How an AI Poem Generator Took on a Life of Its Own Beyond Expectations

From Calculator Sites to AI Poetry: A Bold Experiment

For years, developers Jared Bauman and I have collaborated on AI projects, including a podcast dissecting the success of calculator sites—simple web tools that perform basic conversions or calculations, such as title case formatting or mortgage estimates. These sites often generate staggering traffic and revenue, sometimes exceeding $10,000 per month, despite their minimal codebase.

With the rise of generative AI, we wondered: What if these calculator-style sites could produce paragraphs of creative text instead of just numbers? The idea of an AI-powered poem generator was born.

Building the AI Poem Generator in Under an Hour

To test the concept, we registered PoemAIGenerator.com, a domain chosen for its keyword relevance. Using ChatGPT and OpenAI’s Assistants platform, we created a custom LLM-powered tool tailored for poetry generation. The platform allowed us to build a version of ChatGPT specifically for this use case, accessible via API for seamless integration.

Our primary concern was misuse. To prevent abuse, such as hacking attempts, we embedded strict safety rules into the system. The AI was programmed to:

• Refuse to generate offensive or inappropriate content
• Maintain a G-rated output
• Block requests for personal information or targeted harassment

Adapting to Any Poetry Style with Precision

The AI poem generator was designed to adapt to diverse poetic demands. Whether users requested:

• A haiku (requiring exactly 17 syllables)
• An iambic pentameter in the style of Maya Angelou
• Or any other structured poetic form

The system would comply, demonstrating the versatility of modern LLMs. The experiment proved that poetry, often considered a uniquely human endeavor, could be replicated—and even enhanced—by AI.

Unexpected Outcomes and User Behavior

Despite our precautions, the AI poem generator took on a life of its own. Users pushed the boundaries of its capabilities, exploring creative and sometimes surprising applications. The project highlighted both the potential and the unpredictability of generative AI in real-world scenarios.