Anchor text ratio is the share of your inbound links that use each type of link text — your brand name, a target keyword verbatim, a keyword inside a longer phrase, a generic phrase like "click here", or the bare URL. This calculator takes the counts from an existing backlink export and turns them into a distribution, then flags the categories that sit outside the ranges you set. It is a diagnostic instrument pointed at a profile you already have. It is not a blueprint for one to build.
That distinction is the whole point of the page, so Arb Digital will state it plainly rather than bury it in a disclaimer. Deliberately engineering the anchor text of inbound links in order to manipulate rankings is a link scheme. Google's spam policies name link buying, excessive link exchanges, and "using automated programs or services to create links to your site" as violations, and they specifically call out links with "optimized anchor text in articles or press releases distributed on other sites". A tool that told you to go and acquire forty branded anchors to dilute your exact-match percentage would be telling you to commit the exact violation it claims to protect you from.
What This Anchor Text Ratio Calculator Does
You enter how many of your inbound links fall into each of five anchor categories. The calculator converts those counts into percentage shares of the total, displays the distribution as a bar breakdown, and compares each share against an editable threshold — a ceiling for the four non-branded categories and a floor for branded anchors. Anything outside its threshold contributes to a single flag score.
The flag score is not a penalty risk score and cannot be one, because Google publishes no anchor text thresholds and no public signal that would let anyone calculate such a thing. What the score measures is much narrower and much more honest: how far this profile sits from the shape you told the tool to expect. A high score means the distribution is unusual enough to be worth explaining. Sometimes the explanation is entirely innocent — a viral news article, a widely syndicated tool, a brand name that happens to be a common phrase. Sometimes the explanation is a link-building campaign somebody ran in 2019 and never disclosed.
How to Use It
- Export your backlink anchors. Use whichever backlink data provider you already pay for, or Search Console's links report for the subset Google shows you. Deduplicate to one link per referring domain if you want the distribution to reflect sources rather than sitewide footer links.
- Sort each anchor into one of the five buckets. Judgement calls are unavoidable — decide once whether a product name that contains a keyword counts as branded or exact-match, then apply that rule consistently across the whole export.
- Enter the counts. Raw counts, not percentages. The tool handles the arithmetic and will renormalise whatever total you give it.
- Adjust the thresholds to your own market. The defaults are Arb Digital's modelled starting points, not published figures. A brand with a generic dictionary-word name will legitimately show a lower branded share than a brand with an invented name.
- Read the deviations as questions, not verdicts. Each flagged category is something to go and investigate in the link data itself.
The Formula / How It's Calculated
Shares are straightforward. For each category, Share % = (category count ÷ total links) × 100. With the default figures, the total is 600 links: 180 branded (30.0%), 95 exact-match (15.8%), 120 partial-match (20.0%), 140 generic (23.3%) and 65 naked URL (10.8%).
Deviation is measured only in the direction that matters. For the four non-branded categories, Deviation = max(0, share − ceiling), because a low exact-match share is not a problem to be corrected. For branded anchors the test is reversed: Deviation = max(0, floor − share), since an unusually small branded share is the thing worth noticing. The flag score is the sum of all deviations, capped at 100.
Running the defaults through that: exact-match is 15.8% against a 10% ceiling, contributing 5.8. Branded is 30.0% against a 40% floor, contributing 10.0. Partial-match at 20.0% is under its 25% ceiling and contributes nothing, as do generic and naked URL. Total flag score 15.8 — a profile leaning harder on keyword anchors and lighter on brand mentions than the model expects, but not dramatically so. The largest single deviation is the branded shortfall, which is the more interesting of the two findings, because it suggests the site is being linked to as a resource for a keyword rather than cited as a company.
Why the Thresholds Are Inputs and Not Constants
Every published "ideal anchor text ratio" you have ever read is somebody's inference. Google has never released target percentages, has never confirmed that a threshold exists, and does not describe its link handling in terms anyone outside the company could turn into a percentage. The numbers that circulate — 50% branded, under 5% exact-match, and so on — come from vendors sampling profiles of sites that happen to rank and then reporting the average. That is a description of a sample, not a rule the algorithm enforces, and treating it as a rule is how a diagnostic becomes a link-building brief.
The defaults in this calculator are Arb Digital's modelled starting points, drawn from the profiles we work with, and we would rather you overwrote them than trusted them. They are almost certainly wrong for your market. A software company whose name is also a common noun will show branded anchors that any classifier reads as generic. A local trades business with fifty citations will show a naked-URL share that would look alarming on a publisher. A tool or template that gets embedded on other sites will accumulate generic and image anchors at a rate no editorial site ever matches.
The useful comparison is never against a published benchmark. It is against your own profile six months ago, using the same classification rules. That comparison tells you which direction the profile is moving, which is a question the data can actually answer.
Where the Boundary With the Anchor Text Generator Sits
Two adjacent tools, two different jobs, and the difference is worth one clear sentence: the anchor text generator writes anchor text for links you control — your own internal links, and the wording you suggest when a page you have written is being placed — while this calculator only reads and classifies the anchors a profile already has and never tells you what to acquire. If you want help phrasing internal links across your own site, that is the generator's job and it is entirely legitimate; internal linking is your own editorial decision on your own pages. If you want to understand a profile of external links that already exists, stay here.
For internal work specifically, the internal link opportunity calculator finds where links are missing on your own site, which is the one place where deliberately choosing descriptive anchor text is straightforwardly good practice rather than a policy problem — Google's own SEO starter guide encourages descriptive link text on your own pages for exactly that reason.
