Anchor Text Optimization: You Only Control Half of It
The anchor text people obsess over is the anchor text they cannot write. The anchor text they can write is sitting unoptimized on ten thousand internal links.
Anchor text is the visible, clickable text of a link. Anchor text optimization is the practice of choosing that text deliberately instead of by accident.
Almost every guide on the subject is written about backlinks. Ratio tables, percentage targets, how much exact-match is safe, how to "diversify your anchor profile."
There is a problem underneath all of that advice, and it is structural rather than tactical.
You do not write the anchor text on your backlinks. Someone else does. That is what makes them backlinks. The moment you start specifying the exact words a third party must use to link to you, you have stopped doing SEO and started doing the thing Google's policies name.
Meanwhile the anchor text you have absolute, unilateral, free control over — every internal link on your own site — is sitting there saying "click here."
Backlink anchor text ratios are folklore#
Search for anchor text optimization and you will be handed a distribution: roughly this much branded, this much partial-match, this much generic, this much naked URL, and a strict ceiling on exact-match.
Ask where those numbers come from and the trail goes cold. They are not published by Google, and they are not the output of a named study with a stated methodology. They get repeated from post to post until they read as consensus.
Where there is actual measurement, the effect is small.
Ahrefs ran a correlation study on anchor text across 19,840 keywords and 384,614 pages. Exact-match anchors correlated with rankings at a Spearman coefficient of 0.1436. Phrase-match came in at 0.1057. Partial-match at 0.1076. Random anchors at 0.0161.
Spearman runs from -1 to 1. A coefficient of 0.14 is a weak relationship.
Correlation is not causation, and a weak correlation does not mean anchor text is irrelevant. Google's own original paper names anchor text as one of the techniques it uses, and its documentation still treats link text as a signal about the destination page.
What the numbers do rule out is the premise of the ratio table: that there is a precise external anchor distribution you must hit, and that hitting it is worth the effort of trying to engineer it.
Internal anchor text vs backlink anchor text#
Here is the split that matters, and almost no anchor text guide draws it.
Google's spam policies list, under link spam, "links with optimized anchor text in articles, guest posts, or press releases distributed on other sites."
Read that carefully, because it is narrower and more specific than people assume. The problem is not that a link has descriptive anchor text. The problem is placing content on other people's sites in order to carry an anchor you specified. Optimization plus placement is the pattern.
An editorial link where the writer chose the words is fine, whatever words they chose. A guest post where you supplied the paragraph and the anchor is the described behaviour.
Why exact-match anchor text feels dangerous#
The ratio tables did not come from nowhere. They are a fossil, and it helps to know what made them.
On April 24, 2012 Google announced "an important algorithm change targeted at webspam," and said plainly what it would do: "The change will decrease rankings for sites that we believe are violating Google's existing quality guidelines." The industry named it Penguin. Sites with aggressively keyword-stuffed backlink anchors lost rankings across the whole domain, not just on the pages carrying the bad links.
At the time, the fear was rational. If one algorithm run could demote your entire site over your anchor distribution, auditing that distribution was genuinely defensive work. The percentage ceilings people still circulate are a rule of thumb from that era.
Then the mechanism changed. In September 2016 Google folded Penguin into the core algorithm and described the new behaviour directly: "Penguin now devalues spam by adjusting ranking based on spam signals, rather than affecting ranking of the whole site." The same post says Penguin became "more granular" and refreshes in real time.
The entire anchor text panic rests on that difference. The old mechanism was demotion. The current one is devaluation. A link Google distrusts is simply not counted; the penalty does not spread outward to the rest of your site. The downside of a bad anchor is that the link does nothing for you, which is also the downside of never having acquired it.
This is the same reason the disavow file is not the tool people think it is, and why a backlink audit is better run as an asset inventory than a toxicity hunt.
So: yes, be careful about buying anchors, because paying for links is still against the policies and the money is wasted either way. No, you do not need to defend a percentage, because the failure mode that made percentages worth defending was retired a decade ago.
What you can honestly do about backlink anchor text#
You still have one legitimate lever, and it is upstream: what your page is called and how it is described is what people will type when they link to it.
Give a page a clear, specific, memorable name and the anchors that arrive will tend to reflect it. Publish a study called "The 2026 State of X" and people link to it as the 2026 State of X. That is not manipulation. That is naming.
Beyond that, the honest list is short:
- If someone has linked to you with "click here," you can ask them to change it. Most will not bother, and it is rarely worth the relationship capital.
- If someone links with a wrong or misleading anchor, that is worth a polite correction, because it is a factual error on their page.
- You never send a publisher a required anchor string. Suggesting a natural way to reference your resource is fine; issuing the exact keyword phrase you need is the described pattern.
- You do not manage a percentage target you cannot enforce.
If you run outreach at volume, this is the quiet advantage of doing it properly: pitches built around a genuinely useful resource get linked with the resource's own name, which produces the branded, descriptive anchors an "ideal profile" is trying to fake. That is research and relationship work, and it is the part of link building that gets cut first. It is the part GenLink is built to keep running.
Internal link anchor text is the half that is free#
Internal anchor text is the largest untouched optimization surface on most sites, because every single instance of it is yours.
Google's guidance on crawlable links is direct about what it wants: "Good anchor text is descriptive, reasonably concise, and relevant to the page that it's on and to the page it links to." It names the generic options to avoid explicitly: "click here," "read more," "website," "article."
