Anchor text tells search engines what a page is about. It is also the loudest signal in a backlink profile, which means over-optimising it is the fastest way to make a link programme look manufactured. Here is how to vary it safely.
Anchor text is the clickable words in a link. That is the whole definition, and it takes about four seconds to learn. What takes longer is understanding why the phrase you most want — the commercial keyword you are trying to rank for — is the one you should use least. That inversion catches almost everyone, and the reasoning behind it is worth getting right rather than following as a rule.
What anchor text does
In HTML a link looks like this:
<a href="https://example.com/guide">this guide to composting</a>The href is the destination. “This guide to composting” is the anchor text. Search engines use it as a description of the destination page, supplied by a third party. That third-party element is what gives it weight — you can write whatever you like about your own page, but a hundred other sites describing it the same way is independent corroboration. This is also why anchor text works across languages and even for pages a crawler cannot read well. If enough sites link to a PDF using the phrase “annual emissions report,” search engines can infer what the file contains without parsing it.
The types, and what each one signals
Branded. Your company or site name. Linkexchange, the Linkexchange plugin. Signals a genuine reference. Almost impossible to over-use.
Naked URL. The address itself. https://example.com/guide. Extremely natural — people paste links constantly — and underused by link builders because it feels like a wasted opportunity. It is not. Generic. Click here, this article, read more. Passes little topical signal but occurs constantly in real writing, which is exactly why a profile containing none looks odd.
Partial match. Contains your target phrase inside a longer natural phrase. This guide to composting at home for a page targeting “composting at home.” The workhorse category — topical signal without the pattern risk.
Exact match. The target keyword and nothing else. Composting at home. Strongest topical signal and the highest risk, for reasons below.
Image links. When a link is wrapped around an image, the alt text functions as the anchor. Frequently overlooked in audits, and a common source of unintended exact-match concentration when the same badge or logo is placed across many sites. Anchors are one of five quality variables covered in the do-follow vs nofollow guide.
Why exact-match is the risky one
Consider how anchors form naturally. Someone references your page in an article. They are writing for their readers, so they use whatever phrasing fits their sentence — your brand name, the article title, a descriptive fragment, or a bare URL. They are not thinking about your keyword targets and have no reason to use your exact commercial phrase. Across a hundred organic links, you get wide variety with exact-match appearing occasionally by coincidence. Now consider a manufactured profile. Someone is building links to rank a page for a specific term, so they request that term as the anchor every time. The result is thirty links to one page, twenty-two of which use the identical commercial phrase. That distribution does not occur by accident. It is arguably the single most legible signal that links were arranged rather than earned, which is why anchor uniformity sits near the top of the pattern list in the advanced link exchange strategy guide. Google’s spam policies address over-optimised anchor text directly in the context of link schemes. There is no published ratio, and anyone quoting one as official invented it.
What a defensible distribution looks like
Ratios circulate widely and none are authoritative. As a working shape rather than a rule, a commercial page’s inbound anchors typically skew heavily toward branded, naked URL, and generic anchors, with partial match forming a meaningful minority and exact match a small one. Two things matter more than the specific percentages.
Measure per target URL, not per campaign. This is the mistake experienced people still make. Your anchors distribute across a page’s entire inbound profile, from every source, over its whole life. Managing the ratio inside one campaign while ignoring the accumulated total tells you nothing.
Include your organic links in the calculation. Naturally acquired links skew heavily branded and generic, which gives you headroom you may not realise you have. A page with sixty organic links and eight built ones has a very different distribution from a page with eight of each.
A worked audit
A commercial page with 40 inbound links, targeting “project management software.”
Before: 18 exact match, 6 partial, 9 branded, 4 generic, 3 naked URL. Exact match is 45% of the profile. That is a page whose anchor profile only makes sense if someone was building links to it deliberately. The concentration alone communicates intent.
The instinct is to remove links. That is usually the wrong move — you lose the equity, you may not be able to get them removed, and disavowing preemptively causes more damage than it prevents.
The better fix is dilution. Add 25 new links skewed toward branded, naked URL, and partial match. The profile becomes 18 exact out of 65, or 28%. Same exact-match links, materially different shape, and no equity lost.
Practical target: if exact match on a commercial page is climbing past roughly a quarter of its inbound anchors, stop requesting it and build variety until the ratio comes back. The fix is more varied links, not fewer links.
How to audit your own anchors
Fifteen minutes per target page.
