Anchor text: a real signal, and a line Google has policed for 14 years

What anchor text is actually for

Anchor text is the clickable words in a link, and its original job has never changed: tell a reader, and a crawler, what sits on the other end before anyone clicks. Google's own SEO Starter Guide puts the user-facing half of that plainly: "the better your anchor text is, the easier it is for users to navigate," as quoted in Search Engine Journal's anchor text ranking-factors review, published 3 October 2021. The crawler-facing half is the part that got complicated. A link's anchor text is also one of the oldest context signals Google has about the destination page, which is exactly why, for a few years in the late 2000s, stuffing every inbound link with the same exact keyword looked like a free lever. It was never a user benefit. It was an attempt to talk directly to the algorithm, and the algorithm was rebuilt to notice.

April 2012: Penguin starts reading anchor text as a tell

Google's webspam team shipped what it internally called the webspam algorithm update on 24 April 2012, later renamed Penguin. Barry Schwartz reported it live at Search Engine Roundtable the next morning, 25 April 2012, quoting Google's statement to Danny Sullivan that "this change is targeted at webspam, not SEO." The update's stated target was manipulative link patterns and content spam generally, not anchor text by name. But the practical effect on anchor text was immediate and has outlasted everything else about that update. Search Engine Journal's October 2021 review of anchor text as a ranking factor states it directly: anchor text "matters less than it did before the 2012 Penguin update," because Penguin was the moment Google started discounting exact-match, over-optimized anchor patterns specifically, not just spammy links in general. That's the hinge. Before Penguin, ten links with the same stuffed keyword anchor were a plan. After it, they were a liability.

July 2013: Google puts the phrase in writing

Google didn't just adjust an algorithm quietly. A year after Penguin, it rewrote its public link schemes guidelines to name the tactic outright. Search Engine Roundtable covered the update on 29 July 2013, reporting a change that had broken days earlier. Google added three new examples to its own list of link schemes: "large-scale article marketing or guest posting campaigns with keyword-rich anchor text links," "advertorials or native advertising where payment is received for articles that include links that pass PageRank," and, the line that matters most here, "links with optimized anchor text in articles or press releases distributed on other sites." That is Google, in its own guidelines, calling out manufactured anchor text as a named violation, not an inference SEOs drew from watching rankings move.

September 2016: Penguin goes real time, the penalty gets narrower, not softer

Penguin 4.0 rolled out on 23 September 2016, folded permanently into Google's core ranking algorithm so it runs continuously instead of in occasional, dreaded refreshes. Pageonepower's rollout coverage lays out what changed mechanically: Penguin stopped demoting entire sites and started devaluing individual spammy links and the anchor text riding on them, page by page, in real time. Gary Illyes confirmed on 28 September 2016 that old Penguin demotions were already starting to lift as a result. Read that carelessly and it sounds like Google went easier on manipulated anchor text. It didn't. The correction got narrower and faster, not weaker. A page built on keyword-stuffed inbound anchors used to risk the whole site going down with it. Now that link's anchor text just gets ignored, silently, on an ongoing basis, which is a smaller blast radius but no less a judgment on the anchor text itself.

What Google's policy says right now, in 2026

Here is the part that should settle the "maybe this is old news" question. Google's current spam policies for Google web search, the live documentation, not an archived version, still lists "links with optimized anchor text in articles, guest posts, or press releases distributed on other sites" as an example of link spam, sitting under its paid-placement and native-advertising examples. A few lines later it names the same pattern in a different venue: "Forum comments with optimized links in the post or signature." Fourteen years after Penguin and thirteen after the 2013 guideline rewrite, Google has had every opportunity to soften or retire this language. It hasn't. The only thing that's changed is the document has gotten more specific about where the tactic shows up, not less concerned about it.

Mueller, on the record: reads the words, ignores the ratios

Two separate Google statements, four years apart, answer the two questions site owners actually ask about anchor text. The first is whether Google looks at the words at all. John Mueller answered that directly on a Google SEO office-hours livestream on 27 November 2020. "What we try to look at is the specific anchor text there," he said, as reported by Barry Schwartz at Search Engine Roundtable on 30 November 2020. He added that keyword-rich anchor text pointed at a site "would be more problematic," and said he'd pass the specific case being discussed to Google's web spam team. In the same session, Mueller addressed length: he said Google doesn't do "anything special" based on how many words sit inside an anchor, and that the real difference is context, a longer anchor tells Google more about the destination, a short one reads more like a bare keyword, neither inherently better, just different information.

The second question is whether the ratio of anchor text to everything else on a page is something to optimize toward. Mueller closed that door in a Twitter exchange on 9 October 2018, responding to a question that listed word count, anchor-text-to-word ratio, backlink count and external link count together as candidate quality signals. Barry Schwartz reported the exchange the same day. Mueller's answer: "None of those numbers are indicative of quality." That rules out a specific bad habit: running your anchor text percentage against some target range as if there's a correct number to hit. There isn't one, by Google's own account. The risk Google has actually named, twice, in writing, is buying or manufacturing exact-match anchors on other people's sites, not landing on the wrong side of a ratio on your own.

