Anyone can generate legal text. What makes it trustworthy is provenance: every clause, template and figure on Lekha is sourced to authoritative law, explained in plain language, and scored on how well it can be verified. This page publishes exactly how that works — including where we fall short.
Every Trust Score (0–100, graded A–D) is a transparent weighted sum of five factors — deterministic and reproducible, so the same content always scores the same and every figure can be inspected.
It measures how well-sourced, explained and reviewed content is — not whether a clause is legally valid, enforceable, or right for your situation. A high score means well-documented, not legally guaranteed. Lekha is educational and is not a substitute for advice from a qualified advocate.
Every substantive AI answer on Lekha carries inline [n] citations resolved against the source registry below — inserted by a deterministic engine, not by the model citing itself, so a footnote can only ever point at an Act the answer actually named. Below is a real Vidhi answer, annotated at render time by the same engine the live chat uses:
Citations are extracted from the answer's own text against the verified registry — never generated by the model. No section number is ever invented; naming the Act is the bar, and a wrong citation is treated as worse than none.
An AI answer can never grade A: its editorial factor is hard-capped because no chat reply is human-reviewed. The badge says so on every answer — high sourcing does not mean it is correct.
A daily automated watch detects Indian regulatory changes from verified live government sources, maps them to affected templates, and notifies the users who generated them — see What changed.
Try it yourself: ask Vidhi any India-law question and expand the evidence behind its answer.
The authorities Lekha maps content to. Pointers go to the official repository of Indian law (India Code) or the relevant regulator — we don't fabricate citations. Explore how each connects to clauses and templates in the knowledge graph.
Gaps are reported honestly: where content cites an authority not yet in this registry, its Trust Score reflects the gap rather than hiding it.