword stuffing is a violation of its spam policies.
  5. Do not chain links. Google says not to chain up links next to each other, because it is harder for readers to distinguish between links and you lose surrounding text for each link.

Twelve rewrites

The sentences below are hypothetical anchor text examples made for this guide, labeled as such. Each pair fixes one failure; the first five rewrites cover generic labels and anchors that run too long.

Generic labels

  • Before: To see plans, <a href="/pricing">click here</a>.
  • After: See the <a href="/pricing">pricing plans</a>.
  • Before: <a href="/guides/crawl-budget">Read more</a>
  • After: <a href="/guides/crawl-budget">How crawl budget works</a>
  • Before: Our checklist is <a href="/checklist">here</a>.
  • After: Use the <a href="/checklist">launch checklist</a>.

Too long

  • Before: <a href="/guides/audit">Our complete step by step guide to auditing every anchor on your website for beginners</a>
  • After: Our <a href="/guides/audit">anchor audit guide</a> walks through the steps.
  • Before: <a href="/guides/schema">If you want to learn how structured data works and why it matters for your pages</a>
  • After: Read how <a href="/guides/schema">structured data</a> works.

Stuffed

  • Before: <a href="/plumber">best emergency plumber cheap emergency plumbing plumber near me</a>
  • After: Call an <a href="/plumber">emergency plumber</a> in your area.
  • Before: Try our <a href="/crm">best CRM software CRM tool CRM for startups</a>.
  • After: Try our <a href="/crm">CRM for startups</a>.

Chained

  • Before: <a href="/a">Pricing</a> <a href="/b">Plans</a> <a href="/c">Features</a> <a href="/d">Demo</a>
  • After: Compare <a href="/a">pricing</a> before you choose between the <a href="/b">plans</a>; the <a href="/c">feature list</a> explains the differences.
  • Before: <a href="/x">Guide</a> | <a href="/y">Guide 2</a> | <a href="/z">Guide 3</a>
  • After: Start with the <a href="/x">setup guide</a>, then read how <a href="/y">reporting</a> works.

Empty or image links

  • Before: <a href="/home"><img src="logo.png"></a>
  • After: <a href="/home"><img src="logo.png" alt="Acme Plumbing home"></a>
  • Before: <a href="/contact"></a>
  • After: <a href="/contact">Contact the team</a>
  • Before: <a href="/shop"><img src="cart.png" alt="image"></a>
  • After: <a href="/shop"><img src="cart.png" alt="Shop all products"></a>

How long is too long

Google gives no word count, only "reasonably concise". A link should name its destination and stop. Once the linked words read like a sentence of their own, the text around them is doing too little.

Put the meaning in the first words, because that is where readers look. In a Nielsen Norman Group test of truncated links, the winning link, "Gift Cards & E-Gift Certificates", let 85% of users predict its destination from its first 11 characters.

The same test found the opposite at the bottom: for 35% of the links, users shown only the first 11 characters had no idea where they led. An anchor that opens with "Introducing" or "Learn more about" spends those characters on nothing.

A workable test is to read only the linked words out loud. If a stranger can tell what the click delivers, the length is fine. That is also how a screen-reader user meets the link in a list.

The link situation decides the anchor: for an internal body link write a partial-match descriptive phrase, and for a placement you do not control accept whatever the editor wrote. A target mix gives you nothing to apply case by case, while the situation tells you who writes the words and what the risk is. The table covers ten situations.

Link situationAnchor type to useWhat to avoidEvidence
Internal body linkPartial match: the target page's mapped keyword or a close natural variantThe identical exact-match phrase on every link; generic labelsGoogle on internal anchors (descriptive, concise), Zyppy 2022 (correlation)
Navigation or sitewide linkShort descriptive label, usually the page's nameChained links; cramming related keywords into a labelGoogle: do not chain links, write naturally
Image linkAlt text that says where the link goesEmpty alt text; alt such as "image"Google uses the alt attribute; Search Console shows "(empty)"
Editorial backlink you earnWhatever the editor wrote; you cannot chooseAsking the editor for exact-match wordingSearch Console link text report shows what editors wrote
Guest postBranded anchorOptimized anchor textGoogle's spam clause on optimized anchors in guest posts
Paid placementBranded or a plain description of the pagePaid links that pass ranking credit; optimized anchor textGoogle's spam clause names paid articles with links that pass ranking credit, and optimized anchors in guest posts and press releases
User-generated linkWhatever the user wrote on your siteTreating it as a signal you setNo study row; no documented rule
Brand mentionThe brand nameForcing a keyword into the nameGoogle: write naturally
Link inside a list of linksA descriptive phrase per item with a sentence around itStacking bare labels side by sideGoogle: do not chain links
Link to a page whose keyword you also mapPartial match around the mapped keywordThe same phrase on every linkZyppy 2022 and 2024 disagree on variety, so no engineered mix

Download CSV (CC BY 4.0)

Ahrefs gives one rule for one situation: guest posts get branded anchors and you should not manage ratios. This table adds nine more link situations, each with a Google or study row and a thing to avoid. For the internal rows, the anchor should come from the page's mapped keyword; our keyword mapping guide explains how to assign it.

