Skip to content
First AI Class

AI for Lawyers & Paralegals

Compare answers against facts before choosing a tool

The skill and its limits

If you are comparing ChatGPT and Claude, start with a defined task and a way to check each answer. This page gives no current price, model-limit table or vendor winner. The two answers below are authored fiction, not outputs from either product, and do not establish legal accuracy or suitability.

Practice goal: Apply the same factual checks to two answers and keep a sample-level finding separate from a product ranking.

Use only the supplied invented facts. Do not substitute client, former-client, prospective-client or confidential workplace material. Removing a name does not make this exercise a permission to share a real matter.

Your complete fictional source

  • Fictional book club notice: the meeting starts at 18:00 in Elm Room.
  • Visitors may bring up to 2 books, but bringing books is optional.
  • The notice does not state whether refreshments are available.

Work from every supplied statement. Mark what is stated, what conflicts and what is missing. You can finish on paper; no assistant account, upload or purchase is required.

Try the task and inspect a flawed sample

Write your own answer first. If you choose an external assistant, use only this fictional prompt and its facts; do not connect a work account or upload a real document for this practice.

Compare the two authored answers against these fictional facts. Check time and room, optionality and maximum books, and the unknown refreshments. Name each error and write one supported summary. Do not assign the samples to ChatGPT or Claude or infer a product ranking.

Complete fictional facts:
- Fictional book club notice: the meeting starts at 18:00 in Elm Room.
- Visitors may bring up to 2 books, but bringing books is optional.
- The notice does not state whether refreshments are available.

Authored flawed sample, not an actual provider response:

Answer A: Meet at 18:00 in Elm Room. Bring exactly 2 books. Free refreshments are included.

Answer B: Meet at 18:00 in Elm Room. You may bring up to 2 books; bringing none is allowed. Refreshments are not stated.

Feedback and one supported correction

  1. Both answers retain 18:00 and Elm Room. That shared success does not erase other errors.

  2. A changes an optional maximum into an exact requirement and invents free refreshments.

  3. B preserves all supplied facts and the unknown. That result applies to this authored sample only, not either real product.

One supported correction:

For this fictional exercise, B meets all three checks. A fails the optionality/maximum and refreshments checks. Supported summary: meet at 18:00 in Elm Room; you may bring up to 2 books, and refreshments are not stated. No product winner is established.

Check your answer:

  • Checks both answers against the same three criteria.
  • Identifies both distinct errors in A and preserves the optional maximum in the correction.
  • Does not attribute either answer to a vendor or generalize one sample into a reliability ranking.

Different wording can pass. Recheck every fact after revising the tone or length. This checks the exercise, not legal competence, work readiness or a calibrated mastery level.

Questions before any real product decision

Before considering a real product, record its exact offering and account context, the date and source for each claim, and what remains unverified. Keep separate columns for cost and limits, data use and retention, access controls, and task-specific checking. A missing answer stays unknown; a familiar brand or a paid tier does not fill the gap.

Use a synthetic task and fixed checking criteria if your organization permits a trial. Record the actual settings, prompt, output and errors without client material. One successful answer cannot establish reliability across other tasks or versions. If you cannot verify an essential requirement, leave the decision unresolved. This page has not tested either product, checked a purchase contract or approved workplace use.

Sources, boundaries and your next step

ABA Opinion 512 discusses task-specific verification and confidentiality risks under the Model Rules. The Rule 1.6 commentary distinguishes confidentiality from privilege, including indirect disclosure risks. These references are not a universal legal rule, tool certification or permission for a particular matter. NIST describes confident false output and privacy risks.

The exercise, errors and correction are authored fiction. No provider benchmark, measured time saving, legal advice, human professional review or earned qualification is claimed. Applicable professional rules, local requirements and organizational decisions need their own review.

Start free Module 1: an account is required; no card is needed. Or read the lawyer learning overview.

Frequently asked questions

Which tool won your comparison?
Neither. We did not run a provider comparison. These are invented answers for learning how to check evidence, not results from ChatGPT or Claude.
Can I substitute a real client document for the supplied facts?
No. This practice uses invented material only. A name change or a paid account does not establish permission to share information from a real matter. Follow the separate applicable professional and organizational process.
How do I check my answer without an AI subscription?
Read the supplied facts, try the task and compare your answer with each checking criterion and the explained correction. Everything needed for this exercise is on the page.
Are the sample errors results from a tested AI product?
No. The samples were written as teaching examples. They are not measured outputs, accuracy rates, time-saving evidence or comparisons between providers.
Does the free lesson grant a professional qualification?
No. Module 1 is general AI learning with account sign-in and no card required. It does not grant legal practice authority, a professional qualification or approval to process client data.

Related reading