- Ahrefs: Site Explorer, set mode to Exact URL for the page you are checking, then open the Anchors report. Semrush: Backlink Analytics, Anchors tab, with the URL selector set correctly. Moz: Link Explorer, Anchor Text
- Export the list with link counts per anchor
- Categorise into the six types. A spreadsheet column and fifteen minutes of judgement — no tool does this reliably because it depends on knowing your own brand and target terms
- Calculate the share of each category
- Check for single-phrase dominance. One exact phrase appearing on fifteen links is worse than five different exact-match variants appearing three times each
- Repeat for each commercial target page. Sitewide anchor data is close to meaningless, because the distribution that matters is per destination
Controlling anchors in practice
When you can choose, as in an exchange or a placement you are arranging: propose the anchor rather than leaving it open, and propose variety across placements. Requesting the same phrase repeatedly is the failure mode, and it is entirely self-inflicted.
When you cannot choose, as with editorial links: this is fine and it is the point. Naturally chosen anchors are the variety you are trying to manufacture elsewhere — and the how to get backlinks guide notes which tactics give you anchor control and which do not.
Do not demand exact match in exchange criteria. Beyond the profile risk, it narrows your partner pool sharply — experienced publishers know what a page full of exact-match outbound anchors does to their own site and will decline.
Internally, the rules are different. There is no over-optimisation risk from your own internal links the way there is externally, so descriptive internal anchors are simply good practice. That asymmetry surprises people.
The over-correction that also looks wrong
Worth naming, because the advice above pushes people into it. Someone reads that exact match is risky, panics, and swings to using nothing but branded anchors on every built link. Forty links, all reading with the company name, all pointing at one commercial page. That is also a uniform distribution. It is a safer uniformity, but a page with forty inbound links and one anchor phrase is a page whose links were arranged, regardless of which phrase was chosen. The signal is the concentration, not the keyword. The same applies to perfectly engineered ratios. A profile that hits an exact target split across five anchor categories, month after month, is describing a spreadsheet rather than a set of publishers making independent choices. Real distributions are lumpy — some months skew branded, some produce two partial matches and nothing else, and occasionally someone links with a phrase you would never have picked. The goal is variety that looks unplanned, which means accepting some untidiness rather than optimising toward a target ratio. If your anchor report looks like it was designed, it was.
Edge cases
Your brand name is your keyword. If the company is called Composting At Home, branded and exact-match anchors are identical and the distribution question largely dissolves. This is a genuine advantage of exact-match brand names.
Image and logo links. Alt text serves as the anchor. A badge or widget distributed across many sites can create heavy exact-match concentration invisibly, because most audits only look at text anchors.
Empty anchors. Links with no text and no alt attribute appear in reports as blank. Common with icon links and generally harmless, but they inflate your generic count if you categorise them carelessly.
Redirected links. Anchors pointing at a URL that redirects to your target still count toward the destination’s profile. If you have consolidated pages, audit the old URLs too or you will miss part of the picture.
Translated anchors. On multilingual sites, anchors in other languages targeting the same page are variety rather than duplication, and they are usually a healthy sign.
Where anchors actually get decided
Every rule above depends on one thing: whether you get a say in the anchor at all. Editorial links you do not control, and that is fine — those are the natural variety you are trying to produce elsewhere. Arranged links you do control, which means the anchor concentration on any commercial page is largely a record of decisions you made one placement at a time. That is where it slips. Nobody sets out to put the same phrase on twenty-two links. They request it once because it is the term they want to rank for, then again because it worked last time, and the ratio builds without a decision ever being taken. Linkexchange handles the placement side from inside your WordPress dashboard — you post what you want, qualifying publishers take it, settlement runs on credits, referral traffic is attributed per link, and placements are monitored so a removed link surfaces rather than sitting in your last export. What it does not do is choose your anchors. That stays a judgement, and it is worth making it against the whole target page rather than one campaign at a time. The internal linking tools are included free, and they are the place the asymmetry above matters. Internal anchors carry no over-optimisation risk, so descriptive is simply better — and the tools surface which of your pages should be linking to each other in the first place, along with orphan pages, weak pages, and authority flow.
The practical summary
Anchor text is a real ranking input and the temptation is to use your best keyword every time. Resist it, because the concentration communicates more than the keyword does. Favour branded, naked URL, and partial match. Let exact match occur occasionally rather than requesting it. Measure per target page across the whole inbound profile. And when a ratio drifts, dilute rather than delete.
