This walkthrough shows you an alternate way to achieve what Evaluators accomplishes in the Evaluating Grade Level Appropriateness tutorial for comparison.
Once you have a solid framework for evaluating texts, you can leverage a large language model like Gemini to apply that framework consistently and efficiently. This guide will walk you through how to use our prepared prompts to have Gemini perform the four-step text analysis.

​
Requirements

  • Gemini app at https://gemini.google.com/app and pick the model 2.5 Pro.
    Even if you don’t have pro subscription, you can try this model once a day.

​
The two-prompt process

The most effective way to guide Gemini is with a two-step prompting process:
  1. The Setup Prompt: This is where you give Gemini its instructions, including its persona, the process it must follow, and the format for its answer.
  2. The Content Prompt: This is where you provide the actual text you want Gemini to analyze.

​
Step 1: The Setup Prompt

Think of this first prompt as loading the software. You are telling Gemini how to do the job before you tell it what to do. Your setup prompt should clearly define the entire task from start to finish.

​
What to include in your setup prompt:

  1. Persona: Tell Gemini who it should be. For example: 
    You are an expert in English literature education for K-12.
    OR copy and paste the System prompt from the evaluator as described in the Prompts section of the Grade Level Appropriateness evaluator.
  2. The Process: Explicitly list the steps Gemini must take. This is where you insert the four-step analysis process you’ve created (Quantitative, Qualitative, Background Knowledge, Synthesis in the Traditional Method tutorial. OR copy and paste the User prompt from the evaluator as described in the Prompts section of the Grade Level Appropriateness evaluator.
  3. The Output Format: Be very specific about how you want the answer. Requesting a structured format like JSON is ideal because it makes the output predictable and easy to read. For example: 
    Your reasoning and answer need to be in JSON format. Strictly follow the following format... In your reasoning, provide numbered bullet points for each of the analysis in each of the 3 steps. At the end, give me the 4th bullet point called 'synthesis'...
    OR copy and paste the formatting instructions from the evaluator as described in the Prompts section of the Grade Level Appropriateness evaluator.
By combining these elements into one comprehensive setup prompt, you prepare Gemini to execute the task accurately as soon as it receives the text.

​
The content prompt

This prompt is much simpler. After you have sent the setup prompt, you simply provide the text you want to be evaluated. It’s best to clearly label the text to avoid any confusion. You can do this by starting your prompt with a clear marker, like: 
<begin of text to evaluate>

"Tom and his crowd looked down the path and saw two other new-comers approaching..."
Gemini will use the instructions from your first prompt to analyze the text in this second prompt.

​
Interpreting the Results

Because you requested a specific JSON format, Gemini’s output will be structured and easy to interpret. It should contain the key properties you asked for:
  • reasoning: This object will contain the detailed, step-by-step analysis you instructed Gemini to perform, giving you full transparency into its “thought process.”
  • grade: This provides the final, recommended grade level for independent reading.
  • alternative_grade: This gives you an option for using the text with a different grade level.
  • scaffolding_needed: This offers concrete, actionable strategies for how to support the students in the alternative grade level.
By using this structured prompting method, you can turn a complex analytical task into a simple, repeatable workflow, ensuring every text is evaluated against the same rigorous criteria.