A solicitor asks an AI to draft a contract review, then gets a bland summary that misses key risks. The problem is not the tool, it is the vague prompt. In law, loose inputs lead to loose outputs, and that can cause real errors.

Legal prompts are clear instructions given to AI for tasks like drafting contracts, reviewing clauses, or researching cases. They set scope, facts, and format, so the AI knows what to do and what to avoid.

This post introduces the CARE method, a simple framework to make prompts precise: Context, Ask, Rules, Examples. You will see a quick breakdown, practical tips, and worked examples. Use it to save time, cut rework, and improve accuracy.

What is the CARE Method and Why Use It for Legal Prompts?

The CARE method is a simple structure for writing legal prompts that work. CARE stands for Context, Ask, Rules, and Examples. It helps you give the AI enough detail to act with precision, without guessing or padding. In law, where small errors have big costs, CARE reduces risk, speeds up workflows, and improves the quality of first drafts.

• Context: provide background that frames the task.
• Ask: state the request in clear terms.
• Rules: set boundaries for tone, sources, and length.
• Examples: show what good looks like.

Used together, these parts produce outputs that fit the facts, the jurisdiction, and the standard you expect. Vague prompts lead to generic answers, which is unsafe for contracts, disputes, or regulatory advice. CARE closes that gap.

Provide Clear Context to Set the Scene

Give the AI the who, what, where, and why. Add details that matter to the law you want applied, then stop. The goal is a full picture without fluff.

Include:

• Case type: employment, commercial, data protection, planning.
• Jurisdiction: England and Wales, Scotland, EU, or other.
• Key facts: dates, parties, sums, clauses, deadlines.
• Client background: sector, role, risk appetite, objectives.

Short legal example:

• “UK employment dispute, England and Wales. Employee with two years’ service alleges unfair dismissal after a redundancy process. No collective consultation. No settlement offer yet.”

This context lets the AI analyse without assumptions. Keep it concise yet complete. One or two tight paragraphs usually suffice.

State Your Ask Directly and Specifically

Tell the AI exactly what to do and how to present it. Use strong action verbs and define the output shape.

Helpful patterns:

• “Summarise this contract clause in 5 bullet points.”
• “List risks in this scenario under UK consumer law.”
• “Draft a 2‑paragraph email to the client explaining options.”
• “Identify missing warranties and suggest wording.”

In legal work, specify the law or precedent to consider:

• “Apply Employment Rights Act 1996 and recent EAT authority.”
• “Refer to CMA guidance on unfair terms.”

Avoid broad asks like “Review this” or “What do you think?” They invite off‑topic responses and weak conclusions.

Set Rules for Format, Tone, and Limits

Rules act as boundaries that keep the output fit for practice. Define tone, sources, citations, and size.

Examples of effective rules:

• “Use formal language suitable for client advice.”
• “Cite UK statutes and name relevant sections.”
• “Reference publicly available sources only.”
• “Keep under 500 words, with short paragraphs.”
• “Highlight assumptions and uncertainties.”

For sensitive work, add ethical notes:

• “Do not fabricate citations.”
• “Base analysis on the provided facts only.”
• “Flag when further research is needed.”

These constraints cut noise, prevent overreach, and improve trust in the result.

Include Examples to Guide the AI

Examples show style, depth, and structure. One good example is often enough. Two can help when the task varies by context.

How to use examples:

• Provide a short sample input and a model answer.
• Use anonymised facts from a past matter.
• Point out features you want copied, such as headings or risk grading.

Sample format:

• Input: “Share purchase agreement, English law. Identify indemnity gaps.”
• Output example: “Three bullets, each naming the gap, risk level, and fix. Cite Companies Act sections if relevant.”

Examples act like a template. They reduce guesswork and raise consistency across drafts and matters.

Key Tips to Apply the CARE Method in Legal Work

CARE turns vague instructions into precise legal outputs. Start small, practice on low‑risk tasks, and tighten your prompts over a few iterations. Treat AI as a junior colleague, clear on context and rules, and always double‑check its work against law and facts.

Start with Basic Prompts and Build Up

Begin with the core CARE pieces, then layer detail. This helps you see what the model understands before you add complexity.

A simple path:

1. Set a short context.
2. Make one clear ask.
3. Add one or two rules.
4. Provide a quick example or template.

Practical starter tasks:

• Routine emails to clients or opponents.
• Neutral summaries of meeting notes.
• Basic clause comparisons against a house style.

