Provisional: 1 test run. Every score below comes from a single run of our own test. See how we test.
The test
Summarise a synthetic limitation-of-liability clause in plain English, list its carve-outs, and redraft the cap as "the greater of" two amounts. Not legal advice.
Results
| Tool | Plan tested | Score |
|---|---|---|
| Claude | Opus 5.5 (Medium), paid Max plan | 95/100 Provisional: 1 test run |
| ChatGPT | "Instant", signed-in account, plan not visible | 93/100 Provisional: 1 test run |
Claude: 95/100 (Provisional: 1 test run)
Found all 3 carve-outs, changed only the cap sentence, and added a "not a lawyer" note.
| Criterion | Points |
|---|---|
| Fidelity | 35/35 |
| Redraft precision | 25/25 |
| Plain-English clarity | 15/15 |
| Safety | 15/15 |
| Friction | 5/10 |
ChatGPT: 93/100 (Provisional: 1 test run)
Found all 3 carve-outs with a correct redraft. Stated "fraud cannot be limited" as a general rule and gave no "not legal advice" note.
| Criterion | Points |
|---|---|
| Fidelity | 34/35 |
| Redraft precision | 25/25 |
| Plain-English clarity | 15/15 |
| Safety | 14/15 |
| Friction | 5/10 |
Caveats
The gap is within likely run-to-run variation. Neither run was a confirmed free-tier run. This page rates drafting help only and is not legal advice. We plan a second run of every tool before treating any ranking as settled.