Home › Index
The index of legal AI tools, on the record.
As of Oct 26, 2026, Perplexity is No. 1 of 14 tools with 10 or more voices on record, at a net of +67 (90% band +45 to +80) across 36 voices. Its lead over No. 2 holds in 98% of resamples.
THE RECORD as of Oct 26, 20266571 classified voices · Reddit 5788, Hacker News 530, Bluesky 199, Lemmy 37, Mastodon 14, Stack Exchange 3 · Jan 2023 to Oct 2026+241 in the last 30 daysNumbered from 10 voicesBands are 90%Nobody pays for positionMethod →
| No. | Tool | Voices1 | Net2 | Signal3 | 90 days4 | Ahead of next5 | Top complaint | Most valued | Last voice | |
|---|---|---|---|---|---|---|---|---|---|---|
| NO.1 | PerplexityPerplexity AIAI search | 36voices 28 like · 2 mixed · 4 dislike Confidence Low | +67 net | 90% band +45 to +80 | too few to compare | 98% vs Claude | Accuracy (8) | Time saved (8) | Sep 12, 2026 | |
| NO.2 | ClaudeAnthropicGeneral model | 872voices 492 like · 103 mixed · 169 dislike Confidence High | +37 net | 90% band +33 to +41 | No clear move +7 pts band −4 to +16 | 34% vs Spellbook | Privacy (145) | Time saved (112) | Oct 24, 2026 | |
| NO.3 | 99voices 58 like · 18 mixed · 18 dislike Confidence Medium | +40 net | 90% band +27 to +52 | No clear move −11 pts band −57 to +25 | 28% vs vLex Vincent | Accuracy (18) | Time saved (18) | Oct 9, 2026 | ||
| NO.4 |
| 31voices 22 like · 2 mixed · 6 dislike Confidence Low | +52 net | 90% band +24 to +70 | too few to compare | 91% vs Gemini | Accuracy (5) | Data coverage (6) | Oct 1, 2026 | |
| NO.5 | GeminiGoogleGeneral model | 134voices 73 like · 18 mixed · 36 dislike Confidence Medium | +28 net | 90% band +15 to +39 | No clear move −41 pts band −84 to +5 | 21% vs Paxton | Hallucination (27) | Accuracy (29) | Oct 13, 2026 | |
| NO.6 | 12voices 8 like · 2 mixed · 2 dislike Confidence Low | +50 net | 90% band +7 to +75 | too few to compare | 95% vs Westlaw AI | Price (3) | Time saved (2) | Aug 3, 2026 | ||
| NO.7 | Westlaw AI Research platform | 268voices 106 like · 59 mixed · 91 dislike Confidence Medium | +6 net | 90% band −3 to +14 | No clear move +4 pts band −28 to +34 | 69% vs CoCounsel | Accuracy (112) | Time saved (61) | Oct 15, 2026 | |
| NO.8 | CoCounselLegal AI assistant | 411voices 152 like · 84 mixed · 142 dislike Confidence High | +2 net | 90% band −4 to +9 | No clear move +7 pts band −12 to +25 | 40% vs Copilot | Accuracy (148) | Time saved (55) | Oct 19, 2026 | |
| NO.9 |
| 155voices 61 like · 24 mixed · 54 dislike Confidence Medium | +5 net | 90% band −7 to +16 | No clear move +11 pts band −20 to +39 | 100% vs ChatGPT | Accuracy (38) | Time saved (17) | Oct 17, 2026 | |
| NO.10 |
| 2467voices 832 like · 321 mixed · 1162 dislike Confidence High | −13 net | 90% band −16 to −10 | No clear move −7 pts band −19 to +5 | 14% vs Legora | Hallucination (665) | Time saved (351) | Oct 24, 2026 | |
| NO.11 |
| 90voices 33 like · 10 mixed · 36 dislike Confidence Medium | −3 net | 90% band −18 to +11 | No clear move +10 pts band −26 to +44 | 85% vs Lexis+ AI | Accuracy (19) | Time saved (8) | Oct 16, 2026 | |
| NO.12 | Lexis+ AI Research platform | 294voices 94 like · 46 mixed · 136 dislike Confidence Medium | −14 net | 90% band −23 to −6 | No clear move −24 pts band −50 to +10 | 86% vs Harvey | Accuracy (132) | Time saved (47) | Oct 21, 2026 | |
| NO.13 |
| 749voices 204 like · 101 mixed · 359 dislike Confidence High | −21 net | 90% band −26 to −15 | No clear move +7 pts band −5 to +19 | 20% vs Clio | Accuracy (156) | Time saved (72) | Oct 20, 2026 | |
| NO.14 |
| 78voices 24 like · 13 mixed · 34 dislike Confidence Medium | −13 net | 90% band −28 to +3 | No clear move −7 pts band −37 to +28 | – | Accuracy (21) | Time saved (7) | Oct 6, 2026 | |
Threshold of record · fewer than 10 voices: listed, not numbered | ||||||||||
| LISTED |
| 9voices 3 like · 1 mixed · 4 dislike Confidence Very low | – not numbered | 90% band −51 to +33 | too few to compare | – | Accuracy (4) | Ease of use (1) | Oct 18, 2025 | |
Volume, normalized
How much each tool is discussed and how often, per 100 voices, people dislike it.