Internal links are the one place where you control the anchor completely. Google's documentation says paying more attention to the words used for internal links can help both people and Google make sense of your site more easily and find other pages on your site. Internal link anchor text is therefore the lever you control most fully.

The guide rule: use the target page's mapped keyword or a close natural variant, make it descriptive, and never repeat the identical exact-match phrase sitewide. For the mechanics of how many links to place and where, see our guide to internal linking strategy. A page with no links pointing at it has a different problem entirely, which belongs to orphan pages and not to anchor wording.

Several internal links can share one anchor when they point to the same page. The risk sits in the other case: two different pages sharing the same anchor, which creates competing links with the same words and leaves readers guessing which page the phrase belongs to. That pattern is the one worth auditing, and the audit section shows how.

Internal versus external anchor text

External anchors are written by other people, and the Zyppy 2024 study found anchor variations per page negatively correlated with traffic change both internally (-0.337) and externally (-0.352). The two kinds of anchors differ in what you can do about them. You can rewrite every internal anchor in an afternoon. You can only supply or influence external ones.

That asymmetry is the practical argument for treating internal anchors as a craft. Keep them consistent with the mapped keyword where it reads naturally, vary them where the sentence calls for different words, and do not chase a mix as a goal.

Backlink anchor text is mostly written by someone else, so the controllable part is the page the link points to and the wording of anything you supply. The tactics for earning those links belong to our link building guide; this section covers how to read and shape the anchors that arrive.

Anchors you do not control

An editorial link is one a site chose to give you, and the editor writes the anchor. You do not control it, and trying to dictate it is where the trouble starts: Google's spam policies name links with optimized anchor text in articles, guest posts or press releases distributed on other sites. The Search Console links report shows what editors actually wrote, which tells you how others describe your pages.

Anchors you supply

When you provide the link yourself, as in a guest post or a press release, the documented risk applies to you. Use your brand name or a plain description of the page. The quoted policy language is about optimized anchor text, so the safe default for supplied anchors is the brand name.

Google's documentation on rel values says to use nofollow when you would rather Google not associate your site with, or crawl, the linked page, and it says nothing about anchor text on such links, so no documented answer exists. The values themselves are covered in our external links seo guide.

How to audit your anchor text for free

The Search Console Links report lists the link text used in backlinks to your site, grouped by case-insensitive exact match and sorted by count, which makes a free anchor audit possible without a paid backlink tool. The report is called Top linking text, and Google describes it as showing which link text is used in backlinks to your site.

It covers only links from outside your property, so your own internal anchors never appear. Links with no text, such as image links, appear as "(empty)" in the Link Text column and as empty rows in the exported data. Duplicates are combined, where a duplicate is a case-insensitive exact match: "Hello world" and "HELLO WORLD" are one row, while "Hello world" and "Hello worlds" stay two.

Checklist of four audit steps: export the link text, classify each row, compare with the target, fix what you control, with report limits noted
The four audit steps and what the Search Console report does and does not show

Open Search Console, go to Links, and under external links open Top linking text. Export the table to Google Sheets or a CSV. Put the link text in column A and keep Google's count column next to it.

Export rather than read the screen. The on-screen tables stop at 1,000 rows and may be truncated on larger or heavily linked sites, while the Export button takes up to 100,000 rows.

Classify each row

Add a column that labels each row with one of five classes: branded, naked URL, generic, exact match or other. The formula below is a starting point you can copy into a spreadsheet, using hypothetical values, brand name "acme" and target phrase in cell F1.

Code
=IF(A2="(empty)","empty",IF(REGEXMATCH(LOWER(A2),"^https?://|^www\."),"naked URL",IF(REGEXMATCH(LOWER(A2),"acme"),"branded",IF(REGEXMATCH(LOWER(A2),"^(click here|read more|here|this|link|learn more)$"),"generic",IF(LOWER(A2)=LOWER($F$1),"exact match","other")))))
ClassRule the formula applies
emptyThe text equals "(empty)", usually an image link
naked URLThe text starts with http, https or www
brandedThe text contains your brand name
genericThe whole text is a phrase such as click here or read more
exact matchThe whole text equals your target phrase, ignoring case
otherEverything else, which you read by eye

Here is a hypothetical export for an invented plumbing company, "Acme Plumbing", targeting "emergency plumber". Twenty rows, classified by the formula, with the reason beside each. None of this is real data and it carries no traffic numbers.

