Sunday evening, 9 PM. A teacher sits down at the computer. Tomorrow's lesson plan needs preparing, a quiz for 7th graders needs building, feedback on last week's homework needs writing. The familiar scenario: at least two hours.

But this time it's different. The teacher opens Madlen, selects the learning objective, enters a few parameters. Three minutes later the plan is ready, the quiz is ready, the feedback draft is ready. The rest of the time goes into thinking about a "hook" for tomorrow's lesson. The teacher works out how to capture the students' attention.

This scene became part of the daily routine of 18,566 teachers on our platform between October 2024 and April 2026. And for the first time, we pulled together 18 months of this data and published "Designing the Future: The AI Adoption Report of Teachers in Türkiye."

As we prepared the report, our goal wasn't just to line up big numbers. What we really wanted to know was this: how are teachers actually using AI? What are they changing, what are they preserving? Where do they trust it, where do they resist it?

Here is what we found.

How much time does AI save teachers?

308,062 hours. The first reaction when you hear it is "wow." But what excites us isn't the number itself, it's where that time flows.

How did we calculate it? We took as our baseline the processes that take 60 minutes on average to create a podcast with traditional methods, 45 minutes to prepare a comprehensive rubric, and 10 minutes per student for manual essay assessment. For every operation our platform completes in seconds, the difference between the traditional duration and the AI processing time was counted as "time saved."

And where does that time go? Our data paints an interesting picture. 38% of content production happens in the afternoon, right during school hours. Activity intensity peaks between 10:00 and 15:00. In other words, our teachers use us not for lesson preparation at home but right in the middle of the classroom. A quick quiz between lessons, tomorrow's plan during the break, a podcast draft at lunch.

According to the OECD Digital Education Outlook 2026 report, teacher-focused AI tools can reduce preparation time by up to 30%. Our data gives this finding concrete shape on a Türkiye scale. But rather than the saving itself, we look at the opening that saving creates. That opening turns into the teacher making eye contact with a student, asking "how are you today?" in the hallway, drawing that explanatory diagram on the board.

425,669 questions generated, 98% assigned to students untouched

This data made us both proud and thoughtful.

Of the 238,281 questions generated and assigned on our platform, 233,702 were sent directly to students by teachers without a single comma altered. The ratio: 98.08%.

The Stanford SCALE report "The Evidence Base on AI in K-12: A 2026 Review" shows that AI tools designed specifically for education produce more consistent results than general-purpose chatbots. Our question engine isn't a general text generator. It's a purpose-built tool that operates according to learning objectives, difficulty levels, and curriculum frameworks. The 98% rate shows that teachers feel this difference.

But the same data raises a question: does this trust come together with AI literacy? Are our teachers equipped to question the outputs and intervene when necessary? We didn't want to ignore this tension. In our report, we state clearly that AI literacy isn't a future competency but a skill of the present.

How is the AI assistant used in the classroom?

When we examined 25,843 sessions of Madlen Assistant, four distinct usage profiles emerged.

Half of the sessions are short interactions: 3-6 message conversations that close after a few turns. The teacher asks during the break, gets the answer, moves on. One out of every five sessions is a single-shot command: a one-sentence request, instantly generated. These two profiles add up to 70%.

But what really excites us is the remaining 30%. One out of every four sessions turns into a 7-14 message dialogue. And 1,809 sessions exceed 15 messages. These teachers use the assistant not like an encyclopedia but like a colleague thinking alongside them. They design the unit together, developing the content layer by layer.

According to the OECD, combining education-focused AI tools with teacher expertise produces results neither could achieve alone. That 30% "thinking together" profile shows how this model works in practice in Türkiye.

Teachers in Türkiye think about ethics first when it comes to AI

Perhaps the most striking finding of the report came from our webinars. 22 webinars, more than 5,000 participants. And when these participants thought of AI in the classroom, the first need that came to their minds wasn't "making work easier."

It was preserving human values.

Primary priority: teaching students critical thinking and information verification skills. A significant portion view safe-use skills as an inseparable part of the curriculum.

According to the UNESCO "AI and the Future of Education" (2025) report, more than half of academics feel uncertain about the effective pedagogical application of AI. The "curious but cautious" profile we observed in our participants matches this global trend. Our difference is this: we've turned that uncertainty into concrete training programs.

Read the full report

All data in this report was drawn from Madlen's usage logs between October 2024 and April 2026. The data was pulled directly from the platform database and doesn't rely on surveys or self-reporting. The report doesn't claim representativeness; it reflects the profile of actively using schools. We openly share our limitations, methodology notes, and external source references inside the report.

→ Visit our resources to download the full report

→ Request a free demo to explore Madlen

To cite: Madlen. (2026). The AI Adoption Report of Teachers in Türkiye. Istanbul.