Atlas Rankings Survived a Second Test. The Real Story Was the Peer Network Behind Them.
Viclaro froze the ten most visible NYC divorce firms in Atlas, ran them through an internal audit sample, and found the ranking signal held up. The deeper story was not just correlation: the audits exposed the peer network, the content patterns that made those peers win, and the exact artifact a firm could implement.
The ranking was not a fluke
Viclaro froze the ten firms most visible in the latest NYC divorce Atlas snapshot and ran them through an internal audit pipeline. That choice matters. The cohort was fixed before the new results were inspected, which means the test was not built to flatter the ranking after the fact.
The question was blunt: does the Atlas signal survive a fresh sample of buyer-style prompts, or does it collapse once you stop looking at the same data twice?
The answer, on the evidence, was that it survived. Seven of the ten firms were numerically comparable, the audit produced 840 usable provider probes, and the reconciled data tracked Atlas closely enough to be difficult to wave away.
What the numbers actually said
Across the seven firms that could be compared directly, Atlas visibility and independent audit visibility moved together. The reconciled data produced a Pearson correlation of 0.82 and a Spearman rank correlation of 0.79.
That is not a casual result. It is a strong directional match between two separately generated samples. Higher Atlas share generally corresponded to higher audit share under a different set of prompts.
The important part is what the result does not say. It does not prove causality. It does not promise a permanent rank. It does not claim a content change will always move the needle. It does say the ranking surface is real enough to matter.
The audit found something more useful than a score
The audits were not just a second way to count citations. They surfaced the reasons firms win or lose on specific prompts: whether the site gives the model a quotable answer, whether the credential is buried or obvious, whether the firm sounds like a generic practice area or a real answer to a real buyer problem, and whether the public evidence is strong enough to support recommendation language.
That is where the value is. Atlas gives you the map. The audit tells you what kind of terrain you are standing on.
In other words: the leaderboard identifies the firms in the fight. The audit shows the content patterns that explain why those firms are there.
The peer network was the real signal
The most interesting part of the audit was not that a few firms scored higher than others. It was that the firms clustered into a peer network. The same small set of alternatives kept showing up around the top names, which is exactly what you would expect if the assistants were not hallucinating a list but working from a stable competitive field.
Atlas shows the market structure. The audit shows the local comparison set. Put the two together and you get a sharper picture: a consensus core at the top, credible challengers underneath it, and a lower tier that tends to win only when a prompt is phrased in a very specific way.
That changes the commercial story. A firm is not just trying to get “visible.” It is trying to become the answer the model chooses over the same peer set it keeps reaching for now.
The recommendations were not abstract advice
The recommendation artifact matters because it ties the peer network back to the content that created it. It points to the patterns the cited peers already have and the audited firm does not: question-shaped headings, concise answer blocks, quotable credentials, and structured proof the model can lift.
That makes the output actionable. If a peer is being cited because it has a direct answer block for a buyer scenario, the fix is not “do more marketing.” The fix is to publish that same kind of answer, with the firm’s own facts, on the right page.
This is the point where the audit stops being commentary and starts becoming a thing a firm can implement.
What the test does not prove
This was a selected top-ten cohort in one category. It does not generalize automatically to every legal vertical, every city, or every assistant.
It also does not prove that publishing the recommendations causes lift. That remains the next test, and it is the one that would matter commercially.
But the bar was lower and still meaningful: does the ranking signal survive a second pass, and does the audit produce a usable map of the peers and the content behind them? On this evidence, yes.
What happens next
The next real test is a before-and-after run with a cooperative firm that implements verified recommendations and then gets measured again against frozen prompts.
If that moves, the story gets much stronger. If it does not, the company still learns something important about how AI assistants are choosing firms in the wild.
Either way, the product is no longer just a ranking display. It is a measurement system with a diagnostic layer and a repeatable artifact firms can use.
Key takeaways
- Atlas and the internal audit sample moved together in the frozen NYC divorce top ten.
- Seven firms completed cleanly, producing 840 usable provider probes and strong rank agreement.
- The audit exposed the peer set and the content patterns behind it, not just a score.
- The recommendations are implementation artifacts, not abstract advice.
- The result supports correlation, not causation.
Sources and further reading
Primary documentation and research used for this field note. Product behavior changes; check the linked source before treating any implementation detail as permanent.
- 1. NYC divorce top-ten validation report — Viclaro
- 2. Atlas public methodology — Viclaro
- 3. NYC divorce leaderboard — Viclaro
- 4. Construct validity of AI citation audits — Viclaro
Next step
Atlas shows the public map. A Viclaro audit turns that map into the prompts your firm is losing and the page edits most likely to change the next scan.
Read the methodology and live leaderboard.