Title tags, internal structure, and crawl errors are all fully under a site owner’s control. Anchor text in backlinks is the opposite kind of problem: other people write it. The words a linking site wraps around a link to your page are chosen by that site’s editor, not by you, and yet the aggregate of those choices is one of the clearest signals Google uses to judge whether a link profile was earned or engineered.
That inversion of control is the whole reason anchor analysis matters. A profile built by real editorial linking looks one way; a profile assembled through paid placements and link schemes looks another. Google’s Penguin system exists to tell the two apart, which makes anchor analysis one of the first diagnostic steps in any backlink audit worth running.
Why Penguin changed the calculus #
Google launched Penguin in April 2012 specifically to address link and anchor text manipulation. Before Penguin, ranking for a commercial keyword was often a matter of building many links that used that keyword as the anchor, and rankings tended to follow. The tactic worked well enough that it shaped link-building strategy for years.
In September 2016, Google folded Penguin into the core algorithm and made it operate in real time. Two details from Google’s own announcement matter here. First, Penguin now refreshes continuously rather than in periodic updates, so changes are reflected shortly after Google recrawls and reindexes a page. Second, and often overlooked, Penguin 4.0 became more granular: rather than demoting a whole site for bad links, it tends to discount the offending links so they stop counting toward ranking. That granularity is why anchor problems often show up as a specific page losing traction rather than an entire domain collapsing.
The practical result for 2026: anchor strategy is no longer about pushing target keywords into anchors. It is about maintaining a distribution that resembles what would emerge if real people were genuinely linking to your content, while still carrying enough relevance to support ranking. The balance is narrow, and getting it wrong is costly.
The six categories of anchor text #
Every backlink anchor falls into one of a handful of categories. The mix across all the anchors pointing at a page is that page’s anchor profile.
- Branded anchors use the brand name or domain: “Apple,” “Nike,” “yoursite.com.” These are the most common anchors in genuinely organic profiles, because people writing about a brand naturally reference it by name. They signal entity recognition even though they carry little keyword relevance for non-branded queries.
- Naked URL anchors are the bare URL shown as the link text, such as “https://yoursite.com/page-name.” They appear when authors paste a link without adding descriptive text, which is common in casual references, social posts, and copy-paste citations.
- Generic anchors are non-descriptive phrases: “click here,” “read more,” “this article,” “learn more.” They carry little keyword relevance and reflect ordinary linking behavior, especially on resource pages.
- Partial-match anchors include variations of a target keyword without the exact phrase. For a page targeting “project management software,” partial-match anchors might read “project management tools” or “software for project teams.” They carry relevance with natural variation.
- Exact-match anchors use the precise target keyword. For that same page, the exact-match anchor is “project management software.” These pass the strongest keyword signal, and they draw the most scrutiny when overused.
- Image anchors apply when a link wraps an image, in which case the image’s alt text functions as the anchor. The category is small for most sites but relevant for image-heavy pages.
A natural profile spreads across these categories in proportions that vary by industry, content type, and how often the brand gets mentioned organically. B2B SaaS profiles differ from e-commerce, which differ from local services. There is no single correct shape.
What the distribution benchmarks are, and what they are not #
SEO tools and industry analyses publish rough “ideal” distributions: branded anchors making up roughly a third to half the profile, exact-match anchors in the low single digits, generics and naked URLs filling out the rest. Be precise about what these numbers are. They are descriptive observations drawn from tool datasets and competitive analysis, not thresholds Google has published or confirmed. Google has never released an official anchor-percentage table, and John Mueller has repeatedly rejected the idea that internal anchor optimization carries a penalty at all. Treat any specific percentage as a general pattern, not a firm rule.
Two things hold even without exact numbers. Exact-match anchors are consistently the smallest share of a genuinely organic profile, because real people write in their own words and rarely reproduce a precise commercial phrase. And branded and generic anchors dominate organic profiles, because that is how people reference sites in casual writing.
The more useful reframing: the benchmarks describe what natural profiles tend to look like, not what a brand should target. A profile that hits the “ideal” numbers by deliberate construction still carries the manipulation signature of any other engineered profile. What produces natural distributions is real editorial linking from sites that reference the content for a reader’s benefit; the distribution emerges as a byproduct and cannot be reverse-engineered without recreating the patterns Penguin looks for.
One naming note, since it recurs in older SEO writing: the “topically related” anchors sometimes labeled “LSI anchors” are a real category with a wrong label. Mueller has stated plainly that LSI is not part of Google’s ranking systems. The category exists; the term for it is a misattribution.
What over-optimization looks like #
Rather than fixating on a percentage, it is more accurate to watch for shapes that do not occur in organic linking. The most consistent ones:
- Exact-match concentration that is high relative to the page’s topic. There is no published cutoff, but a commercial-keyword anchor making up a large share of a page’s profile reads as built rather than earned. What counts as “high” depends on the topic: a page whose subject genuinely is a specific phrase can carry more of that phrase naturally than a page with many natural phrasings.
- Repeated identical anchors across many domains. A page with thirty backlinks where most use the exact same anchor text indicates coordinated placement. Natural linking produces variation; coordinated placement does not.
- Anchors disproportionate to the page topic. A “wedding photographer Boston” page whose anchors are overwhelmingly “wedding photographer” looks built for the keyword rather than earned through editorial coverage.
- Commercial-intent anchors in concentration. Phrases like “buy cheap SEO services” or “best affordable web hosting” are unambiguously commercial, and high concentrations of them point to paid placement or scheme participation.
- Sudden distribution shifts. A profile that was mostly branded for years and swings toward exact-match within a few months signals an active campaign.
- Anchor velocity bursts. A page acquiring dozens of new backlinks with similar anchors in a single week looks unnatural regardless of the distribution.
