RankAssurance
Owner Tests · Relevancy · Updated 6 September 2026

The Anchor-Text Test.

When the Clickable Words Help a Person — and When They Are a Spam Example

The Short Answer

The Anchor-Text Test asks whether a reader would know the destination from the clickable words alone — and whether any paid, optimized anchor is marked.

Anchor text is the visible words on a link. Official guidance is short: make them descriptive, reasonably concise, and true to the destination. There is no published exact-match percentage. A placed article stuffed with keyword follows is the named spam example. This page is not a recipe for ranking any site for any word.

The Question This Article Answers

Is there a safe exact-match anchor percentage? None is published — no five-to-ten-percent rule, no cap. The Anchor-Text Test asks whether a reader would know the destination from the clickable words alone, and whether any paid, optimized anchor is marked.

The clickable words tell a person — and a crawler — what they are about to open. That is useful. It is not a cheat that ranks a page for a phrase the page does not discuss. Folklore still sells a cap (five to ten percent exact match) and a recipe (use “click here” for the rest so the graph looks natural). Official link guidance says the opposite of that second half: avoid “click here,” “read more,” “website.”

Would a reader know the destination from these words alone?
Three jobs for the words on a link: internal helps a reader find your other page; earned is a publisher describing what they cite; paid must be marked sponsored or nofollow if it exists to pass credit.
Figure 1. There is no official exact-match percentage. The job of the link decides the words.
Two panels contrast anchor text written for a person, a list of cheese types pointing to the Cheese Types page, with anchors built for a package: keyword links reading wedding rings, best ring, and buy flowers that all point to the same money URL, buynow.example.com/pay. The official spam example uses optimized anchors in a placed article.
Figure 2. The official spam example is a placed article whose links are the product. The cheese-style label is official good practice: a person can predict the page.
Decision tree: would a reader know the destination from these words alone? Yes: use them, keep them short and true. No: rewrite generic click-here, or mark a stuffed paid follow as sponsored.
Figure 3. Same filter as Would-This-Exist. If the words only make sense as a ranking purchase, they fail.

What the Documents Say

LabelOfficial LineUse
Descriptive, conciseTells people and Google about the destination. Relates to both pages.Default for internal and earned
“Click here,” “read more”Too generic. Rewrite.Do not use as a camouflage mix
A whole paragraph as the linkToo long. Tighten.Do not stuff the sentence
Chained one-word linksThe docs’ own example: do not split their sample phrase “so many times this year” into five separate linksOne thought, one link
Image linkDescriptive alt is the labelNot “no anchor”
Paid follow, optimized wordsNamed spam example in guest posts and press releasesMark sponsored / nofollow or do not buy the credit
Forum signature keyword pileNamed spam exampleDo not do it

The good-practice example in Google’s own link documentation is a cheese shop: the recommended anchor names the list of cheeses you are about to open, so a person can predict the page from the words alone. That is the whole standard. The named spam example is its mirror — a placed article whose keyword-stuffed follow links are the product being sold.

Large sites have used “click here” in their own help articles. That is a writing miss, not a ranking system. Do not copy it as a safe majority mix.

Negative-SEO dumps of junk anchors exist. Google’s December 2022 link spam update announcement says its spam systems work to neutralize the impact of unnatural links. A pile of weird labels is not, by itself, a reason to buy a disavow package. If a manual action is in Search Console, that is the Suppression Type Test.

Three Generic Links, Rewritten

BeforeAfterWhy
“click here”…to see our service area“our service area map”The destination is in the words
“read more”…after a boiler-guide teaser“the full boiler-replacement guide”Predictable before the click
“website”…pointing at a permit office“the Springfield permit office”Names the source being cited

A Small Anchor Audit, Executed

Six links from one page, each read with the table above. This is the whole method — no percentage was computed at any step.

The Clickable WordsWhat It IsThe Call
“our wiring safety checklist”Internal, descriptiveKeep
“click here”Internal, genericRewrite so the words name the page
“Acme Electric”A publisher named the brand in a news citeNormal earned language. Not yours to edit
“emergency electrician dallas”Exact-match follow in a placed article you paid forThe named spam example. Mark sponsored or decline
A logo image, empty altImage link with no labelAdd descriptive alt — that is the anchor
A three-line sentence, all linkedToo long to predictTighten to the phrase that names the destination

Labels on the AI Surface

Answer engines quote labels. A descriptive anchor survives quotation into an AI answer and still names the destination; “click here” carries nothing anywhere. The words that help a reader predict a page are the same words that let an engine describe and route to it — one more reason the label’s job is description, not keyword placement. Aboutness itself is a different test: anchors do not make a page about a query it does not cover — Topic-Fit.

Calculator

Anchor-Text Calculator

Do not do tonight: order a mix of exact-match follows “capped at 10 percent,” or rewrite every internal link to a target keyword.

Official-Path Next Step

  1. Rewrite internal labels so a person can predict the page. That also helps crawl.
  2. When you cite a source, name the source. That is normal outbound writing, not a juice leak.
  3. If someone else mentioned you, you do not control their words. Earned-Mention Test.
  4. If you paid for the follow, mark it. Qualify outbound links.
  5. If the question is whether the linking page itself is worth being on, that is the Link-Quality Test; a row of your own hosts holding the anchors is Sister-Site.
The sentence to keep. The words on a link are for a person first. A purchased exact-match follow is not a label. It is the spam example.

Questions

Is There an Official Exact-Match Percentage?

No. Optimized anchors in paid guest posts and press releases that pass credit are the named problem — not a 5–10 percent recipe.

Can Anchors Rank a Page for a Word That Is Not on the Page?

Do not plan on it. The page still has to be about the query. Mentions are evidence. They are not a substitute for the First-Index relevancy gate.

Do Internal Anchors Matter as Much as Inbound?

They are the half you fully control. The official guidance covers both, and descriptive internal labels also help crawl and help a reader move. Start there — it costs nothing and needs nobody’s permission.

Is a Branded Anchor Wasted?

No. A publisher naming the brand or the article title is normal earned language — it is what honest citation looks like. Wanting keywords there instead is how the named spam example gets ordered.

Should I Vary Anchors to Look Natural?

Managing a ratio is the folklore this page opened with. Earned words are the publisher’s choice; yours only need to be descriptive. A pattern you have to engineer to look natural is not natural.

Do Nofollow Anchors Do Anything?

The rel attribute qualifies the credit; official docs treat those attributes as hints about how to handle the link. The words still do their human job either way — a reader follows a clear label regardless of rel.

Sources