Introduction: When Language Models Pick Up the Pencil
The AI race is heating up—not just in logic or text, but in art. A recent experiment put leading language models to the test, asking them to “draw” the Mona Lisa using only text prompts and colored pencils, rendered in ASCII or step-by-step instructions. Why does this matter for AI creators, marketers, and entrepreneurs? Let’s break down what happened, and what you can learn from it.
How the Experiment Worked
The Challenge
A team at tryai.dev asked top LLMs (large language models) to recreate the world’s most famous painting—the Mona Lisa. But no image generators allowed: models had to use text alone, either as ASCII art or detailed drawing instructions for colored pencils.
The Contenders
Four of 2024’s leading AI models faced off:
- OpenAI GPT-5.6
- Anthropic Claude
- Google Gemini
- xAI Grok
Each model received the exact same prompt and instructions, so their “artistic” skills could be fairly compared.
The Results
Each AI gave us a unique Mona Lisa, showing different levels of detail, style, and interpretation. Some tried ASCII art, some offered drawing guides, and others focused on capturing the painting’s mood in words.
Why This Matters for Creators and Marketers
1. Visual Storytelling—Powered by Text
The experiment proves that modern LLMs can create not just words, but visual ideas through text. That means:
- Instantly illustrated instructions
- Quick sketches for presentations and MVPs
- Fast idea visualization for marketing campaigns
2. Creative Variety—Pick the Right Model for the Job
Each AI interprets the same prompt differently. As a creator, you can test multiple models to find the one that matches your style, audience, or accuracy needs.
3. Prototype Visuals—No Designer Needed
Need a banner, logo, or infographic idea fast? You don’t need an artist—simple text-based visuals can get your point across quickly to teams or clients.
4. Know the Limits
Even the best models don’t produce perfect “drawings.” Results are often creative interpretations, not replicas. Clear prompts help, but don’t expect photorealism—yet.
5. The Multimodal Trend
Models are rapidly learning to not just write, but also draw, narrate, and edit video. Staying on top of these trends keeps you one step ahead in the market.
Model Prompts and Responses: Highlights
GPT-5.6
- Generated detailed ASCII art, aiming to capture the face and background—though the result was abstract.
- Gave color guidance and suggested shading techniques.
Claude
- Provided a drawing plus advice on achieving the Mona Lisa’s “mysterious smile” with pencils.
- Reflected on the painting’s meaning beyond just the image.
Gemini
- Attempted ASCII art, but the outcome was very schematic.
- Focused on step-by-step tips for beginners on drawing faces.
Grok
- Leaned into humor and irony in its descriptions.
- Delivered the most “avant-garde” version—testers found it amusing.
How to Use This in Your Work
- AI Creators: Rapidly prototype visual ideas for websites, apps, or social posts using LLMs.
- Marketers: Craft unique visual stories—even without Photoshop skills—by developing strong prompts.
- Business Owners & Entrepreneurs: Experiment with different LLMs to find your visual style and save on early-stage design costs.
What to do right now
- Test your LLMs—Give them a creative visual prompt (e.g., “draw a cat in ASCII” or “describe how to sketch a coffee shop logo”).
- Compare outputs—Try the same prompt in several models to see which suits your needs.
- Draft prompts for your niche—Develop instructions for your typical visual tasks (guides, sketches, storytelling).
- Discuss with your team—Brainstorm how these tools could save time and speed up ideation.
- Track new multimodal features—Experiment early and often as new capabilities arrive.
More hands-on AI tools and templates — in my courses.
Source: Based on materials from Hacker News


