Anchor text: what it is, its types, and how much exact match still looks natural

Anchor text is the part of a link people click on: the underlined word, the phrase, or the button label. Technically it sits inside the <a> tag, between the opening and closing markup, as in <a href="...">anchor text</a>. There is also the title attribute, which shows up as a tooltip on hover, but Google treats it as a much weaker signal than the visible link text itself.

Since the early days of PageRank, anchor text has been one of the main sources Google uses to understand what a linked page is about. If ten independent sites link to a document using the phrase "accounting software for freelancers," Google associates that document with that topic, regardless of what the page itself says. That makes anchor text one of the few ranking factors a page owner cannot fully control on the receiving end, only influence, for example through link building outreach or editorial mentions.

It is worth separating anchor text from the link destination itself: anchor text describes the target page's content from the linking site's point of view, not the target owner's. That outside perspective is exactly what gives it value, and what makes it possible to manipulate.

Google also weighs the words directly before and after a link as a weaker, supporting context signal, an effect often called co-occurrence. That is why a generic anchor like "read more here" still works fine inside a sentence with a clear topic: the surrounding text supplies what the anchor text alone does not show.

Not every link has visible text. When an image links to another page, say a product photo pointing to a category page, that image's alt attribute usually takes over the anchor text's role. If the alt attribute is missing too, Google and screen readers are left with only the surrounding page context, a noticeably weaker signal. For image links, it is worth treating alt text the way you treat anchor text on a text link: descriptive, but not the same keyword stamped on every single page.

Anchor text is also an accessibility issue. Screen readers often read visually impaired users only a page's list of links, without the surrounding body copy. An anchor like "click here" is meaningless in that context, while "pricing for business customers" makes the destination clear immediately. Good anchor text practice is not just an SEO question, it also improves usability for an audience that consumes a page differently than a sighted user with a mouse.

In practice, several anchor text types exist. A branded anchor uses the company or brand name, for example "ZDS" or "ZDS Digital." A naked URL shows the web address itself as the link text, for example "https://zds.es." Generic anchors use phrases with no keyword relevance, such as "click here," "learn more," or "this page." A partial-match anchor embeds part of the target keyword inside a natural sentence, for example "a guide to internal linking." An exact-match anchor, by contrast, uses the target keyword verbatim, for example "anchor text" as the link text pointing to this very article.

A natural link profile, one built without deliberate steering, mixes these types. Journalists, bloggers, and other site owners tend to link using the brand name, the naked URL, or a generic phrase, because they are not thinking about SEO keywords while writing. Exact-match anchors are the exception in an organically grown profile, not the rule. A rough industry rule of thumb puts exact-match anchors at roughly one to five percent of a profile's total anchor text (not officially confirmed by Google, but widely cited across the SEO industry). More important than the exact figure is the ratio: if exact match dominates a profile, it reads as engineered rather than earned.

That natural mix is also why aggressive link building campaigns, which stamp the same keyword anchor on every link, leave a telltale pattern behind. A real audience does not link uniformly, it links differently each time.

The actual distribution also depends on a domain's size and reputation. An established brand, say a large online retailer, automatically picks up plenty of branded anchors because users and press already know and use the company name. A new or local site has not earned that advantage yet, so its profile leans more heavily on generic anchors and naked URLs, since it does not yet have a name recognisable enough for someone to remember. Pace matters too: dozens of links appearing with the same exact-match anchor within a few weeks looks more suspicious to an algorithmic pattern than the same growth spread across two years.

A practical example: a mid-size furniture retailer earns 40 organic backlinks over two years. Twelve name the company, ten show the naked domain, nine use generic phrasing like "order here" or "shop now," six include a partial-match phrase such as "sustainable solid wood furniture from this collection," and only three use the exact target phrase "buy solid wood furniture." That ratio, brand and generic as the largest blocks, exact match as the smallest group, is typical of a profile built from real mentions rather than purchased placements.

Beyond these five core types there is also a semantic variant: anchor text that uses a related word instead of the exact keyword phrase, for example "wood furniture" instead of "buy solid wood furniture." Google has understood such synonyms and topically related terms well enough for some time to associate them with the same destination page without needing a word-for-word match. Anyone actively writing anchor text for link building can use this variant to stay close to the keyword without sliding into exact match. As for length, there is no fixed character count, but short and precise beats long and fully spelled out: two to five words is usually enough to signal the destination page's topic without overloading the surrounding sentence.

Google deliberately targeted anchor text manipulation once it became clear that link building agencies were buying exact-match anchors in bulk to steer rankings directly. The first Penguin update went live on April 24, 2012 (verified), and explicitly targeted unnatural link patterns, anchor text spam included. Sites with a conspicuous share of exact-match anchors sometimes lost rankings overnight.

After several follow-up versions, Penguin was folded into Google's core algorithm on September 23, 2016, and has run in real time ever since (verified). The effect has shifted since then: instead of penalizing an entire domain, Google now tends to neutralize the suspicious links themselves, they simply stop counting toward ranking. That lowers the risk of a full-site penalty, but it does not make anchor text spam risk-free, because a conspicuous pattern repeated across many links still functions as a signal Google factors into its assessment of overall link quality.

In practice, this means anyone running active link building or placing guest posts should never dictate the exact target keyword as anchor text, and instead leave the wording to the linking site or vary it deliberately. When a press mention reads as natural, it usually links with the brand name or a generic phrase, not with the keyword the site wants to rank for.

Google names the problem directly in its own spam policies. Under "link spam," Google lists "optimized anchor text in articles, guest posts, or press releases" as one example, illustrated with repeated keyword anchors such as "wedding rings," "best ring," or "buy flowers" inside the same piece of content (verified, developers.google.com/search/docs/essentials/spam-policies). That exact pattern, the same handful of keywords as link text across multiple placements, is what a natural profile does not produce.

It is also worth separating algorithmic devaluation from a manual action. Penguin has run algorithmically and in real time since 2016, without anyone at Google intervening case by case. In parallel, Google can still apply a manual action against a domain for especially conspicuous, obviously purchased link patterns; that action shows up in Search Console and requires a reconsideration request. For that rarer but more severe case, the disavow tool remains relevant, even though it now carries less weight in Penguin's day-to-day algorithmic scoring than it did before 2016.

For most sites that are not running aggressive link building, the risk is low regardless. The issue becomes concrete mainly when an agency or freelancer buys or places links on a client's behalf and is tempted to stamp the highest-paying keyword onto every single one. That is exactly where deliberate restraint pays off.

A milder version of the same logic applies to internal linking. Google separates internal and external signals in its evaluation, but even so: if a site repeats the same exact-match anchor pointing to the same destination page across twenty subpages, it reads as editorially implausible, even when it never triggers an actual Penguin penalty.

• Review your own backlink profile's anchor text distribution regularly (for example via Google Search Console or a backlink tool) and keep the exact-match share low.
• When actively link building, deliberately vary between branded anchors, generic anchors, naked URLs, and partial matches.
• For internal links, use descriptive but naturally worded anchors that make the destination page's content clear.
• For guest posts, leave the anchor text choice to the publishing site's editorial team where possible, rather than dictating it exactly.
• Keep anchor text readable within the surrounding sentence, it should make sense even without the link.

• Using the same exact-match anchor text across a very large number of backlinks pointing to the same page.
• Repeating the exact same phrasing on every internal link to the same destination page instead of using variants.
• Choosing anchor text that does not match the destination page's actual content (a misleading anchor).
• Presenting paid link placements with imposed keyword anchor text as "natural" link building.
• Avoiding generic anchors like "click here" entirely, which produces an artificially keyword-heavy profile.