It’s back-to-school season, and if the trend of the past few years continues, there will be even greater educator use of artificial intelligence (AI) tools than last year. The degree of AI usage will vary across districts and within schools, partly due to district leaders’ embrace of AI and partly driven by teachers’ openness to using AI tools. As educators’ use of AI tools is more widespread, the question of which tools they use and how they’re deployed becomes more essential.

We’re well past the basic chatbot era of AI tools. Dozens of ed tech companies are offering AI-native products to districts and legacy ed tech products are gaining AI-powered features. Just this past month, OpenAI and Anthropic both announced educator-focused versions of their consumer-facing platforms: ChatGPT’s new educator plugins and Claude for Teachers.

How can educators, district leaders, and state policymakers determine which of these tools are effective versus those that are more sizzle than steak?

That’s the question my Bellwether colleagues answered in partnership with the PIE Network through Signals of Quality.* There are indicators that district leaders can look for when assessing ed tech tools that boil down to a few essential questions:

• Are learning outcomes emphasized more than technology features?

• Does the tool support or undermine productive struggle?

• Is the tool built to support strong pedagogy and instructional practices?

• Can vendors explain — in plain English — how technical choices maximize quality?

• What sustainability risks come with the vendor’s business model, and how are those mitigated?

District leaders aren’t the only ones who can help ensure the use of high-quality ed tech tools in schools. State education agencies (SEAs) can engage in capacity building to support thoughtful procurement in districts and leverage their scale to shape the market and evaluate outcomes. Advocates don’t need to be technical experts to shape ed tech procurement — they can engage by elevating priorities for their communities, assessing what AI tools are currently in use, and monitoring policy developments in other districts and states.

Of course, teachers should be part of ed tech procurement in their schools and districts, but they should also exercise caution with which platforms they test on their own. Even if a product claims to protect private student data, teachers should work with tools that are supported by their districts to ensure they don’t violate student data privacy laws.

There’s an overwhelming number of AI-powered ed tech tools being pitched to educators. It’s up to district leaders, SEAs, and advocates to help ensure that the ones that make it into classrooms are the tools that can best support teachers and students this school year and beyond.

Quick Hits

• An alphabet without an F: AI-generated garbage is showing up on Teachers Pay Teachers, full of typos, errors, and other slop-like characteristics. The record of quality control for material on the popular educational resource and lesson plan marketplace  wasn’t stellar pre-ChatGPT. Now, the marginal cost of generating crummy worksheets on Teachers Pay Teachers is rapidly approaching zero. Let’s hope the growing wave of edu-slop serves as further motivation for teachers to invest in high-quality instructional materials.
• Good fences make good neighbors; what about hyperscalers?: In Issue 11 last July, we flagged that data centers might have more everyday impact on folks than policies that regulate the technical aspects of AI models. It’s certainly a big issue heading into the 2026 midterm elections, and there’s an education angle, too. Nearly 1,900 U.S. public schools sit within a mile of a data center. This can be good for local property tax rolls — a major source of school funding — and can even lead to windfalls for educators. But some states are also providing major sales tax breaks for data centers. What we’re watching: How will the impact on school funding net out?
• Virtual insanity: A school district in western New York announced that it was going to purchase a lifelike robot teacher. The purchase was approved by the school board, but was quickly put on hold after significant backlash from parents and teachers. A good amount of the backlash was tied to the generally dystopian vibe of kids learning from a silicone-skinned robot; the robot manufacturer’s ties to not-safe-for-work products didn’t help either. Let’s hope other districts learn from this and stick to carbon-based teachers.

Odds & Ends

• 🎙️ In case you want more Bellwether perspectives on AI: Andy Rotherham sat down with aiEDU’s Alex Kotran for a wide-ranging conversation about AI in education, including why building knowledge still matters, AI’s equity risks, and how to keep up with the ever-evolving AI landscape. Andy will also be discussing AI tutoring with Reed Hastings and Mike Goldstein in an upcoming Bellwether LinkedIn Live.

• 📖 Currently reading: Wendell Berry’s collection of essays, “What Are People For?,” is a counterweight to techno-optimism. The first entry, “Damage,” is particularly resonant in the dawning Age of AI.

• • *Editor’s note: Some organizations and funders in this newsletter, including the PIE Network, are past or present clients or funders of Bellwether.