One page in our dataset ranks third on a term with 2,500 searches a month. Domain rating: 11. Backlinks to the URL: zero. Referring domains to the URL: zero. That page sits above firms with link profiles that cost real money to build, and it does so right now, in a live Google result.
That anomaly is what made us run the numbers.
What we measured and how
We pulled 44 Google SERPs across 10 Florida markets and five practice areas: personal injury, car accident, DUI, divorce, and wrongful death. We fingerprinted 446 pages and captured 717 ranking placements, giving us 305 organic law firm results to work with. For each ranking URL, we recorded backlink count, referring domain count, URL rating, domain rating, and content signals. I should be upfront about what we did not measure: user behavior, click-through rates, and Google Business Profile signals. The local map pack sits above everything we analyzed, and no content play touches that block. This is one state at one point in time. Those are real limits, and I am not going to pretend otherwise.
What the correlation numbers actually say
Backlinks to the ranking URL correlate with position at -0.179. Referring domains to the URL correlate at -0.244. Both are statistically significant at n=305, but they explain roughly 4 to 6 percent of rank variance. Backlinks was the weakest of the five metrics we measured.
The counter-examples put texture on that number. One page with 47,404 backlinks and the highest URL rating in its entire result set ranks eighth. A site with domain rating 0 ranks tenth on a high-difficulty term. The page I opened with, DR 11, zero links, ranks third. If you have been paying a retainer specifically to accumulate backlinks on the theory that volume predicts position, these results are worth sitting with.
Correlation is not causation, and a null result from a single-state sample is still a small-sample null. I am not claiming we cracked the algorithm. I am saying we went looking for the thing practitioners charge most to build, and we found the weakest signal of the five we tested.
Referring domains work as a floor, not a ladder
Here is where the data gets actionable. Ninety-nine percent of domains that ranked had 300 or more referring domains. The 10th percentile for ranking domains sat between 372 and 441 referring domains across all five practice areas. That threshold held consistently, which means clearing it appears to matter.
What does not appear to matter much is how far above the floor you go. When we compared the referring domain distributions for top-three finishers against bottom-three finishers in the same SERPs, those distributions overlapped heavily. Firms ranked first and firms ranked eighth were drawing from similar referring domain ranges. The floor is real; the ladder is not.
For a firm with under 300 referring domains, some link acquisition is probably worth doing. For a firm already past 400, the marginal spend on more links is harder to justify based on what this data shows.
Does the query wording change everything?
We ran 50 pairs of queries that mean the same thing to a prospective client: "personal injury lawyer Miami" versus "personal injury attorney Miami," and so on across markets and practice areas. The top ten shared only 64.3 percent of its results across those pairs. Median Kendall tau was 0.429, meaning rank order within the shared results was only weakly preserved. Only 4 of the 50 pairs held their full top-ten order.
Why does this matter for interpreting the link data? Because site-level factors like domain rating and referring domains are identical between a lawyer query and an attorney query by construction. The site does not change. The links do not change. And yet roughly 40 percent of the result set changes on a single word. That variation has to come from somewhere other than the link profile, and it suggests that query-level signals, page-level content, and whatever Google is doing with intent classification account for a substantial share of what we were attributing to authority metrics.
A firm optimizing one page for one phrasing and assuming it captures the full market is working from an incomplete map.
What junk link profiles on ranking sites tell us
Several sites in our dataset had link profiles that were easy to characterize: referring domains with names like buybacklinks.agency and pbnseolinks.shop appeared in the referring domain lists of pages that rank. Those sites rank anyway, sitting in positions two through six on terms with meaningful search volume.
I am not recommending purchased links. I am noting what the data shows, which is that the spend on that link profile is not what holds those positions. If it were, the sites with clean, expensive, editorially earned links would outrank them consistently. Some do. Many do not.
The constraint most firms are actually hitting
If link volume explains 4 to 6 percent of rank variance, and query wording alone shifts 40 percent of a result set, the more productive question for most law firms is whether they have a page for every query a client actually types. Not a page optimized for "personal injury lawyer Tampa" that also tries to capture car accident and wrongful death and DUI. One page, one query, built around what a real person types when they need that specific thing.
That is a content coverage problem, and it is cheaper to fix than a link deficit. It is also the kind of problem that does not resolve itself by spending another quarter on a link building retainer.
We will keep running markets. If you want the Florida dataset methodology or want to see the same study run for your state and practice area, the contact is on the Ghosts site.