Skip to content
CasePrompts
Build

Free local CITE Check

Check your litigation AI prompt structure.

Paste the instruction text you plan to use. Fixed browser-local rules mark the four CITE blocks and five review safeguards as present, thin, or missing—without calling a model provider.

Prompt text only—never paste case documents. What you paste stays in this browser; the check makes no model or provider request.

Paste a prompt above to check its structure.

What CITE Check looks for

C — Context

CITE Check looks for a stated role, matter, document, and issues in scope. It flags a missing or thin Context block. It does not determine whether the context is accurate, complete, or appropriate for the matter.

I — Instructions

It looks for a defined task instead of an open-ended request such as “find the important parts.” CITE Check can suggest concrete subtasks, but it does not know which issues are material or whether the instructions reflect your strategy.

T — Template

It looks for a requested output format—a table, timeline, or memo with fixed headings. A named format can make the response easier to review; it does not make the content complete, accurate, or ready to use.

E — Evidence

CITE Check looks for a source-location requirement and an instruction to report when support is not located. Evidence counts as one of four equally weighted structural blocks. The instruction cannot stop a model from inventing a citation or missing material, so verify the result against the source.

How to use CITE Check

  1. 01

    Paste your prompt

    Paste only the instruction text you want to check. The on-device check does not upload the prompt or call a model provider. If optional allowlisted product analytics are enabled, CasePrompts may record non-sensitive event metadata. Do not include case documents, privileged facts, or confidential client information.

  2. 02

    Check the structure

    Choose “Check structure” when the prompt is ready. The checker applies fixed, equally weighted rules to the four CITE blocks and separately reports five review safeguards. A single keyword stays thin; “present” requires multiple structural cues.

  3. 03

    Repair and check again

    Copy useful repair language, edit the prompt, and choose “Check again.” The scorecard measures structure rather than model performance, so independently verify any later output against the source.

Common questions

Is CITE Check free?
Yes. The on-device CITE Check is free, with no signup or card. AI Deep Review is a controlled cloud experiment and is not enabled on this deployment.
Does CITE Check see my case file?
There is no document-upload feature. The on-device check processes prompt text locally and does not send it to a model provider. If optional allowlisted product analytics are enabled, CasePrompts may record non-sensitive event metadata. Do not paste case documents, privileged facts, or confidential client information.
Does it check prompts for a specific AI tool?
The structural check is vendor-neutral and does not evaluate a specific model. A high score means the prompt contains the signals CITE Check looks for; it does not predict performance in Claude, ChatGPT, Copilot, CoCounsel, or any other tool.
Will a high score guarantee good output?
No. The score covers prompt structure only. It does not test the underlying record, model, product configuration, accuracy, completeness, confidentiality, or legal sufficiency. Verify every material statement and citation before relying on the output.
Do I need to know prompt engineering to use it?
No. The checker teaches the method as it runs — each flag explains what is missing and how to write it, so you improve by using it.

New to the framework?

CITE Check applies the CITE method — Context, Instructions, Template, Evidence. Read the full walkthrough, then bring your prompt back here to check its structure.

Read the CITE method →Browse the prompt library

Curious how the AI tools handle your data?See where the models run →