Where the evidence says anchor text actually pays off: your own internal links

Everything above is about anchor text you don't fully control, links other sites point at you. The data gets a lot more useful once you look at the links you do control. Cyrus Shepard's Zyppy study of 23 million internal links across 1,800 websites, published 12 July 2022 and updated 23 February 2026, found that pages with at least one exact-match anchor pointing to their target keyword saw at least five times more traffic than pages with no exact match anywhere in their inbound internal anchors. Naked URL anchors, where the visible text is just the raw address, still beat having no real anchor at all, by roughly 50%. Only 6% of the 23 million links in the dataset carried empty anchor text, meaning that specific mistake was already rare by the time the study ran.

That distinction matters more than it looks. Google's link spam policy, the one quoted above, is written entirely around links on other sites, guest posts, press releases, forum signatures and paid placements, not the anchor text on your own internal navigation and body copy. You cannot be penalized by someone else's guidelines for writing a clear, descriptive internal link, and our own guide to whether internal linking still matters covers the full Zyppy breakdown, including the point where the benefit of adding more links reverses past roughly 45 to 50 per page. If the Penguin-era risk lives almost entirely in anchors you don't control, the Zyppy-era reward lives almost entirely in anchors you do. Fix the second before worrying about the first.

One thing does not work as a substitute. Nofollowing your own internal links to redirect authority toward other pages was a real tactic once, and Google closed it on purpose in 2009; our guide to whether PageRank sculpting still works covers why that specific workaround has been dead for over fifteen years. Clear, varied, descriptive anchor text on the links you add is the thing that replaced it, not a clever way to block or reroute the links you already have.

Does any of this carry over to AI search?

Nobody has published a study that isolates anchor text wording and measures it against AI citation rate the way Google's own guidelines, Penguin and Mueller have addressed it for classic rankings. The closest thing we have is indirect, and it comes from backlinks rather than anchor text specifically. Our guide to whether backlinks work on AI search covers a Semrush study that split follow links from nofollow links and found them correlating with AI mentions almost identically, 0.504 against 0.509 on Spearman. A nofollow link passes no ranking credit by design. If follow and nofollow links move the AI-mention number by roughly the same amount, the href attribute itself is doing very little of the work, which weakly suggests the literal wording inside the anchor matters even less than the attribute does. That's an inference carried over from a related finding, not a direct test of anchor text and AI citation, and we're flagging it as exactly that rather than dressing it up as measured.

What to actually do this week

1. Stop treating anchor text ratio as a number to hit. Mueller said directly that word count, anchor text ratio and link count numbers aren't indicative of quality on their own. There's no correct percentage of exact-match anchors to aim for.
2. Never request or pay for specific exact-match anchor text in a guest post, press release, forum signature or sponsored placement. That's the literal, named example in Google's current spam policy, not a grey area.
3. Audit your own inbound anchor text distribution for pattern, not ratio. A profile that's almost entirely branded or naked-URL anchors, with a small, naturally earned minority of exact match, looks nothing like a manufactured pattern. A profile where every link uses the same exact phrase does, regardless of what percentage that works out to.
4. Fix internal link anchors first. It's the one place the data shows a clear payoff and the one place Google's link spam policy has no jurisdiction at all, since it's your own site. Replace "click here" and bare page titles with anchors that describe the destination.

What we could not verify

We could not find a dated, official Google statement from 2025 or 2026 that specifically reaffirms anchor text's current weight as a ranking signal. The evidence here is a chain, Penguin in 2012, the guideline rewrite in 2013, Penguin going real time in 2016, Mueller's comments in 2018 and 2020, and a spam policy that's still live today, rather than one fresh restatement. Google has also never published a number for how much Penguin reduced anchor text's weight; Search Engine Journal's "matters less" is an editorial read of Google's actions over time, not a disclosed percentage. And the AI search section above borrows an inference from a backlinks study rather than citing a study that measured anchor text and AI citation directly, because we could not find one.

Sources

The 2012 Penguin launch and Google's "targeted at webspam, not SEO" quote come from Barry Schwartz, It's Live: Google Over Optimization Algorithm, Search Engine Roundtable, 25 April 2012. Anchor text's reduced weight after Penguin, and the Google Starter Guide quote, come from Search Engine Journal's anchor text ranking-factors review, published 3 October 2021. The 2013 link schemes guideline rewrite and its three new examples come from Search Engine Roundtable's coverage, published 29 July 2013. The Penguin 4.0 rollout and the shift from demoting to devaluing come from Pageonepower's Penguin 4.0 coverage, 23 September 2016. Google's current, live spam policy and the "optimized anchor text" and "optimized links" wording come from Spam policies for Google web search, Google Search Central documentation, accessed 3 November 2026. Mueller's November 2020 comments on reading specific anchor text and anchor length come from Barry Schwartz, Yes, Anchor Text Is A Google Ranking Factor, Search Engine Roundtable, 30 November 2020. Mueller's October 2018 comment that word count, link count and anchor text ratios are not indicative of quality comes from Barry Schwartz, Google: Word Count, Link Count, Anchor Text Ratios Are Not Indicative Of Quality, Search Engine Roundtable, 9 October 2018. The internal-link anchor text findings come from Cyrus Shepard, 23 Million Internal Links, an SEO Case Study, Zyppy, published 12 July 2022 and updated 23 February 2026. The follow-versus-nofollow AI mention correlation comes from our own guide to backlinks and AI search, sourcing Semrush's October 2025 study.

Common questions