Example prompt using CARE on a low‑risk email:

• Context: “England and Wales, commercial matter. Client seeks a short update on a missed delivery deadline.”
• Ask: “Draft a 120‑word client email with next steps.”
• Rules: “Plain English, no legal advice, bullet points for actions.”
• Examples: “Follow this tone: concise, calm, practical.”

Build up from there:

• Add statutes or guidance once the base works.
• Specify structure, such as headings or risk tiers.
• Paste short extracts before whole contracts.

Tip: keep iterations short. Try one change at a time, such as adding word limits or naming the jurisdiction. This reduces noise and gives you control.

Tailor Prompts to Your Legal Niche

The more your prompt reflects your practice area, the better the output. Name the jurisdiction, the statute, and the outcome you need.

Ideas by niche:

• Family law: “Apply Matrimonial Causes Act 1973 and FPR. Focus on needs and conduct. England and Wales.”
• Corporate compliance: “Assess policy gaps against the Companies Act 2006, Bribery Act 2010, and SMCR expectations.”
• Data protection: “Summarise risks under the Data Protection Act 2018 and UK GDPR. Flag lawful basis, data minimisation, and DPIA triggers.”

Targeted example for data protection:

• Context: “UK tech SME, B2C app. Processes email, location, and usage statistics.”
• Ask: “List high‑risk processing points and practical mitigations.”
• Rules: “Cite DPA 2018 principles where relevant. No speculation. Keep to 200 words.”
• Examples: “Output like: ‘Issue, risk, mitigation’ per line.”

Add concrete signals:

• Acts and guidance: name the statute, key principles, or regulator focus.
• Sector: healthcare, fintech, public sector.
• Client priorities: cost control, speed, low litigation risk.
• House style: headings you use in advice notes, such as Facts, Issues, Analysis, Next steps.

This level of context prevents generic answers and reduces rework.

Always Review and Refine AI Responses

Treat AI outputs as drafts. You are responsible for accuracy and ethics.

Review checks:

• Law: are statutes, sections, and tests correct for the jurisdiction and date?
• Facts: has the output added assumptions or missed key details?
• Bias: watch for loaded language, unfair stereotypes, or skewed risk scoring.
• Omissions: look for missing defences, exceptions, or counter‑arguments.
• Confidentiality: remove or anonymise client data where needed.

If the draft falls short, re‑prompt with CARE adjustments:

• Context: restate jurisdiction, parties, dates, and contract type.
• Ask: narrow the task, such as “identify only termination risks.”
• Rules: set citation style, word limits, and “no invented sources.”
• Examples: paste a short model answer from a past matter.

Helpful re‑prompt pattern:

• “You missed s.13 ERA 1996 on unlawful deductions. Revise the analysis to include it. Keep to 150 words, bullet points only, England and Wales.”

Ethics and practice hygiene:

• Verify against primary sources before relying on output.
• Keep a short audit note of what you checked and what you changed.
• For data protection work, align advice with DPA 2018 principles, such as lawful basis, data minimisation, and storage limits.

This cycle of review and refinement protects clients and builds trust in your workflow.

Examples of CARE Prompts for Common Legal Tasks

Use these worked CARE prompts to set clear tasks, keep outputs consistent, and cut rework. Each example shows the four parts in action, then adds a short note on what a good response should look like.

Example: Reviewing a Contract Clause

CARE prompt to analyse a non-compete clause in a UK employment contract.

• Context:
  • Role: Senior account manager with key client contact, basic salary £85,000.
  • Company: UK-based B2B SaaS provider, national client base, sales cycles of 6 to 12 months.
  • Contract: England and Wales, employer seeks to protect client connections and confidential information.
• Ask:
  • Identify enforceability risks in the non-compete clause.
  • Assess reasonableness of duration, geography, and scope of restricted activities.
  • Flag drafting defects and suggest fixes.
  • Cite UK authorities where relevant.
• Rules:
  • Formal tone suitable for a client advisory note.
  • England and Wales law, restraint of trade principles.
  • Refer to cases such as Tillman v Egon Zehnder [2019] UKSC 32 and core tests on legitimate business interest and reasonableness.
  • Use headings: Summary risk rating, Issues, Analysis, Suggested wording, Next steps.
  • Keep to 300 words.
• Examples:
  • Clause snippet: “For 12 months after termination, the Employee shall not be employed by, engaged with, or interested in any business which competes with the Employer anywhere in the United Kingdom.”
  • Model analysis style:
    • Summary risk rating: High.
    • Issues: Nationwide scope, 12-month duration, “interested in” captures minor shareholdings.
    • Fix: Limit to named competitors or client segments, 6 months, carve out passive shareholdings up to 5 percent.