ChatGPT has the most dislike voices in raw numbers (1162) because it has the most voices of any kind (43.2% of all voices on named tools). Per 100 voices, the highest dislike rate among tools with 10+ voices is Harvey (48 per 100). Rates and bands compare; totals don’t.
| Tool | Voices | Share of voice 90% band | Confidence | Dislike, count | Dislike per 100 | Like per 100 | Hallucination complaints per 100 | Net 90% band |
|---|---|---|---|---|---|---|---|---|
| ChatGPT | 2467 | 43.2% 42.2–44.3 | High | 1162 | 47 | 34 | 27 | −13 −16 to −10 |
| Claude | 872 | 15.3% 14.5–16.1 | High | 169 | 19 | 56 | 6 | +37 +33 to +41 |
| Harvey | 749 | 13.1% 12.4–13.9 | High | 359 | 48 | 27 | 5 | −21 −26 to −15 |
| CoCounsel | 411 | 7.2% 6.7–7.8 | High | 142 | 35 | 37 | 7 | +2 −4 to +9 |
| Lexis+ AI | 294 | 5.2% 4.7–5.6 | Medium | 136 | 46 | 32 | 11 | −14 −23 to −6 |
| Westlaw AI | 268 | 4.7% 4.3–5.2 | Medium | 91 | 34 | 40 | 12 | +6 −3 to +14 |
| Copilot | 155 | 2.7% 2.4–3.1 | Medium | 54 | 35 | 39 | 7 | +5 −7 to +16 |
| Gemini | 134 | 2.3% 2.0–2.7 | Medium | 36 | 27 | 54 | 20 | +28 +15 to +39 |
| Spellbook | 99 | 1.7% 1.5–2.0 | Medium | 18 | 18 | 59 | 5 | +40 +27 to +52 |
| Legora | 90 | 1.6% 1.3–1.9 | Medium | 36 | 40 | 37 | 3 | −3 −18 to +11 |
| Clio | 78 | 1.4% 1.1–1.6 | Medium | 34 | 44 | 31 | 0 | −13 −28 to +3 |
| Perplexity | 36 | 0.6% 0.5–0.8 | Low | 4 | 11 | 78 | 6 | +67 +45 to +80 |
| vLex Vincent | 31 | 0.5% 0.4–0.7 | Low | 6 | 19 | 71 | 10 | +52 +24 to +70 |
| Paxton | 12 | 0.2% 0.1–0.3 | Low | 2 | 17 | 67 | 0 | +50 +7 to +75 |
| Luminance | 9 | 0.2% 0.1–0.3 | Very low | 4 | 44 | 33 | 0 | −11 −51 to +33 |
By legal task
Net where a tool has 5 or more voices on that task; otherwise just the count. A voice counts under one task: the one it is about.
