Does anchor text type actually correlate with Google rankings?

Anchor text diversity advice is everywhere: keep exact match under some threshold, mix in branded and generic anchors, avoid looking “unnatural.” Almost none of it points to a number. Ahrefs’ anchor text study is the closest thing the industry has to one, because it measured the actual correlation between anchor text type and rank, at scale, and published the coefficients.

What Ahrefs actually measured#

Ahrefs pulled the top 20 Google results for 19,840 keywords with 2,000 to 5,000 monthly searches, each 2 to 4 English words with no special characters, for a total of 384,614 web pages. To isolate anchor text from the other obvious confound, backlink authority, it kept only keywords where the top 10 ranking pages had similar URL Rating scores, so pages in the sample were roughly matched on link authority before anchor text was measured.

For every page, Ahrefs calculated what share of its backlinks used exact-match, phrase-match, partial-match or random anchor text, then ran a Spearman correlation between each share and the page’s ranking position. Spearman correlation, like the Pearson coefficient used in other ranking-factor studies, runs from -1 to 1, and values below roughly 0.3 are conventionally read as weak, not values a site owner could use to predict rank from anchor text alone.

The numbers, and why none of them are strong#

Exact-match anchor text came out highest, at 0.1436 average and 0.1869 median, still below the conventional threshold for even a moderate correlation. Phrase-match and partial-match landed close together, 0.1057 and 0.1076 on average, statistically indistinguishable from each other in this dataset. Random anchor text, links using text unrelated to the page’s target keyword, correlated at 0.0161, functionally no relationship at all.

Ahrefs ran a second, separate test on the words surrounding a link rather than the anchor text itself, using a reduced sample of 16,000 pages, 800 at each ranking position from 1 to 20. Exact-match keywords in the surrounding text correlated at only 0.0640, and pages with one or more matching words nearby scored a slightly negative -0.0701, meaning having relevant words near a link showed no positive association with rank in that sample, and if anything a faint inverse one.

Why we care#

The consistent pattern across both tests is that anchor text type is a real but small signal, not the mechanism most link-building advice implies it to be. Exact match beats random match, which is a defensible, expected result. But the gap between them is the difference between a correlation of roughly 0.14 and roughly 0.02, both weak, not the difference between a strong signal and a nonexistent one. Anyone treating anchor text ratios as a lever that reliably moves rankings is reading more certainty into these numbers than Ahrefs’ own study supports; the honest takeaway is that anchor text relevance helps marginally, and no combination of match types substitutes for the URL Rating and topical relevance Ahrefs controlled for to isolate it in the first place.