We asked three AI engines who to buy for legal practice management software, 12 different ways, 3 times each. Clio came back in 100% of them. More than half of the 43 vendors we measured came back in none of them.
Published · 43 vendors · 36 engine responses
of the 43 legal practice management software vendors we measured were never named once, in any engine, across all 12 checks.
Buyers increasingly ask an AI engine before they ask a search engine. So we asked the engines directly: who should someone buy for legal practice management software? We asked 12 different ways, across 3 engines, and repeated the whole thing 3 times to separate real results from model noise.
More than half of the 43 vendors were never named once. Not ranked low. Not mentioned in passing. Absent from every response, every time. Of those, 28 produced that same zero on every single repeat, which makes it a measurement rather than an unlucky sample.
15 of 43 vendors in the sampling frame were named at least once. Ranked by how often, so the shape of who actually wins the category is visible at a glance. 12 of those have a score that reproduced on every repeat; 3 appeared in only one of 3 and are marked 'not reproduced'.
| Vendor | Named in | Rate | ShareVisibility | Reproducibility | Status |
|---|---|---|---|---|---|
| Clio# | 12 of 12 | 100% | stable | Cited | |
| Smokeball# | 12 of 12 | 97% | noisy | Cited | |
| PracticePanther# | 11 of 12 | 89% | noisy | Cited | |
| Filevine# | 10 of 12 | 83% | noisy | Cited | |
| CARET Legal# | 9 of 12 | 78% | noisy | Cited | |
| CosmoLex# | 6 of 12 | 50% | noisy | Cited | |
| Assembly Software# | 5 of 12 | 44% | noisy | Cited | |
| Neos# | 3 of 12 | 25% | stable | Cited | |
| Actionstep# | 2 of 12 | 22% | noisy | Cited | |
| 8am MyCase# | 1 of 12 | 11% | noisy | Cited | |
| 8am CasePeer# | 1 of 12 | 6% | noisy | Cited | |
| Rocket Matter# | 0 of 12 | 6% | noisy | Not reproduced | |
| Tabs3# | 1 of 12 | 6% | noisy | Cited | |
| AbacusLaw# | 0 of 12 | 3% | noisy | Not reproduced | |
| Centerbase# | 0 of 12 | 3% | noisy | Not reproduced |
The other 28 vendors never came up, not once, in 12 checks repeated 3 times. Collapsed here because a list of identical zeros isn't something to read row by row -- it's the shape of the finding. An asterisk marks a zero that reproduced on every single repeat, the strongest version of this claim.
Every response above came from somewhere. Across all 36 responses for this category, these are the pages the engines cited most, ranked by how often each was pulled from (237 citations captured).
| Source | Cited | Example |
|---|---|---|
| referent.law | 38× | Best Legal Practice Management Software (Ranked) | Referent |
| learn.g2.com | 33× | I Explored G2's 6 Best Legal Practice Management Software |
| youtube.com | 33× | 5 Best Legal Practice Management Software (Full Demo & Comparison) |
| quickbase.com | 29× | The Best Legal Case & Practice Management Software ... |
| counselstack.io | 27× | CounselStack — Legal Practice Management Software Reviews |
| g2.com | 18× | Best Legal Practice Management Software |
| legalsoft.com | 10× | 8 Best Legal Case Management Software for Law Firms ... |
| capterra.com | 9× | Best Law Practice Management Software 2026 |
| pagelightprime.com | 7× | 2026 Guide: Best Legal Practice Management Software |
| rankings.io | 6× | 10 Best Law Practice Management Software In 2026 |
This section exists so the numbers can be checked rather than trusted. Anything that could not be reproduced is labelled as such and is not quoted as a figure anywhere on this page.
All 43 vendors came from one published category list. Nothing was removed for looking like a weak fit. Filtering a sampling frame after seeing it turns a census into an opinion.
Perplexity (sonar, live web search); ChatGPT (gpt-4o-search-preview, live web search); Gemini (Flash, grounded with Google Search). Each performs live web search rather than answering from training data.
Every query ran 3 times. These models are non-deterministic, so a single run is an anecdote. Results that did not hold across repeats are marked noisy and are not quoted as figures.
The vendor is named in the engine's answer to a buyer question. We never ask about a vendor by name, because that only proves the model has heard of them.
Buyer-style questions, not brand searches. Asking an engine about a company by name proves only that it has heard of them.
This measures one category on one date, from one published list. AI answers change as the web changes, so these numbers describe July 2026 and not a permanent state. Vendors absent from the source list are absent here too, which is a property of the frame and not a judgment. Exact non-zero counts drift by a check or two between runs, which is why reproducibility is published per row instead of hidden.
Clio. We asked 3 live-search AI engines (Perplexity, ChatGPT, Gemini) 12 different buyer-style questions, 3 times each, on July 30, 2026. Clio was named in 100% of those checks.
More than half of the 43 vendors we measured were never named once, in any engine, across all 12 checks. That is not ranking low. It is being absent from every answer, every time.
3 AI engines with live web search enabled (Perplexity, ChatGPT, Gemini), 12 buyer-style questions per vendor, and the entire run repeated 3 times to separate a real result from model noise. The sampling frame was 43 vendors. Exact queries and limitations are published on this page.
Because these engines answer from the sources they can cite, not from a vendor's own marketing. A company can rank well in Google, publish constantly, and still be absent from every AI answer if it is not present in the pages those engines actually quote. This is a citation gap, not a content gap.
Left: every vendor AI named at least once, tallest first. Right: the 28 that were never named, shown as one block because 28 identical zeros carry a single fact. Two panels, not one axis.
Re-measured on a schedule. Each run is added, never overwritten, so a figure quoted from an earlier date stays checkable.
| Measured | Never named | Leader | Vendors |
|---|---|---|---|
| 2026-07-30 | 65% | Clio 100% | 43 |
| 2026-07-31 | 67% | Clio 100% | 43 |
The Watchlist is the same measurement, re-run monthly, with the parts a one-off study cannot show you: who moved, which pages the engines started quoting, and which of those a competitor owns.
$249/mo, or $2,490/yr.
See what is in the monthly reportType a vendor name to see how it scored. The full table above already contains every result, so this is a shortcut, not a gate. No email required.