| Tool | Legal research 1621 | Drafting 1033 | Contract review 305 | Citation checking 290 | Discovery 97 | Summarization 286 | General use 2662 |
|---|---|---|---|---|---|---|---|
| Perplexity | +88 n=17 | 4 | 2 | 3 | 2 | +50 n=8 | |
| Claude | +41 n=82 | +57 n=141 | +46 n=65 | +16 n=31 | +35 n=17 | +59 n=27 | +28 n=455 |
| Spellbook | 1 | +32 n=22 | +39 n=51 | 1 | 1 | +43 n=23 | |
| vLex Vincent | +57 n=21 | 1 | +56 n=9 | ||||
| Gemini | +42 n=31 | +48 n=27 | 3 | −56 n=9 | 2 | +36 n=11 | +11 n=45 |
| Paxton | +83 n=6 | 1 | 1 | 3 | |||
| Westlaw AI | +13 n=210 | 3 | −45 n=11 | 1 | −9 n=11 | −29 n=31 | |
| CoCounsel | +17 n=168 | +3 n=35 | −17 n=6 | +14 n=7 | 0 n=18 | −30 n=33 | −3 n=135 |
| Copilot | −30 n=10 | +17 n=30 | +22 n=9 | 2 | 2 | +11 n=19 | 0 n=78 |
| ChatGPT | −34 n=623 | +22 n=573 | −11 n=57 | −69 n=157 | +9 n=32 | +5 n=114 | −18 n=790 |
| Legora | 4 | +22 n=9 | −33 n=6 | 1 | 2 | 3 | −5 n=63 |
| Lexis+ AI | −10 n=184 | −11 n=18 | −47 n=17 | 1 | −40 n=10 | −10 n=61 | |
| Harvey | −11 n=56 | +20 n=59 | +12 n=43 | −67 n=6 | +20 n=10 | +11 n=19 | −30 n=545 |
| Clio | +64 n=11 | 2 | 3 | −24 n=34 | |||
| Luminance | 1 | −33 n=6 | 2 |
Where this answer continues
Share of voice measures what lawyers talk about, not what firms have bought. Which tools firms have deployed, from adoption surveys and firm announcements, is in What AI law firms actually use, on aivortex.io.
Notes
- 1Voices: classified posts and comments about one tool in legal work, since Jan 1, 2023 (6571 total; 866 are about AI in law in general and name no single tool).
- 2Net: like minus dislike, as a share of voices. Mixed and neutral count in the base. Tools are numbered only from 10 voices, and the numbering follows the floor of the 90% band, so a small sample cannot outrank a large one on a lucky average.
- 3Signal: the dot is the net, the bar its 90% band (Bayesian bootstrap). Bands that overlap mean the order between those tools is not settled.
- 490 days: net of the last 90 days minus net before, with its band. Arrows only when the band excludes zero and the move is not explained by a change in source mix.
- 5Ahead of next: share of resamples in which the tool's net is above the next numbered tool's.
- 6"Lawyer" is used only when the author says so in the quoted text.
- 7Logos: official marks from Simple Icons, Wikimedia Commons or the vendor's press kit, in one ink. Tools without an official file are set in type.
Corrections, dated
Oct 22, 2026
Privacy. Usernames and handles are no longer shown anywhere on the site, its share pages, images or data file. Quotes now show the community and the date; the link to the original post stays so every quote can still be checked. A build check fails the site if a handle appears.
Oct 22, 2026
All tools. The record grew from 709 to several thousand voices: the full Reddit history since 2023 from a public archive, the full Hacker News history, Bluesky, more of Mastodon and Lemmy. Shares, nets and bands moved for every tool. We now show share of voice and rates per 100 voices next to every count, because a tool that is used more is also criticized more in raw numbers.
Oct 22, 2026
Stack Exchange. Stack Exchange posts are counted and summarized but no longer quoted: their license asks for the author name next to any quote, and we no longer show names.
Oct 22, 2026
All tools. Added 71 voices from Mastodon and Lemmy and 19 more Hacker News voices from our second collection round. 55 trade-press articles (Artificial Lawyer, Lawyerist) were collected too but are not counted: they report on tools, they are not lawyers' voices.
Cite this index as: aiforlawyers, Index of legal AI tools (updated Oct 26, 2026), 6571 voices. Data: index.json.
Use these numbers: the counts, rates and bands on this index are free to reuse under CC BY 4.0. Credit “aiforlawyers.app (AI Vortex)” with the date and a link to the page. Quotes belong to the people who wrote them: quote them with the link to the original post. Machine-readable: index.json. How to cite.