Link text (hypothetical)ClassWhy
Acme PlumbingbrandedContains the brand
https://acmeplumbing.examplenaked URLStarts with https
(empty)emptyAn image link with no alt
emergency plumberexact matchEquals the target phrase
click heregenericIn the generic list
emergency plumber in LeedsotherPartial match, read by eye
Acme Plumbing's emergency callout pagebrandedContains the brand
read moregenericIn the generic list
this guideotherVague, but the formula only knows five phrases
www.acmeplumbing.example/blognaked URLStarts with www
how to fix a leaking tapotherAn article title
best emergency plumberotherPartial match
acmebrandedContains the brand
a local plumber we usedotherDescriptive, no target
emergency plumbersotherThe plural differs from the target, so it is not an exact match
Plumbing tips from AcmebrandedContains the brand
heregenericIn the generic list
plumber near meotherPartial match
leaky pipe repair guideotherDescriptive, no target
Learn more about Acme PlumbingbrandedContains the brand, though the opener is generic

Two rows teach the most. "emergency plumbers" reads as an exact match by eye but lands in "other" by formula, and Search Console treats a plural as a different row in the same way. "Learn more about Acme Plumbing" is branded by the formula even though it opens generically, so read the "other" rows and a sample of the rest by hand.

Compare with the page's target

Pick one page and read its rows against the phrase you want it known for. Look for the concentration check from the ratio section: is one non-branded phrase dominating, and does anything look like a pattern of identical anchors? A mix of branded rows, naked URLs and descriptive "other" rows is what editors writing for themselves produce.

Fix what you control

You cannot edit backlink anchors on other sites, but you can fix the rest. Add alt text to your image links, rewrite generic internal anchors, and remove any chained links. For internal anchors, the Links report is blind, so crawl the site with a crawler such as Screaming Frog or the Site Audit tool in your SEO suite, export the anchors, and classify them with the same formula.

Common anchor text mistakes

The most common anchor text mistakes are generic labels, empty anchors, keyword stuffing and chained links. Each ties back to a Google quote or a study row from earlier.

  • Generic labels. "Click here" and "read more" make the purpose of the link unclear. WCAG 2.4.4 asks for link purpose from the text alone or its programmatic context, and survey respondents rank links that do not make sense third among problems.
  • An empty anchor. An image link with no alt text or a link with no text at all leaves Google with only the title fallback. Search Console lists such links as "(empty)", and Zyppy's 2022 data found empty anchors in over 6% of links.
  • Keyword stuffing. Cramming every related keyword into an anchor goes against Google's advice to write naturally.
  • Chained links. Links placed directly beside each other make it harder to distinguish between them and strip the surrounding text, as Google warns.
  • JavaScript anchors nobody checked. Text inserted by script should appear in the rendered HTML, which the URL Inspection Tool confirms.
  • Chasing a ratio. No published ratio exists, and the correlation studies disagree with each other in direction.

The rules repeated on page one have no causal source, and what Google documents is narrower. Choose each anchor for the link situation in front of you. Your next step is to export Top linking text from Search Console for your most important page and classify it with the formula above. Pages with no inbound link have no anchor to classify; orphan pages covers finding them.

Frequently asked questions

What is a good anchor text?

Google says good anchor text is descriptive, reasonably concise, and relevant to the page it is on and the page it links to. The writing section above has the rules and rewrites.

What types of anchor text are there?

Eight by wording: exact match, partial match, branded, brand plus keyword, naked URL, generic, image and article title. The types table lists an example, when to use each and the evidence.

Do all links need anchor text?

Every link should say where it goes. Image links use alt text as their anchor, and Google falls back to the title attribute only when the a element is empty. Search Console shows links with no text as "(empty)".

How many keywords should an anchor have?

Google gives no number. It says to write as naturally as possible and resist the urge to cram every related keyword into the anchor.

What is a good anchor text ratio?

No primary source publishes one. The 1% to 5% figure on Ahrefs' page names no source, only a screenshot captioned as coming from another SEO blog, and the studies measure correlation with different outcomes. Check concentration and empty anchors instead.

Can I use the same anchor text for several internal links?

Yes, when the links go to the same page. Avoid giving two different pages the same anchor, and avoid the identical exact-match phrase sitewide. The internal links section explains why.

Can over-optimized anchor text get a site penalized?

Google's spam policies list links with optimized anchor text in guest posts and press releases distributed on other sites as link spam. Since 2016, Penguin devalues spam signals instead of affecting ranking of the whole site.

How do you write anchor text?

Name the destination in a few words, write the way the sentence would read without the link, and check that the linked words make sense read alone. Twelve before and after rewrites are in the writing section.

Cite this page

Aktaş, F. (2026, November 15). What Is Anchor Text? Types, Studies and What Google Reads. Mission Growth. https://missiongrowth.io/blog/anchor-text

Up nextOrphan Pages in SEO: How to Find Them Beyond Search Console
25 min read
Next step

Put these playbooks to work

Start with a free audit. See where the lift is before you commit.

How it works

  1. 01

    30-minute audit call

    We map your funnel against your goal and pull live data from your channels.

  2. 02

    Lift estimate

    You get a written estimate of where the lift is, with a 30-day plan to capture it.

  3. 03

    You decide

    Run it with us, run it in-house, or shelve it. No commitment from the audit.