Why this works: The context ties the restriction to real interests, the ask focuses on enforceability, the rules anchor the legal tests and tone, and the example shows depth and format. Together, you get a precise, case-backed risk view, not a generic checklist.

Expected output: A concise, case-cited risk analysis with targeted edits, such as narrowing to clients the employee dealt with in the past 12 months, a 6-month term, and a passive investment carve-out, with reference to Tillman on severance and “interested in” wording.

Example: Summarising a Court Case

CARE prompt to summarise a UK Supreme Court ruling on privacy.

• Context:
  • Case: Lloyd v Google LLC [2021] UKSC 50.
  • Issue: Whether a representative action under CPR 19.6 could recover uniform “loss of control” damages for alleged breaches of the Data Protection Act 1998 without proof of individual damage.
  • Relevance: Collective privacy claims, class action strategy, damages under UK law.
• Ask:
  • Summarise the key points and holding in 200 words.
  • Focus on procedural route, test for “same interest,” and approach to damages.
  • Note practical impact on privacy class actions.
• Rules:
  • Neutral tone.
  • Bullet format, each bullet one sentence.
  • England and Wales law.
  • No speculation, cite the case name once at the start.
• Examples:
  • Sample output style:
    • Lloyd v Google LLC [2021] UKSC 50 considered a representative claim for data protection breaches.
    • The Court held that the class did not share the same interest for damages without individual assessment.
    • Uniform damages for “loss of control” were not available under the DPA 1998 absent proof of material damage or distress.
    • The decision limits opt-out privacy claims that cannot show common issues on loss and causation.
    • Claimants may still seek a bifurcated route, certification on liability issues first, then individualised damages.

How CARE keeps focus: The context narrows the legal frame, the ask sets a tight length and scope, the rules force neutral, scannable bullets, and the example demonstrates the tone and level. The model cannot drift into commentary or outdated doctrine.

Expected output: Five to seven crisp bullets, under 200 words, covering the holding, reasoning on “same interest,” damages approach, and practical impact on privacy litigation strategy.

Example: Checking Compliance with Regulations

CARE prompt to verify GDPR compliance in a data policy.

• Context:
  • Business: UK ecommerce retailer selling direct to consumers, stores customer names, emails, order history, and browsing data for marketing analytics.
  • Audience: Internal compliance review ahead of a website relaunch.
  • Scope: Privacy notice and internal data retention policy, UK GDPR and Data Protection Act 2018.
• Ask:
  • List compliance gaps and risks against UK GDPR principles and duties.
  • Prioritise gaps High, Medium, Low.
  • Provide a compliant rewrite for one policy snippet.
  • Recommend practical fixes and owner actions.
• Rules:
  • Cite UK GDPR Articles where relevant, such as Art 5 (principles), Art 6 (lawful basis), Art 13 (transparency), Art 21 (right to object), Art 30 (records), Art 32 (security), Art 35 (DPIA).
  • Plain English, 250 words limit.
  • Output structure: Findings, Compliant wording, Next steps.
  • No legal jargon beyond article references.
• Examples:
  • Policy snippet: “We keep your data for as long as needed and share it with partners to improve our services.”
  • Compliant version example:
    • “Retention: We keep account data for 6 years after your last order to meet tax and accounting duties, then delete or anonymise it.”
    • “Sharing: We share email addresses with our email provider to send order updates. We have a data processing agreement and do not allow use for other purposes.”

Method’s role in precision: Context ties the assessment to ecommerce realities, the ask demands gap spotting and fixes, the rules anchor the review in UK GDPR Articles, and the examples show what a compliant line looks like. You get practical, article-linked edits, not theory.

Expected output: A short gap list with article citations, a tightened policy sentence pair that names lawful basis and retention, and three action items, such as adding an Art 21 opt-out line for direct marketing, recording processing in Art 30 logs, and setting defined retention schedules by data category.

Conclusion

The CARE method helps turn vague legal prompts into precise instructions that produce reliable first drafts. By setting clear Context, a direct Ask, firm Rules, and practical Examples, you cut generic output, save time, and reduce mistakes.

Put it to work now. Try one CARE prompt this week on a low‑risk task, then refine and reuse the pattern for your practice area. Share what worked and what you changed, and keep building a prompt library that raises accuracy and efficiency. Thank you for reading.