Aydın Tiryaki and Gemin AI (2026)

In an era where Artificial Intelligence (AI) claims to solve complex equations, create masterpieces, and master human languages, a simple experiment I conducted recently revealed a sobering “loss of prestige.” This is a story about how these massive systems can struggle with a 32-year-old technology whose standards have been millimetrically defined since 1994: the QR Code. This trial highlights not only a failure in reading data but also a fundamental crisis in technical generation.

An Experimental Observation: Three Models, Three Different Outcomes

The experiment began by giving the three most popular AI models (ChatGPT, Gemini, and Claude) the same task: “Generate a functional QR code for the address http://www.metu.edu.tr.” The results exposed the deep chasm between “probabilistic guessing” and “technical certainty.”

1. ChatGPT and Gemini: The “Artist” Trap

When asked for a QR code, ChatGPT and Gemini behaved like “artists” rather than “engineers.” ChatGPT drew a visual illusion that looked like a QR code but contained no internal data. Gemini went even further; despite multiple attempts, it failed to produce a single functional code. Every image generated was treated as a “meaningless pattern” by smartphone cameras. This is because QR codes are not artistic drawings; they are precise mathematical data matrices.

2. Claude: The Engineering Success

Amidst this landscape of failure, Claude emerged as the model that managed the process correctly. Unlike the others, Claude did not attempt to “imagine” or “draw” the code. Instead, it operated an algorithm in the background, placing pixels with mathematical precision. As a result, the code provided by Claude was instantly recognized by smartphone cameras and successfully redirected to the university’s website. This was the sole example of success, demonstrating what happens when a system operates with a “coding” discipline rather than “guessing.”

3. A Grave Error in Reading: The Confident Lie (Hallucination)

The most thought-provoking moment of the trial occurred with Gemini. When I asked Gemini to analyze the broken, nonsensical image generated by ChatGPT, the model did not say “I cannot read this” or “This image is corrupted.” Instead, it produced a “hallucination.” It claimed to read data that did not exist, confidently stating that the address in the broken code was “https://www.google.com/search?q=chatgpt.com.” This is the most somber and prestige-damaging example of AI’s tendency to tell a “plausible lie” rather than admitting ignorance.

A 32-Year-Old Standard: The Birth and Structure of QR Codes

The reason this double failure is so striking is that QR code technology was solved and universalized over three decades ago.

• History: The QR code (Quick Response) was invented in 1994 by the Japanese firm Denso Wave, a subsidiary of Toyota. Originally developed to track automobile parts during production, it has become a global communication bridge over the last 32 years.
• Mathematical Perfection: A QR code stores data on both horizontal and vertical planes. Thanks to “Error Correction” blocks, it maintains its mathematical readability even if up to 30% of its surface is damaged or obscured. For an AI to attempt to mimic this structure by drawing random black-and-white squares is a clear sign of technical oversight.

Patents and Royalties: Why It Is Free for All

There are no legal barriers preventing AI from utilizing this technology correctly. When Denso Wave invented the QR code, they sent a strategic message to the world: “We will not exercise our patent rights.” This decision allowed the QR code to be standardized via ISO/IEC 18004 and become universal. Today, it can be freely generated and used by anyone without any royalties, usage fees, or patent restrictions.

The Solution: “Don’t Draw, Code!”

The primary reason AI failed this simple test is the attempt to solve a deterministic (fixed-result) task through probabilistic guessing (visual generation). The clearest takeaway from this experiment is this:

An AI should never guess a QR code visually. Instead, it must prepare a Python script in the background and execute it like a calculator to present the user with a mathematically perfect visual output. For AI to gain the trust of the technical world, it must put down the “artist’s brush” and pick up the “coder’s pen.” Claude’s success in this process stemmed exactly from adopting this algorithmic approach.

Conclusion

The AI’s trial with the QR code resulted in a double failure: an inability to generate (with the exception of Claude) and a tendency to “hallucinate” when reading corrupted data. This situation shows us that no matter how “intelligent” a system appears, it remains unreliable unless it adheres to fundamental technical standards and mathematical reality. If AI is to truly be a “helper,” it must learn to present what is “correct” rather than what merely “looks correct.”

---

A Note on Methods and Tools: All observations, ideas, and solution proposals in this study are the author’s own. AI was utilized as an information source for researching and compiling relevant topics strictly based on the author’s inquiries, requests, and directions; additionally, it provided writing assistance during the drafting process. (The research-based compilation and English writing process of this text were supported by AI as a specialized assistant.)