It also gives the test, which takes two seconds per link: "Try reading only the anchor text (out of context) and check if it's specific enough to make sense by itself."
A practical internal anchor pass, in order:
| Pass | What you are looking for | Why it pays |
|---|---|---|
| Generic anchors | Every "click here", "read more", "learn more", "this page" | Free descriptive signal you are currently throwing away |
| Naked URLs in body copy | Bare URLs used as link text mid-sentence | Reads badly and describes nothing |
| Image links | Links wrapped around images | Google uses the image's alt attribute as the anchor text |
| Repeated identical anchors | The same phrase pointing at five different pages | You are telling Google five pages are the same thing |
| Money pages | Your highest-value pages, and every internal link pointing at them | These usually have the fewest descriptive internal anchors |
The last row is the one that produces results fastest. Most sites have a commercial page that everything links to as "our product" and nothing links to by what it actually is.
Running that pass is a crawl, not a project. Any site crawler that exports an inlinks report gives you every internal link on the site with its anchor text attached, and from there the work is three filters:
- Sort anchors by frequency. The generic ones rise to the top immediately, because "read more" is never used once. This also surfaces the repeated-identical-anchor problem for free.
- Filter the anchor column against Google's own bad list — click here, read more, website, article — plus learn more, this page, here, and bare
http. That is your work queue, already prioritised. - Invert it for your money pages. Pull every internal link pointing at your highest-value URLs and read just the anchor column. If you cannot tell what the page sells from that column alone, neither can Google.
Fix them in the source content, not with a redirect or a plugin. The anchor is a word in a sentence, and the sentence usually reads better once the link describes where it goes.
Why the anchor text report looks like a dial#
It is worth being fair to the ratio table, because there is a real observation buried under it.
Sites that manufactured their links do tend to have strange anchor profiles. If you paid for two hundred placements and specified the anchor each time, the distribution will look nothing like that of a site people cite because they want to. An odd-looking anchor profile is a symptom that correlates with a cause, which is genuinely useful diagnostically.
The mistake is inverting it. Reading "manufactured link profiles look unnatural" as "make your profile look natural" gets the causation exactly backwards, and the action it produces is to manufacture a different set of anchors, which is the same behaviour with extra steps.
The profile is downstream of how you got the links. Change the acquisition method and the profile changes on its own. Manage the profile directly and you are, definitionally, manufacturing anchors.
There is also a reporting trap here. Anchor distribution is easy to put in a dashboard, so it ends up in the monthly deck next to numbers that genuinely are levers. Put it in the deck if you like, but label it a description, not a target. Nobody should be given an anchor-ratio OKR.
Anchor text optimization, minus the folklore#
Stop running an anchor text report on your backlink profile as if it were a control panel. It is a mirror, not a dial. It tells you how the web describes you, which is useful information about your positioning, and it is not a set of numbers you can move directly without doing something you should not do.
If you want to change how the web describes you, change what you publish and what you call it. Then go and fix the ten thousand internal links you have been ignoring while you were reading ratio tables.
For the rest of the picture, see what a backlink is, dofollow vs nofollow links, and how to run a backlink audit that inventories assets rather than hunting for toxicity.
Frequently asked questions#
What is anchor text optimization?
Anchor text optimization is deliberately choosing the visible text of a link so it describes the page being linked to. On your own internal links you control it completely. On backlinks from other sites the publisher chooses it, so your influence is indirect at best.
What is the ideal anchor text ratio for backlinks?
There is no published ideal ratio from Google, and the distributions circulated in SEO guides do not trace back to a named study with a stated methodology. Ahrefs' correlation study across 19,840 keywords found exact-match anchors correlated with rankings at only 0.1436 Spearman, which is a weak relationship. Treat your anchor profile as a description of how the web refers to you, not a dial you set.
Is exact-match anchor text dangerous?
The anchor itself is not the issue. Google's spam policies specifically name "links with optimized anchor text in articles, guest posts, or press releases distributed on other sites" — that is, anchors you specified on content you placed elsewhere. An editorial link where the writer happened to use your keyword is not that. The site-wide demotion people fear dates to the 2012 Penguin update; since 2016 Google has said Penguin "devalues spam by adjusting ranking based on spam signals, rather than affecting ranking of the whole site." A distrusted link is discounted, not turned into a penalty.
What makes good anchor text?
Google's documentation says good anchor text is "descriptive, reasonably concise, and relevant to the page that it's on and to the page it links to." It names "click here," "read more," "website" and "article" as generic options to avoid, and gives a test: read the anchor text on its own and check whether it is specific enough to make sense by itself.
Does anchor text matter for internal links?
Yes, and it is the half of anchor text optimization you can act on unilaterally. Every internal link on your site is text you wrote, it is free to change, and on most sites a large share of it is generic. Fixing it is the highest-certainty anchor text work available.
What anchor text does Google use for an image link?
When a link is wrapped around an image, Google uses the alt attribute of the img element as the anchor text. An image link with no alt text is a link with no anchor text.
The half you do not control still has to be earned. GenLink is a backlink agent that runs the research half of that job continuously: it finds sites relevant to your topic, finds the right person there, writes a pitch specific to what they publish, sends it from a warmed mailbox, follows up, and triages the replies. Links earned that way arrive with the anchors people choose for you, which is the only anchor profile worth having.
