A page with zero backlinks was sitting at position three on a Florida personal injury search that gets 2,500 clicks a month. Zero referring domains, not a handful. I had pulled the data expecting link counts to explain the rankings, and the first anomaly I found broke the premise entirely.
What I went looking for
I have spent 25 years in SEO, and for most of that time I believed what the industry believes: you rank law firms by building links. It is the category's foundational claim, the thing agencies charge the most for, and the metric that shows up first in every monthly report. So I decided to actually check it. I pulled 44 Google results across ten Florida markets and five practice areas, recorded each firm's ranking position, and matched it against the referring-domain count in Ahrefs. Then I ran the correlations.
They came back weak. Backlinks were the weakest predictor I measured, behind page-level signals and a few things that are harder to quantify. I wrote up a draft that evening, and I will be honest: I was pleased with myself. A contrarian finding is a satisfying thing to have, especially one that implicates an entire service category.
The number that looked like data
Before it went out, one question stopped me: had I looked at which domains were in those link profiles, or only at the counts?
I had only looked at the counts. I had treated the Ahrefs referring-domain number as a clean input. It is not.
When I opened the actual link lists, the firm sitting at number ten in one market had 340 referring domains on paper. Scrolling the list, the names told the story: variations on "buy backlinks" and "pbn seo links", thirty of them, all on the cheap new top-level domains that link sellers register by the hundred. Spam networks list real websites automatically, with no contact with the owner, as a way of advertising their own packages. That firm had never bought anything from them. Google is almost certainly ignoring the lot. Strip them out and only five of those 340 domains had any real audience at all, and three of the five were scrapers. I had spent a night correlating that number against rankings as if it meant something.
Noise makes correlations look weaker than they are. My weak correlation between backlinks and rankings was not evidence that links do not matter. It was evidence that the standard link count is a broken measurement.
What the real links looked like
I went back through the data and filtered each profile to domains with measurable real-world traffic, the kind of domain that exists because people use it, not because a link vendor spun it up. The picture inverted.
The firm at number ten in that same market had five qualifying domains once I stripped the noise, and three of those were scrapers pulling from a directory it had not actively sought out. The firm at number two had more than a hundred. Cornell Law, Justia, Avvo, Martindale, the Jacksonville Business Journal, the local NPR affiliate, the University of North Florida, Florida State, county school districts. These were earned links. They were citations accumulated because the firm had been genuinely present in that city for years, quoted in coverage, listed in the places institutions link to when they mean it.
That is a different thing than buying a link package. It is also a harder thing to fake, which is probably why it holds up.
What this means if you are the one paying
Three things I would tell any managing partner who has been writing a check for link building.
Ask to see the list, not the count. A referring-domain number with no list attached is not a report; it is a summary of something your agency may not have looked at either. If the list is full of dot shop domains and hyphenated exact-match garbage, you are being shown spam as evidence of progress.
What separated the firms ranking well was local embeddedness, not link volume. Bar association citations, university mentions, real local press, the directories that actual humans use to find lawyers. You accumulate those by being present, by doing work worth mentioning, by being the firm a journalist calls when they need a quote about the new courthouse. That does not come in a package, and that is precisely why Google keeps rewarding it.
Being wrong in public is part of the job. I published an early version of this analysis that overstated the case against links, and the thing that caught it was going back to the raw list instead of trusting the summary number. That is also the lesson for anyone reading an SEO report: the summary number is where the story gets simplified past the point of usefulness.
The finding that survived all of it
After I corrected the analysis, one result held up that I had not expected to find interesting.
I ran 50 pairs of searches that differ by one word: personal injury lawyer versus personal injury attorney, car accident lawyer versus car accident attorney, across ten markets, same city and same day in each pair. The top ten overlapped about two thirds of the time. Same firms, same link profiles, same domain authority scores, and one different word in the query changed roughly a third of who appeared.
That number is worth sitting with, because everything a link report measures is held still between those two searches. Whatever moved is happening at the level of the specific page and the specific phrasing.
I want to be careful about what I can and cannot claim here. I did not establish that the firms holding steady across both phrasings were doing so because they had a dedicated page for each one. I checked, and only a small fraction of them served a different URL for the two searches. So the mechanism is mostly about which firms Google considers relevant to each phrasing, and I do not know precisely what drives that.
What I would still tell a firm is this: if you have one practice-area page carrying six different ways a client might ask the same question, the link budget is probably not your binding constraint. Coverage is the cheaper thing to test first, and it is testable. Build the page for the exact phrasing and see what happens.
Links matter. Real ones, from real places, earned over time by being an actual part of your city. The referring-domain count in your monthly report, as a raw number, almost certainly is not measuring that. Ask for the list.