It’s been well over a year since ChatGPT was introduced, yet EdWeek reports that only one-third of K-12 teachers say they had used artificial intelligence-driven tools in their classroom by the end of 2023.

But it doesn’t take into account the AI tools teachers are exploring outside of the classroom.

As the creator of Tom’s Takes: AI Tools and Views, a popular website dedicated to curating AI resources for educators, I have a unique vantage point on what types of AI tools teachers are actively seeking out. 

I say that because Tom’s Takes: AI Tools and Views consistently appears in the top 5 (non-sponsored) Google Search results for queries related to AI tools and K12 teaching, such as:

• AI tools for history teachers (average position 1.6)

• AI tools for math teachers (average position 1.9)

• AI grading tools for teachers (average position 2.2)

• AI for spanish teachers (average position 2.7)

• AI for math teachers (average position 4.5)

• AI tools for science teachers (average position 4.7)

• AI for science teachers (average position 4.7)

Google Search accounts for about 65% of my website traffic and direct access accounts for another 24%. Together with other traffic sources, my website registered 5.3K pageviews last month.

“AI grading” is the Google search query that drives the most traffic to my website. Across the 10 most popular search queries, the term "AI grading" appears in 4 of them, accounting for over 1000 clicks in recent weeks. “AI Tools for Teacher Grading” has been the most popular page at my website for months now and “AI Tools for Essay Feedback,” a related page, is the 4th most popular page.

That AI grading is such a popular query doesn’t surprise me because it aligns with what I see as a broad swath of teachers seeking to leverage AI to automate time-consuming responsibilities outside the classroom.

An EdWeek study in 2022 found that teachers spend only 46% of their in-school time actually instructing students, with the remaining hours devoted to planning, grading, communication, and record-keeping. AI presents an opportunity to offload some of these non-teaching duties, potentially freeing up more time for meaningful student engagement.

“AI for math teachers” is the second most popular Google search query driving traffic to my site and accounts for only slightly fewer clicks than “AI grading”. That does surprise me. ChatGPT has faced ongoing criticism for its perceived inability to handle complex mathematical operations – and sometimes basic ones. I’m also suprised because it’s so easy for students to find AI to get answers to their math homework, complete with steps. I figured many math teachers would simply stay clear of AI tools. 

That said, recent advancements like AlphaGeometry and InternLM-Math have demonstrated remarkable progress in AI's math reasoning capabilities. AlphaGeometry, a Google AI system, is now capable of solving complex geometry problems at a level approaching a human Olympiad gold-medalist. InternLM-Math, a recently introduced math reasoning large language model (LLM), achieves “state-of-the-art performance” in informal and formal math reasoning benchmarks. 

One potential explanation for teachers' curiosity about AI math tools could be the increasing availability of applications that guide students through problem-solving processes, such as Khanmigo and Sizzle. (Khanmigo, for one, has received extensive press coverage.) These intelligent tutoring systems can serve as AI tutors and dissuade students from immediately seeking an answer. In other words, they prompt students to think through math problems.

But I am perplexed as to why “AI tools for math teachers” drives so much more traffic to my website than “AI tools for science teachers”. The “AI tools for science teachers” query generates only 10% of the traffic that “AI tools for math teachers” does, though my website appears in the first 5 Google Search results for both queries. Moreover, there are constant news stories extolling the benefits of AI in science research. For instance, Science magazine recently proclaims that “AI is transforming how science is done” and “even transforming the nature of science itself.” 

Are math teachers simply more willing to use AI than science teachers?

I’m also a bit surprised that my “AI Tools for Lesson Planning” page receives a lot less traffic than my “AI Tools for Teacher Grading” page. Last year, "AI lesson planning" was a hot query driving substantial traffic to my site, with educators finding my reviews of tools like Eduaide and MagicSchool. Now, “AI lesson planning” is no longer a top query. A telling point might have been when Google searches for “eduaide” and “magicschool” were driving traffic to my website. It’s likely that early AI adopters in schools soon discovered eduaide, magicschool, and other lesson-plan generators, and they spread the word to other educators. Google Trends reports that “magicschool” is now the second most popular Google search query under the topic of “AI tools for teachers.” 

MagicSchool and Eduaide are popular because they’re impressive AI-powered "teacher assistants" that provide a myriad of tools in addition to lesson planning. They essentially serve as a Swiss army knife of tools, all accessible from a single central location, with the primary purpose of streamlining teaching tasks for teachers. So, teachers can find much of what they want to authomate their teaching in one place.

But, the issue here is that AI lesson plan generators are designed to assist teachers, not students. Students typically don’t interact directly with teacher-assistant AI (though MagicSchool has just introduced a student service). 

AI may be valuable for automating non-teaching tasks, but its true potential lies in its ability to personalize education by adapting to each student's strengths, weaknesses, and learning preferences in real-time. The most exciting promise of AI lies in its ability to offer a tailored educational experience, catering to the diverse needs of our students.

But I see little evidence that personalized student tools are at the top of a teacher's AI wish list. My “AI Tools for Personalized Learning” page is visited so infrequently that I haven’t bothered to update it in months.

It’s striking, because the latest language model updates are multimodal, equipped chatbots with voice, vision, and memory capabilities so they can carry on in-depth discussions, role-play with users, adjust instructional methods, and cater to diverse user needs.

All of this potential remains untapped if AI remains solely in the hands of teachers. According to the aforementioned EdWeek report, ChatGPT was the only AI identified by a majority of the teachers using AI in the classroom.

But, there’s hope. The EdWeek article reports that 7% of respondents have introduced Khanmigo in their classroom. And others are exploring multimodal tutoring. That’s a start.

Clearly, my website data only provides a partial picture of teacher AI tools preferences which leaves plenty of room for speculation. But I believe it does demonstrate the allure using AI to automate grading and feedback tasks. Teachers are approaching AI in the classroom with caution, but teachers are actively exploring AI tools. By staying attuned to the needs and concerns of teachers, we might be able to shape the future of AI integration so that it empowers both educators and learners alike.