The rise of large language models (LLMs) like ChatGPT has transformed how students learn to code, particularly when it comes to tackling database queries with SQL. Traditionally, students would hit the web, scouring Stack Overflow and coding blogs for snippets and solutions. But now, they have a powerful new tool at their fingertips. This study explored how students use LLMs versus traditional web search when writing SQL queries. Researchers observed students in a real-world database course, comparing their interactions with standard ChatGPT, a specially tuned instructor LLM (designed to avoid giving direct answers), and good old web search. Surprisingly, the study revealed that students using the instructor-tuned LLM interacted far more with the tool than those using regular ChatGPT or web search. Why? Perhaps the guardrails and prompts encouraged a more conversational, back-and-forth approach. Interestingly, there wasn't a significant difference in the quality of the final queries across the three groups. Though the LLM users edged out the web searchers slightly in terms of correctness. What's even more intriguing is that students using the instructor-tuned LLM reported a lower mental burden. This suggests that while interacting more, the tailored guidance made the process feel less taxing. This research offers a glimpse into how AI is changing the learning landscape. While concerns linger about over-reliance and potential for misuse, LLMs like ChatGPT clearly offer a powerful new way for students to grapple with coding challenges. As these tools evolve and instructors find creative ways to integrate them, we can expect even more engaging and personalized learning experiences. It remains to be seen how these technologies will influence the future of education, but it’s clear that students don't seem to mind having AI tutors in the coding dojo.