12AM Agency · Law-firm research
Which law firms appear in AI answers?
What does the classified-firm census show?
Of 61,409 U.S. law firms classified in CounselIndex, 97.9% did not appear in any recorded AI answer.
Research, not a client result. 2,094 recorded AI-naming observations across 332 U.S. legal markets, data through August 2026, methodology v1.0.
The index contains 127,000+ U.S. law-firm offices. The census covers 61,409 classified firms. Offices, firms, and recorded observations are different units.
What do the visibility states mean?
| State | Definition |
|---|---|
| Invisible | Fails the machine-visibility gate. |
| Findable | Machine-visible, not yet in a recorded answer. |
| AI-Named | Appeared in at least one recorded answer. |
| AI-Selected | Named first or named repeatedly. |
The AI Visibility Score uses a 100-point scale. Recorded states and score are distinct measurements.
What did the blinded study find?
A blinded, stratified study covered 400 law firms, with 100 per visibility state. Independent research sessions replicated the community-discussion finding twice. These are correlations, not causation.
| Signal | Observed result | Statistical context |
|---|---|---|
| Organic community discussion | AI-Selected 43%; Invisible 12%; 3.6× rate | Trend p = 1.3×10⁻⁷ unadjusted; 6.7×10⁻⁷ after Bonferroni correction |
| Legacy directory standing | No measurable association; Avvo, Justia, FindLaw | ρ = 0.04, exploratory |
| Recognized accreditations | AI-Selected 71%; Invisible 48%; Findable 46% | p = 6.9×10⁻⁵ unadjusted; the pattern does not rise at every state |
| Mean star ratings | Nearly flat across states: 4.65–4.71 | Descriptive comparison |
Community discussion includes Reddit, forums, and Q&A. Recognized accreditations include Super Lawyers, AV, and board certification.
How should the research be interpreted?
The classified-firm census summarizes recorded AI-answer observations through August 2026 under methodology v1.0. The blinded study compares independently assessed signals across visibility states. Neither design establishes that increasing the AI Visibility Score causes AI naming. The research describes visibility, not legal skill or client outcomes.
Review signals contribute to the classification gate. Review-volume ratios can partly measure that classification rule itself and should not guide a causal sales claim.
How does the earlier practice-matched sweep differ?
The July 27–30, 2026 practice-matched sweep found that 96.6% of 13,053 practice-matched firms did not appear in its recorded answers. That dated sweep uses a narrower cohort. It is separate from the full classified-firm census above.
How should you cite this research?
12AM Agency, CounselIndex law-firm AI visibility research. Data through August 2026, methodology v1.0. Specify whether citing the classified-firm census, blinded study, or July practice-matched sweep, and retain the relevant denominator.
Canonical source: https://12amagency.com/our-work/chatgpt-law-firm-recommendations-ai-visibility-study/