What We Actually Tell the AI

This is the exact prompt we send to GPT-4o. It takes the raw form inputs and turns them into clear, natural messages your team would be comfortable sending.

Role Instructioning → Sets the assistant’s identity as a product comparison bot that helps users make quick purchase decisions based on needs.

‍Zero Shot Prompting → Tells the assistant to recommend one product without examples, relying entirely on the described user need.

‍Input Reference → Supplies product data including name, specs, reviews, and price for evaluation and comparison.

‍Selection Rule Enforcement → Requires the assistant to choose only one product, clearly stating why it is the best fit based on user criteria.

‍Output Constraints → Forbids any lists, alternatives, or extra commentary. Response must include only the selected product name and one line of justification.

Role Instructioning → Sets the assistant’s identity as a news content generator to define its role and tone of response.

‍Zero Shot Prompting → Instructs the assistant to create a concise subheadline using only the given input, without examples.

‍Input Reference → Provides the headline, title, snippet, and URL for each news item to inform subheadline generation.

‍Output Constraints → Requires the assistant to return only the subheadline per article. No extra text, explanation, or formatting is allowed.

Role Instructioning → Sets the assistant’s identity as a social media engagement expert to establish purpose and evaluation tone.

‍Zero Shot Prompting → Instructs the assistant to evaluate each news article’s viral potential using specified weighted criteria without example-based training.

‍Input Reference → Provides title, snippet, and URL for each article to guide scoring.

‍Output Constraints → Requires returning only the Viral Score as a number from 0 to 100 per article. No explanation, extra text, or formatting is allowed.