When you create a custom prompt in Insights IA, the response might not be what you expected. In most cases, this is due to how the instruction is written. The good news: improving the wording is usually enough to get much more useful results.
To help you, we’ve created a checklist to guide you when reviewing your prompt, with clear examples and a simple structure. You don’t need to be an AI expert to apply them—just a few simple tweaks and your analyses will become more precise and relevant.
🧠 Clearly define the role and methodology
Provide details about a specific role. Instead of using generic roles like “social media manager,” offer more context.
👉 Example: “You are a senior social media analyst specialised in performance metrics.”
🪄 Useful tips
Specify the industry you’re working in.
Add years of experience and learning.
Include a specific approach or methodology (JTBD to review buyer personas, AIDA to analyse copy, Jungian archetypes for style and values, SWOT for benchmarks, etc.)
📋 Specify the task to be performed
Describe in detail what you need. Avoid instructions like “do an analysis.” Use precise action verbs that describe your analytical needs.
👉 Example: “Compare the ER by followers of videos versus carousels.”
🪄 Useful tips
Break complex tasks into simple steps.
Avoid vague and general phrases.
Replace subjective evaluations with objective data. For example, don’t ask for “the posts that were most liked” but rather for “the posts with more than 500 interactions and at least 2% ER.”
Clearly state the order of execution. For example: “first analyse the content by themes, then compare the result by formats, and finally create a table showing the differences by brand.”
🎯 Explain the purpose of the analysis
Sometimes results aren’t optimal because the purpose of the analysis is not well defined. Remember, it’s not the same to analyse for learning as it is for decision-making. Clearly state how you will use the result (to present to a client, for internal consultation, for strategic decisions, to identify improvements, to find market niches, etc.)
Clarify whether the objective is strategic or tactical.
👉 Example: “We want to optimise posting frequency based on engagement.”Connect it to a business objective.
👉 Example: “We need to prioritise content with higher virality to drive website traffic.”Avoid generic statements. Replace “to improve on social media” with “to identify which types of content generate the most interactions.” The more specific, the better.
📝 Specify the response format
Request the format in great detail. Always define whether you want a table, list, summary, etc., and also specify how you want it to appear visually.
👉 Example: “Present a table with one column per brand and one row per theme; include the number of interactions per brand and theme in the corresponding cell.”
🪄 Useful tips
Define the structure or blocks. Specify what sections the response should contain.
Indicate whether it should focus on insights, recommendations, comparisons, or quantitative data.
Adjust the level of depth. Mention if you only need H1s and headlines with short descriptions or technical details for an executive report.
🧱 Structure the prompt properly
Present your ideas in separate blocks and with line breaks. AI struggles more when everything is written in one paragraph. Avoid writing everything in one go.
Separate ideas with dashes and use line breaks or capital letters to help.
Make sure each block has a clear purpose.
Write a short final sentence in capital letters after the word “IMPORTANT” to highlight the main goal of the prompt.
And that’s the end of this guide to writing prompts smartly! Adapt these tips to your needs and you’ll see your results improve. If you need inspiration, check out our Insights AI prompt templates or contact the support team.