Anchor Text Experiment: Exact Match at Scale

Anchor text is where link building's greed meets Google's pattern-matching: exact-match anchors demonstrably communicate relevance, over-using them is the most documented penalty signature in the field's history, and everyone wants to know where the line sits. This writeup combines a small controlled observation with the more valuable dataset — anchor distributions across the healthy and penalised profiles we've audited — into the practical anchor doctrine. Spoiler: the line is blurrier and lower than the optimisers hope.

The observation: exact match versus branded, matched pages

Within one campaign's placements, comparable target pages received links differing mainly in anchor treatment: one group's placements carried partial/exact topical anchors (natural-sounding, editorially plausible — "guide to [topic]" style), the other's carried branded/URL anchors, volumes and venue quality matched. At three months: the topical-anchor group showed modestly faster movement on the anchored terms — a couple of positions' head start that the branded group substantially closed by month five. Reading: anchor relevance is real and front-loaded — it accelerates term association; it isn't the difference between ranking and not, and its edge decays as the page's own signals accumulate. Worth having where it occurs naturally; not worth engineering risk for.

The audit dataset: what profiles actually look like

Across the profiles we've audited, the distributions tell the enforcement story better than any test: healthy, never-penalised profiles run overwhelmingly branded, URL and generic anchors ("BacklinksMedia", "backlinksmedia.com", "this study", "here") with exact-match commercial anchors in the low single digits percentage-wise — because real editors write sentences, not keyword insertions. The penalised and devalued profiles (including the recovery case's) inverted this: money anchors at 20–60%, the unmistakable fingerprint of placements bought by keyword. The tell isn't any single anchor — it's a distribution no natural process produces, and distribution analysis is cheap at Google's scale.

The doctrine

1. Don't specify anchors in earned outreach — ever. Editors choosing their own words is both the definition of editorial and the natural distribution's source; anchor requests are the tell that converts a placement from citation to scheme, per the paid-link line.
2. Engineer anchors only where you legitimately control them — your own internal links (where descriptive anchors demonstrably help, per the experiment, and no spam pattern applies) and your syndicated bylines/bios (branded, varied).
3. Make the title do the anchor's job: pages get linked by their titles — a title containing the target phrase naturally harvests partial-match anchors from every organic citation, which is the white-hat version of anchor optimisation and the only scalable one.
4. Audit by distribution, not incident: the quarterly read flags drift toward money-anchor concentration — usually the first visible symptom of a vendor cutting corners, caught here before it's caught elsewhere.

Frequently asked questions

Is there a "safe percentage" of exact-match anchors?

Every published threshold is folklore — enforcement reads distributions in context (niche norms, velocity, venue quality), not one ratio. The operational answer: if you're choosing a percentage you're already on the wrong side of the process; earned profiles land in the safe zone automatically because nobody's choosing.

My competitor runs 40% money anchors and ranks. Why?

The series' survivorship answer: detection is probabilistic, lag is real, and the modern response is often silent devaluation — their anchors may already be passing nothing while their content ranks on other merits. Copying the visible half of someone's strategy without seeing the enforcement half is how the recovery stories begin.

Should I dilute an over-optimised profile with branded links?

After genuine cleanup of the scheme-shaped sources (per the workflow), continuing normal earned acquisition naturally rebalances the distribution — that's remediation. "Dilution campaigns" that just pile new links atop an untouched scheme layer treat the symptom the algorithm isn't even reading. The durable position is the boring one: anchors you never think about, from placements editors control (our standard practice).