Why should I trust it?
Most tools in this space will tell you how accurate they are.
This page is the opposite, and it is the most important one on this site.
The judge
It gives you a score. We will not stand behind it yet.
LexGrit's judge scores your argument across several criteria. Those scores are not calibrated — nobody has yet checked them against a human expert's scores on the same sessions. So they are withheld wherever the number would be treated as evidence.
Absent from the exportable practice record
That document exists so a credentialing body can see the activity happened, and a credentialing record is the single worst place to put an unvalidated number. It says an assessment was performed and on which criteria. It does not say you scored 78.
Absent from your long-run patterns
A trend line drawn from an unmeasured score looks confident and means nothing.
The machinery to fix this is built — a calibration harness that reports variance beside every score, and a floor below which a change between versions is called noise rather than improvement. What is missing is the human scoring, and until that exists, we say so instead of estimating.
The objection engine
It has not published its accuracy either.
It is the most distinctive thing here, and it interrupts you mid-sentence. An objection engine that interrupts wrongly is not neutral — it teaches a reflex you will carry into a courtroom.
So it will be measured against fragments captured from real sessions and labelled by a human, and those figures will be published per tier and per ground when they exist.They do not exist yet.
You are reading the honest version of a claim most products would simply have made.
What is measured, today
Not everything here is uncertain. These are counted, not judged.
Which must-hit points you addressed
Each with the verbatim line where you addressed it. A hit claimed without a quotable line is recorded as a miss.
Which objection grounds actually landed against you
And how often, across every session — counted once per session, so one heated afternoon cannot look like a standing weakness.
How long you took to recover after an interruption
Measured from turn timings in your browser.
Where every ruling comes from
The ruling carries the passages it relied on.
Which parts of your pleading, which sections of the court's rules — quoted rather than paraphrased. If a ruling cited something the retrieval could not support, the citation is flagged in the moment it is spoken, not discovered by you afterwards.
And your voice is never kept
Session audio is not recorded, stored, or sent anywhere for storage.
It is your client's case in your own voice; the numbers derived from it are enough, and they are computed as you speak.