Ask fifty witnesses to describe a getaway car and you get fifty versions. Blue. Dark green. "Some kind of SUV."
If all fifty use the exact same sentence, a good detective stops listening to the story. He starts asking who coached them.
Search engines play that detective with links. The words a site uses to link to a page are a description of that page, written by someone else.
Exact match anchor text is the coached witness of SEO. It is a link that repeats, word for word, the keyword its target page wants to rank for.
The fear of "safe limits" comes from there, and for links from other sites it rests on something real. On your own site, every description comes from the same witness. The limit there is of another kind, and it is not a percentage.
Where the fear of exact match anchors comes from
Google reads the anchors other sites use as independent descriptions of a page. When too many repeat a commercial keyword word for word, they look orchestrated. That suspicion is real. The "safe percentage" of exact match anchor text built on top of it has no known source.
In 2012, Google launched Penguin, an update aimed at manipulated links.
Enforcement has grown quieter since. After the December 2022 link spam update, Google mostly ignores unnatural links instead of punishing the site.
The credit those links passed is simply lost. Manual actions remain for flagrant cases.
SEO practice then turned the suspicion into a rule of thumb. Keep exact match below a small share of a page's backlinks, typically "no more than 5 in 100." The numbers tell a less tidy story.
- 65% or more. The share of money-keyword backlinks on sites that dropped after Penguin, in an analysis of its first victims. Even that level did not guarantee a drop.
- 1 to 5%, or 5 to 10%. The "safe" thresholds repeated across guides today. None of those we reviewed points to a study or a Google statement.
- Zero. The number of exact match backlinks on the typical top-20 page, for the query studied, in a 2019 study of 384,614 ranking pages.
The danger exists at the extremes. The safe line in the middle was never measured.
And an ignored backlink passes none of the ranking value that links transfer, often called link juice. It is less a threat than a zero.
A ratio measures a crowd, and your site has one author
A percentage of exact match anchors is built to spot coordination among many independent authors. On your own site, every anchor has the same author. There is no crowd to measure, and Google says there is no internal over-optimization penalty.
Three facts point the same way.
- Google's spam detection counts crowds. Internal documentation leaked in 2024 shows it tracks spam phrases across unique domains and over time.
- Internal and external anchors can be counted apart. The same leaked documentation describes separate counts and scores for on-site and off-site anchors.
- Google says it outright. In a 2019 answer on internal over-optimization, Gary Illyes wrote that "you can abuse your internal links as much as you want," as far as he knew.
The leak shows what Google stores, not how much each signal weighs. But the picture is consistent.
The usual anchor categories, from branded to partial match, grew out of backlink analysis. They fit awkwardly on a site's own links.
Our deduction. A percentage of exact match internal anchors measures nothing Google needs to watch. The real question is no longer site-wide. It is whether each anchor fits its sentence.
What really limits exact match on your own site
No penalty does not mean no limit. Inside a site, exact match anchors hit two ceilings. Repeating one phrase seems to add less and less. Rotating synonyms to avoid it may leave its own fingerprint. Both come from writing anchors for the engine.
Exact match only touches one of the three layers of a working internal link. Position and the surrounding sentence carry the rest.
Identical anchors stop adding up
- The first one seems to matter. In a study of 23 million internal links, pages with at least one exact match internal anchor had at least 5x more Google traffic. It is a correlation, and those pages may simply be better optimized.
- The fortieth one may not count. Google's leaked documentation tracks anchors discarded as redundant beyond a phrase cap per linking domain. Its exact rule, and whether it covers internal links, are unknown.
Our deduction is simple. Repetition hits a ceiling, not a penalty.
What it costs is opportunity. A page ranks for many queries, and each repeated anchor could have described another one.
Unveil SEO was built to find those other descriptions at scale. It reads each paragraph and proposes an anchor from what that paragraph already says about the target.
Engineered variety leaves a fingerprint too
The usual cure for repetition is to vary anchors. The 23-million-link study backed it, finding variety highly correlated with traffic.
Then its author, Cyrus Shepard, looked at 50 sites that won or lost traffic in Google's 2023 and 2024 updates. The sign flipped.
Losing sites had more internal anchor variations per page, with a correlation of -0.337 against traffic change. The raw number of links made no significant difference.
Fifty sites is a small sample, and the result is correlational. The author says he does not know why it happens.
Our hypothesis is that hammering and rotating are one habit in two costumes. Both pick anchors for a search engine, not for a reader.
How to use exact match anchors without counting them
Give each important page at least one exact match anchor, where a sentence naturally calls for it. Let every other anchor come from its own paragraph. Then run each one through the question Google asks in its own link guidance.
Take a page about trail running shoes, and the same link written three ways.
- Hammered. "Shop our trail running shoes," on every page that mentions running.
- Rotated. "Best trail shoes," "off-road footwear," "trail sneakers," swapped in to dodge repetition.
- Written by the sentence. "Rocky descents punish thin soles, which is why trail shoes with a rock plate outlast road models."
Only the third tells the reader something about the destination. It also reads as if nobody was thinking about rankings.
- 1
List the anchors pointing to the page
A crawler such as Screaming Frog, free up to 500 URLs, lists every internal link to a page with its anchor. An internal link audit pays off first on the pages that bring in revenue.
- 2
Keep one exact match where it reads naturally
Pick the sentence where the keyword already sits comfortably, often the one that introduces the topic.
- 3
Rewrite the repeats from their paragraph
Ask what the paragraph says about the target, and make that the anchor. This works for contextual links, placed inside a paragraph rather than a menu, because a menu has no sentence to borrow from.
- 4
Apply Google's reader test
Google's link best practices ask whether the reader needs these keywords to understand the next page. If an anchor would sound odd to someone who never heard of SEO, rewrite it.
For links from other sites, the rule is shorter. Don't choose their anchors.
Google's spam policies target optimized anchors in guest posts, press releases and articles you place on other sites. Those are the links where you hold the pen.
The safe limit for exact match anchor text was never a number. It is an anchor that would still make sense if search engines did not exist.
