ChatGPT for Lawyers, Tested: What Works, Where It Fails
Practical prompts for legal research, NDA review, intake and client letters, with tested results, verified plan prices and confidentiality cautions.
AI for lawyers can help draft contract clauses, outline case research and structure client intake notes for review. You remain responsible for the work: verify legal authorities against official sources in your jurisdiction, and keep confidential client information out of tools whose data handling you have not vetted.
Start with the task in front of you, not the tool. Choose case research, contract drafting or client intake, then adapt a tested prompt to your matter. This hub includes dated, rubric-graded test runs on named model versions; use them to examine the output rather than assume a tool will suit your work. Treat generated text as a draft and check legal authorities against primary sources. Review data handling before entering client information. Read the jurisdiction notes for the US, UK, Australia and India, then verify the applicable guidance with the official source where you practice. You remain responsible.
Practical prompts for legal research, NDA review, intake and client letters, with tested results, verified plan prices and confidentiality cautions.
Compares a counterparty NDA with your client's playbook clause by clause: issue table, risk level, proposed redline language and questions for the supervising attorney.
The comparison does not establish a legal-work winner; check every draft against its sources and your jurisdiction's requirements.
Use the benchmark to shortlist tools, then verify legal conclusions, confidentiality terms, and your jurisdiction's professional rules.
Set source boundaries and flag missing information; independently verify citations and legal conclusions before using a draft.
Check organizational data terms, professional confidentiality duties and draft accuracy before using Copilot for client work.
Not legal advice. Regulators and professional bodies that set the limits for this work.
Rule-by-rule checklist for using generative AI under ABA Formal Opinion 512, with a citation-check prompt and the 2026 California update.
Not legal advice. One official source per country, with the date we checked it. Verify with a licensed professional in your jurisdiction.
The State Bar of California's AI resource page lists its practical guidance and ABA Formal Opinion 512. The guidance maps AI use to competence, diligence and confidentiality duties, noting that AI products may use inputs to train or refine the model.
State Bar of California — Artificial Intelligence resources (incl. ABA Formal Opinion 512)Checked Oct 19, 2026
The Law Society of England and Wales guidance 'Generative AI – the essentials' (updated 30 Jun 2026) flags misuse or disclosure of confidential information, data protection and privacy, and cybersecurity risks of generative AI tools.
The Law Society — Generative AI: the essentials (updated 30 Jun 2026)Checked Oct 19, 2026
The Supreme Court of NSW's Practice Note SC Gen 23 governs the use of generative AI in proceedings. It was amended on 28 January 2025, with minor amendments in effect from 3 February 2025.
Supreme Court of NSW — Generative AI (Practice Note SC Gen 23)Checked Oct 19, 2026
The Press Information Bureau reported on 14 Nov 2025 that the Government of India notified the Digital Personal Data Protection Rules, 2025, with an 18-month phased compliance timeline. Data Fiduciaries must give standalone, clear consent notices.
PIB — Government notifies DPDP Rules 2025 (14 Nov 2025)Checked Oct 19, 2026
Yes, as drafting or research support, with your review and appropriate handling of client information. Start with a contract draft or research outline and check the output against your sources. The State Bar of California's AI resource page lists its practical guidance and ABA Formal Opinion 512; the California guidance maps AI use to competence, diligence and confidentiality duties. Verify applicable guidance with the official source in your jurisdiction.
Compare them on your task and data-handling requirements rather than assume an overall winner. Use the same nonconfidential source material and instructions, then check each output against that material. Review whether inputs are used for training and what retention settings apply. Consult the dated, rubric-graded runs on this hub, but test your own workflow before relying on a model.
Not by default; first review how the tool handles information and the rules applicable to your matter. The State Bar of California's guidance notes that AI products may use inputs to train or refine the model. Do not treat a no-training setting or removal of names as proof that disclosure is appropriate. Keep confidential details out until you have verified the relevant safeguards and official guidance.
Include the task, jurisdiction, permitted source material and output format, with instructions to flag uncertainty. For a contract drafting task, specify the clause's purpose and ask for assumptions to be identified for your review. Start with a tested prompt from this hub, use nonconfidential material while evaluating it, and verify the result rather than treating the prompt as a safeguard.
Start with client intake, contract drafting or case research, the tasks with tested prompts on this hub. Choose a bounded assignment, such as an intake outline or a draft clause, and decide what you will check before running it. Review the dated test run for that prompt, then evaluate your own output against the source material and your client's instructions.
Tested prompts and skills for your profession, with dates and limits. No sponsors, no popups.