Generative AI has moved from a novelty to a daily study habit for many Georgian university students — but not all students use it the same way, and lecturers are still working out how to respond. A new research report from the Business and Technology University (BTU) looks closely at both sides of the classroom: how students use tools like ChatGPT, Gemini, and Claude, and how lecturers are experimenting with AI in their teaching.
The findings come from two sources: a quantitative survey of 167 BA and MA students at BTU (from finance, management, digital marketing, and IT programs), conducted in June-July 2025, plus interviews, surveys, and informal conversations with lecturers.
How Are Students Actually Using AI?
The starting point matters. Students who began using generative AI earlier are today the most frequent users — the study found a statistically significant link between when a student started using AI and how often they use it now (rho=0.19, p=0.015). Among students who adopted AI in 2022 or earlier, 60% are now active users, compared to 58% of those who started in 2023 and just 41% of those who began in 2024 or later.
Almost everyone uses AI for the basics — finding materials and understanding topics. But active users go further: idea generation, editing and drafting papers, summarizing texts, and even solving math problems. The report groups these behaviors into four categories: basic tasks (long-established, still common), fast-growing tasks (recently adopted but already widespread, like explaining concepts), growing tasks (new and not yet mainstream), and niche tasks (early but limited).
57% of students say they use AI mainly to save time, and a similarly popular motivation is improving their learning process. Encouragingly, students who use AI for time-saving don’t skip verification — they tend to check AI-generated answers more often, suggesting a fairly balanced approach.
Do Students Trust What AI Tells Them?
Most students check AI-generated answers at least sometimes. But students who check least often turn out to be the most confident in AI’s accuracy — a weak but statistically significant negative correlation was found between verification frequency and confidence (rho=-0.16, p=0.037). Students who rarely verify AI outputs rated their confidence at 3.78 out of 5, compared to 3.38 among those who always verify. In other words, the students least equipped to catch AI’s mistakes are the ones most likely to accept them without question.
Which AI Tools Do Students Actually Use?
ChatGPT remains the default tool — nearly every surveyed student uses it for studying. But newer tools are gaining ground: roughly 2 in 5 students also use Gemini, and 1 in 5 use Claude. Adoption of these alternatives is recent — 62% of Gemini users and 70% of Claude users started using them for studying only recently, suggesting real momentum behind tool diversification.
What Frustrates Students Most?
Students’ complaints are less about their ability to use AI and more about the quality of what they get back. The most cited problems: poor communication in Georgian, weak performance on math and finance topics, fabricated information or sources, overly long or off-target answers, contradictory responses, and limited access to premium features.
What Challenges Do Lecturers Face?
For lecturers, the most pressing short-term problem is assessment. Many struggle to evaluate student work objectively when a growing share is AI-generated — original thinking is harder to find, verification is exhausting, and AI-assisted writing often produces unnatural phrasing. The report concludes that Georgian universities need a deeper rethink of teaching and assessment practices, not just new rules.
What New Teaching Approaches Are Emerging?
Lecturers across BTU are experimenting with practical, often low-cost approaches:
- NotebookLM to turn lecture slides and materials into structured summaries and audio/video overviews — ideal for flipped-classroom formats.
- Custom GPT Builder chatbots trained on course materials, giving students an always-available assistant without requiring any coding.
- AI presentation tools like Gamma, Tome, and Canva Magic Design to quickly turn lecture notes into polished slides.
- AI image generation tools (GPT-4o, Midjourney) and Canva to visualize abstract concepts — though these still struggle with Georgian-language text.
- AI-generated test questions and quizzes drawn directly from course materials to save time on assessment design.
- AI-generated learning materials for subjects where no adequate Georgian-language textbook exists.
- Comparative AI ethics exercises where lecturers run the same prompt through ChatGPT, Claude, and Gemini to build critical digital literacy.
The Bottom Line
Generative AI is no longer optional in Georgian higher education. The open questions now are less about whether to use AI and more about how to use it well: building trust calibrated to actual accuracy, developing tools that work in Georgian, and rethinking assessment for a classroom where AI assistance is the norm.
This article summarizes findings from the Digital Ecosystem Digest — Summer 2025 report, ‘Integration of Generative Artificial Intelligence in Teaching and Learning: A University Practice Example from Georgia,’ published by the Business and Technology University. Read the full report here.
Keywords: Generative AI, Higher Education, Georgia, ChatGPT, Gemini, Claude, Student Behavior, AI Literacy, Trust in AI, Teaching Innovation