- Anchors that do not match the destination. If the incoming anchors are about commercial roofing but the page is about residential plumbing, the mismatch points to error or manipulation.
Penguin’s evaluation is contextual, not a fixed formula. The same exact-match ratio can be fine on a page whose topic is genuinely defined by that phrase and problematic on a page with many natural phrasings.
Internal versus external: two different rule sets #
A recurring point of confusion is that internal and external anchor text follow different logic, because the control situation is different.
Internal links are under the site’s full control, and Google understands that. The owner chooses every internal anchor, so descriptive, keyword-relevant internal anchors are appropriate and helpful, and Google has confirmed there is no over-optimization penalty for them. Practical guidance is still to avoid “click here” for important internal links, since generic anchors waste the context internal linking provides, and to vary the anchor across multiple internal links to the same page so Google can read the target’s topic from several angles.
External anchors are the opposite. The linking site chooses its own text, so the result is naturally varied, and the page-level distribution is what Google evaluates. The primary external risk is exact-match concentration on commercial pages. In short: internal anchors should be descriptive, varied, and keyword-aware; external anchors are whatever the linking site chose, with the overall profile monitored.
Working with partial control #
Some acquisition situations give a brand partial influence over the anchor: guest contributions where the author has input into the bio or in-article link, sponsored placements (which require nofollow or sponsored attributes), directory listings with form fields, and editorial relationships where a publisher will accommodate a request. What to do with that influence:
- Skew toward branded and partial-match anchors. The exact-match temptation is the trap; resisting it protects the profile.
- Suggest natural-sounding options. “Project management software” as a partial match works better than “the best project management software for agencies.”
- Vary the anchor across placements. Five guest posts linking to the same page should use five variations, not one anchor five times.
- Match the anchor to the surrounding sentence. An anchor that forces the text to be twisted around it stands out.
- Avoid commercial-keyword anchors entirely on guest contributions. “Best [product] for [use case]” reads as paid placement whether or not payment was involved.
- Accept the editor’s choice when they change it. An editor’s revision is usually correcting an over-optimization the brand proposed.
The pattern holds even with influence: the right move is mostly not pushing for the optimization the algorithm flags. The links that survive scrutiny are the ones where the anchor was never optimized aggressively in the first place.
Auditing at the page level #
Anchor ratios apply per landing page, not per domain, because Penguin evaluates anchor patterns per destination. A homepage can be overwhelmingly branded and look fine while a services page carries a heavy exact-match commercial load and sits at risk. Sitewide averages hide that.
A workable audit: export all backlinks through a tool such as Ahrefs or Semrush, group them by destination URL to see which pages carry the most external links, and calculate the anchor distribution for each significant page. Flag pages where exact-match anchors are high relative to the page’s topic, where a single anchor variant repeats across many different domains, or where the distribution looks substantially unlike an organic profile. The commercial pages targeting specific keywords are usually where the trouble concentrates, because they attract the link-building attention that produces over-optimization.
If a page’s anchor issues correlate with ranking suppression, recovery is a matter of dilution more than deletion. Stop adding exact-match or commercial anchors to the affected page, build new links using branded, generic, or naked-URL anchors to shift the ratio, and adjust internal links toward more varied text. Selective disavow helps only where the over-optimization came from clearly manipulative sources, and it should stay conservative, since broad disavow often removes links that were contributing positive signal. Because Penguin runs in real time but recrawl frequency varies, recalibration takes weeks to months rather than showing up overnight. The frequent mistake is trying to accelerate recovery by aggressively removing links; patient dilution through new clean links tends to outperform aggressive cleanup.
The AI-search overlay #
Anchor strategy in 2026 has gained one more consideration. AI engines such as ChatGPT, Perplexity, and Google’s AI Overviews use semantic retrieval, so the words around a link tend to matter more than the anchor itself. A descriptive sentence containing the link, even with a generic anchor, can be more useful for AI citation than an exact-match anchor in thin context. The two goals reinforce each other: the restraint that protects against Penguin (branded, generic, and naked-URL anchors set in context-rich sentences) also produces the placements AI systems prefer to retrieve. The old tension between “optimized for Google” and “natural for readers” has largely collapsed into one approach that serves both.
FAQ #
Is there an official percentage of exact-match anchors that triggers a penalty?
No. Google has not published an anchor-percentage table, and the specific ratios circulated in SEO tools are industry heuristics from competitive data, not confirmed algorithmic thresholds. Exact-match anchors are reliably the smallest share of an organic profile, so a large concentration on a commercial page is a warning sign, but there is no fixed cutoff.
Does anchor text on internal links carry the same risk as backlink anchors?
No. Google has confirmed there is no over-optimization penalty for internal anchor text. Descriptive, varied internal anchors are appropriate and useful; the concentration risk applies to external backlinks.
Do the ratios apply to my whole site or to individual pages?
Individual pages. Penguin evaluates anchor patterns per destination URL, so a healthy sitewide average can mask a single commercial page with a problematic profile. Audit each significant landing page separately.
Are “LSI anchors” a real category?
The related-term anchors sometimes called that are real, but the label is wrong. Google has stated that LSI is not part of its ranking systems, so the underlying category exists even though the name is a misattribution.
Anchor text in 2026 is a practice of restraint. The tactic that worked in the early 2010s, building many links with target-keyword anchors, is precisely what the algorithm now catches. The profile that holds up is the one that looks like what genuine editorial linking would produce, with enough partial-match variation to carry relevance for the queries that matter. If you have to engineer a profile to hit a benchmark, you have already built the pattern the benchmark was meant to describe. Earn the links through content worth citing, audit the commercial pages quarterly, and let the distribution take care of itself.