Backlink analyse: anchor text myths next to what holds up
Six beliefs that keep showing up in anchor discussions, each followed by what the evidence supports and what a link database can and cannot show you.
Anchor Social research team ·
The six, side by side
Short form. The sections below give the reasoning.
Myth
Reality
What to measure
There is a safe 70/30 branded to exact split
Google has published no ratio
Your mix against 3 to 4 peers
Every exact-match anchor is a risk
Context and pattern matter
Domains behind the exact rows
Generic anchors are worthless
They make profiles look natural
Generic share, tracked monthly
Row count equals link count that matters
One site can produce hundreds of rows
Shares by linking domain
A spike means a penalty is coming
Spikes have many harmless causes
Source of the new rows
Anchor tools agree with each other
Each index differs and refreshes differently
Trend inside one tool
Why each myth survives
Every one of these started with a grain of something true. The trouble is the leap from a pattern to a rule.
The ratio myth
Before Penguin 4.0 in September 2016, over-optimised anchors could trigger a sitewide demotion, and people wanted a threshold to stay under. Since then evaluation is real-time and mostly devaluing, and still no number was released.
The exact-match scare
Exact anchors earned by editorial mentions look different from 200 identical anchors bought in one week. The text matches, the pattern does not. Read the worked example for the domain-level count.
The row-count trap
A footer link on a 3,000-page site yields 3,000 rows and one real vote. Always add a linking-domain column before computing a share.
The tool-agreement hope
Our index holds 181,132,097 domains and refreshes about monthly. Ahrefs, Semrush and Majestic crawl on their own schedules. Expect differences and compare trends, not absolutes.
Spikes, and what usually causes them
Not every jump is hostile. A press mention syndicated to 40 sites will add 40 rows overnight. A site migration can double your count when old URLs get re-crawled. A scraper network may copy your page and its links.
Dated Google events help here. If your anchor mix shifted in March 2024 or August 2025, both periods of spam-focused updates, check whether the rows came from sites that were later hit. Correlation is cheap, so look at the actual linking domains first.
Then ask the plainest question. Did anything on your side cause it? Walk through the migration checklist before blaming outsiders.
One last source of false alarms is the tool itself. Switching from one data source to another mid-year changes counts overnight without a single link moving.
Position
Treat anchor shares as a gauge, not a gate
A gauge tells you the state of things and lets you notice change. A gate suggests a pass or fail line that does not exist. We built the tooling for the first use.
Compare against peers
Count by domain
Watch the trend across at least three snapshots
Two follow-ups we hear often
Does nofollow change the anchor picture? Not the text, only the weight a link may carry. A nofollow link still appears in a backlink list with its anchor, so decide up front whether shares include or exclude them, then keep that rule fixed.
Should exact-match rows be disavowed? Rarely. Only when the links were not earned and cannot be removed, since Google has said for years that most sites never need a disavow file, and a blanket disavow can discard good links along with the bad.