What a Skewed Distribution Actually Tells You
A heavy exact-match share is most often a fossil record. Somebody bought links, ran a guest-post campaign with prescribed anchors, or used a directory network, and the anchors from that period never went away. The distribution is evidence of past activity, and the remedy is a disclosure and cleanup question rather than a dilution question.
A very low branded share on an established company usually means the links are pointing at utility rather than at the business — a tool, a statistic, a template. That is not a defect, but it does mean brand search demand and link demand are decoupled, which you can check directly with the brand search share calculator.
An overwhelming generic share often means the links are structural: sidebar widgets, sponsor logos, directory entries, or image links with empty alt text. Structural links behave differently from editorial ones and it is worth separating them before drawing any conclusion about the profile at all.
And a naked-URL share above roughly a third usually indicates the profile is dominated by citations, forum posts, and automated mentions rather than by editorial links. If the underlying question is whether those links are worth anything, the backlink value calculator is a better lens than the anchor distribution.
Counting Links Versus Counting Domains
The single biggest source of misleading anchor ratios is counting raw links instead of referring domains. One sitewide footer link on a 40,000-page site can appear in an export 40,000 times with identical anchor text. Feed that into a distribution and one relationship dominates the profile, and every conclusion you draw is about that footer rather than your site.
Deduplicating to one link per referring domain fixes it, at the cost of understating genuinely valuable multi-page relationships — a publisher that has linked to you from thirty different articles is a stronger signal than one that linked once, and domain-level counting flattens that difference. There is no universally right answer. The workable rule is to run the analysis both ways when the two totals differ wildly, because the gap between them is itself a finding.
Whichever you choose, record it alongside the numbers, or the distribution cannot be compared to anything later — including your own previous analysis.
Anchors You Cannot Classify
Real exports are messy in ways a five-bucket model does not anticipate. Foreign-language anchors pointing at an English site, anchors that are a person's name rather than a brand or a keyword, image links carrying alt text that reads as a caption, anchors in scripts your classifier cannot read, and anchors that are a single punctuation mark or emoji all turn up regularly.
The temptation is to force them into the generic bucket, which quietly inflates one category and makes the whole distribution less trustworthy. A better habit is to keep an explicit "unclassified" tally outside this tool, and to check its size before believing any of the percentages. If more than about a tenth of the export is unclassifiable, the distribution is describing a subset of your links rather than your profile, and you should say so when you report it.
The same applies to dead and disavowed links. Providers keep historical records long after a link is gone, so a profile that looks keyword-heavy today may be counting anchors from pages deleted two years ago. Cross-check a sample against live pages first.
Arb Digital's SEO team audits existing backlink profiles, separates structural links from editorial ones, documents what a previous agency did, and builds visibility through content and digital PR rather than anchor engineering.
SEO Services Talk to Arb DigitalCommon Mistakes to Avoid
- Treating the distribution as a target. Acquiring links to move a percentage is a link scheme, regardless of how natural the resulting ratio looks. The distribution is a readout, not a goal.
- Counting raw links rather than referring domains, letting a single sitewide footer link define the entire profile.
- Classifying inconsistently — deciding a product name is branded on Monday and exact-match on Thursday makes the trend line meaningless.
- Comparing against a published "ideal ratio" from a vendor study. Those are sample averages, not thresholds, and Google has never published anchor percentages.
- Ignoring dead and disavowed links still sitting in the export, which can make a profile look far more keyword-heavy than it is today.
Related Free Tools From Arb Digital
Pair this with the link velocity calculator to see whether the profile's growth rate is as unusual as its distribution, and the domain authority checker for a source-quality view. If the wider question is site health rather than links alone, the SEO audit score calculator puts off-page authority alongside the other pillars. Browse the full free online tools hub for more.
Frequently Asked Questions
No published one exists. Google has never released target anchor percentages and does not describe its link handling in terms that could produce them. Every "ideal ratio" in circulation is a vendor's average across sampled profiles, which describes a sample rather than a rule. That is why every threshold in this tool is an editable input.
Building links specifically to alter your anchor distribution is a link scheme under Google's spam policies, which name link buying, excessive exchanges and automated link creation as violations. The legitimate responses are to disavow or remove links you know were manipulative, and to earn new coverage on its own merits.
Not automatically, and this tool does not claim to measure penalty risk because no public data would allow that. A high share is a signal to go and look at where those links came from. A profile full of exact-match anchors from paid placements is a different situation from one where a widely used industry term happens to match a page title.
Referring domains give a more representative distribution, because one sitewide footer link can otherwise contribute thousands of identical anchors. Counting links preserves the weight of a publisher that has linked to you many times. Run both when the totals differ sharply, and always record which basis produced the numbers you are reporting.
Either bucket is defensible; consistency is what matters. Most teams treat brand-plus-keyword as partial-match, reserving branded for the name alone. Write your rule down once and apply it to every export, because changing the rule changes the distribution without a single link changing.
It measures the total distance between your profile and the thresholds you entered, nothing more. It is not a penalty probability, a Google metric, or a quality judgement. If you change the thresholds, the score changes while your links stay exactly the same, which is the clearest possible demonstration that it is a model rather than a measurement.
No. Enter external inbound links only. Internal anchor text is your own editorial choice on your own pages and is a normal part of on-site optimisation, so mixing it into an inbound-profile analysis distorts both. Handle internal anchors separately with the internal linking tools instead.
The flag score and thresholds here are an Arb Digital diagnostic model, not metrics published or endorsed by any search engine, and they do not measure penalty risk or predict rankings.
