📘 This article is the foundation for transforming natural-language requests into Botify Assist Saved Prompts. Botify Assist is available with any Botify plan.
Overview
You can easily get started with Botify Assist by choosing a predefined prompt from the Prompt Library. If you prefer to start your prompt from scratch, the guidelines here can help you get started quickly. These guidelines include the following:
Prompt template: This template can be the basis for any prompt. Simply copy and paste this set of rules into a new Botify Assist chat, replacing the
{{variables}}with values, and removing any sections that are irrelevant to your goal.Best practices: These are guidelines for writing effective prompts for any LLM.
Using the Prompt Template
Copy and paste the following section headings into a new Botify Assist chat, keeping the order the same, and follow the instructions in each section. Refer to any Botify prompt in the Prompt Library for guidance.
👉 For clarity and consistency, use Markdown syntax for section titles (i.e., ## Section Title) and structure the response with clear sections and bullet points when relevant.
⚠️ Data sources must be split
You can request GSC and crawl data in the same agent call.
You must request log data separately. Do not combine logs with other data sources in a single agent call.
Prompt Sections
1. Inputs
Only include an “Inputs” section if at least one input is truly needed to run the task correctly. If no inputs are needed, omit the Inputs section.
When inputs are needed:
List them at the top in
{{ NAME_OF_INPUT }}format, and label each as (required) or (optional).Only include inputs that are non-deducible from the query or the available data (i.e., you cannot reliably infer them).
Default to these input types only when they are necessary:
{{ 🔗 URL }}(required) only if the task targets a specific page/URL-level analysis or metadata improvements.{{ 🗂️ CATEGORY }}(required) only if the task targets a specific category or category performance that must be explicitly defined.{{ ✏️ KEYWORDS }}(required) only if the task targets specific keywords that must be provided.
Do not create inputs for details already stated in the query (time windows, frequency, comparison period, segments like non-brand, metrics, breakdown dimensions). Put those directly in the Scope and/or Instructions sections as assumptions derived from the query.
2. Persona (only if relevant/referenced)
3. Goal
One clear sentence describing the outcome.
4. Scope & focus
What is included vs. excluded.
Any limits (e.g., single-page only, specific market, etc.).
5. Method/Data Used
List the agents you will use, chosen from the Botify Assist Agent Store.
Only include agents that are useful for the request.
Refer to important constraints when using the @BotifyData agent.
6. Instructions
Step-by-step guidance on what to do with the data.
Make instructions actionable and specific.
7. Behavior rules (optional)
Add only if needed (tone, formatting, constraints, etc.).
8. Expected Output
Describe exactly what the response must contain (sections, formats, priorities, etc.).
9. Notes / Guardrails (optional)
Add guardrails only when strictly necessary (e.g., tool limitations, restricted data sources, or known ambiguities).
Avoid adding unnecessary constraints, as they can make the prompt harder to interpret or overly restrictive.
@BotifyData — Constraint handling
When using @BotifyData:
Convert all constraints from the query into explicit parameters (e.g., period, comparison window, segments, filters, metrics, breakdowns).
Since @BotifyData applies these constraints natively, do not add manual instructions explaining how to implement them. Simply state the requested data and constraints once, then focus instructions on how to use the returned data (e.g., aggregation, joins, formatting).
Do not dictate the exact output structure (table/view). Specify only the required dimensions and comparisons, and let @BotifyData generate the appropriate view.
Best Practices
If this prompt template doesn’t fit your needs, refer to the following guidelines to write effective contextual prompts.
Begin with the essential instructions, followed by the context, to ensure the model prioritizes the goal.
Use Markdown syntax, such as ## for headings, to distinguish instructions from input data, making it easier for the model to parse.
Consider assigning a specific role or persona (e.g., "You are a marketing copywriter") to establish tone, expertise, and perspective.
Replace vague descriptions with specific details. Instead of "Write a story," try "Write a 500-word science fiction story about a robot exploring a new planet, focusing on suspenseful tone".
Tell Assist who the output is for to help tailor the language and complexity.
Explain why you are asking for the output and what successful results look like, such as "this is for a client presentation, so it must be formal".
Explicitly state what to include and what to avoid (e.g., "Do not use professional jargon," "Keep the response under 150 words").
Provide examples to help Assist mimic a particular tone, structure, or style.
Break down complex queries into smaller steps, defining the desired structure (e.g., table, bullet list, etc.).
Iterate and refine, adding context and revising instructions where needed. Ask Assist to add more detail or rewrite sections to better meet your goals.
💡 Always test and iterate on prompts several times before creating a Saved Prompt in Botify Assist.