The AI in Your Curriculum Didn’t Ask Permission

Most districts think of curriculum adoption as a decision they make once. A committee reviews the options, the board approves, teachers get trained, and the materials stay put for years. AI in instructional materials is quietly breaking that assumption. The product you approved last spring may not be the product your students are using this fall.

That is the real story in this week’s coverage of a new EdReports brief. Not that the evidence is thin. We knew that. The story is that the AI is arriving inside the tools we already trust, sometimes without anyone telling us.

And that turns a purchasing question into a judgment question.

What’s happening with AI in instructional materials

On September 29, EdSurge’s Ira Apfel reported on a study finding that ed tech is rushing AI integration before proving it works. The study is EdReports’ brief, AI in K-12 Instructional Materials: What We’re Seeing, released earlier this month. EdReports looked at 10 K-12 providers: four major publishers whose materials are among the top 10 purchased, three mid-sized curriculum companies and three digital-first ed tech firms. Only one of the 10 provided third-party evidence that its AI features improved student outcomes.

Courtney Allison, EdReports’ chief academic officer, told EdSurge: “I knew that there wasn’t going to be much evidence, but I was surprised at how little there was.”

Two details from the coverage matter most for leaders. First, EdSurge notes that even identifying whether instructional materials contain AI can be hard, because it can be introduced without notification from the vendor. Second, Jean-Claude Brizard, president and CEO of Digital Promise, named the worst case plainly: “quality assurance at adoption no longer guarantees quality at instruction.”

The EdReports brief itself draws a distinction every leader should notice. In its words, “Some tools keep educators ‘in the loop’ before AI-generated content reaches students, while others interact with students and adapt their experiences in real time.” It also found that “the field has made more progress evaluating AI-related risks than instructional quality.” Privacy, safety, bias and transparency have frameworks. Whether AI actually strengthens learning, or keeps a curriculum coherent, does not yet have a consistent way to be judged.

On September 30, EdSurge followed up with a podcast episode, Can Schools Trust AI Features in Edtech Products?, featuring Allison and EdReports CEO Lewis Ferebee, the former DC Public Schools chancellor. The episode covers the difference between tools that let a teacher review content before students see it and tools that put content directly in front of students in real time, why the old line between core and supplemental materials is starting to disappear, and the questions Ferebee would ask a vendor sitting across the table.

What this reveals for school leaders

In Chapter 13 of Human Still Required, I write about what schools must protect: critical thinking, productive struggle, reflection and human feedback that matters. The line from that chapter that keeps coming back to me this week is this one:

What schools protect reveals what they value. What they automate reveals what they’re willing to lose.

Human Still Required

Here’s the uncomfortable part. When AI shows up inside a curriculum through a product update, a district hasn’t decided what to automate. A vendor has. If a tool adapts a student’s experience in real time with no teacher in the loop, someone made a choice about what instruction looks like for that child. It just wasn’t us.

That doesn’t make the tools bad. Some of them may help. The point is that the decision about what gets automated in core instruction belongs to educators, made on purpose, with eyes open.

Chapter 7 is about why judgment can’t be automated, and it applies directly here. Badges, certifications and polished feature lists are a form of recommendation. They can be useful. They are not the decision.

Wise leaders don’t ask, ‘Is this recommended?’ They ask, ‘Is this right for us?’

Human Still Required

“Right for us” requires knowing what you are trying to do for students before a sales demo tells you. The experts in the EdSurge coverage made the same point: start from the instructional vision for students, then look at the tool. Without that vision, every new AI feature looks like progress.

There’s a quieter issue too. Brizard’s warning about quality at adoption versus quality at instruction is really a statement about where leadership attention needs to go. For years, we treated the adoption meeting as the moment of judgment. In an AI era, the judgment has to continue into classrooms, where teachers see what the tool actually does with kids.

Data informs. Humans decide. That applies to the curriculum we buy as much as to anything else we do.

What leaders can do now

• Run an AI inventory of your core materials. Ask every curriculum and ed tech provider, in writing, which AI features are currently active in your instance, when they were added and whether they can be turned off.
• Sort features by who is in the loop. Use the EdReports distinction. Does a teacher review AI-generated content before students see it, or does the tool adapt the student’s experience on its own? Set a higher bar for the second category.
• Make opt-in the default in contracts. New AI features should require district approval before they reach students. Put notification and opt-in language into renewals now.
• Ask what the badge actually demonstrates. When a vendor points to a certification, ask what it measures. Privacy compliance is not evidence of learning. Ask for third-party evidence specific to the AI feature, not to the product it was added to.
• Keep judgment going after adoption. Give teachers a simple way to report what AI features are doing in their classrooms, and review it each semester. Teachers are the first to notice when instruction changes.

If your leadership team wants a shared starting point for these conversations, the Human Still Required leadership guide is built for exactly this kind of discussion.

Frequently asked questions

How do I know if our curriculum has AI features?

Ask the provider directly and get the answer in writing. EdSurge’s coverage of the EdReports brief notes that AI can be introduced without notification from the vendor, so an annual check of what is active in your district’s version of each product is a reasonable practice.

Is there evidence that AI features in instructional materials improve learning?

Not much yet. In the EdReports brief, only one of 10 providers offered third-party evidence that its AI features improved student outcomes. EdReports also notes that research on an established product may not apply to newly added AI features.

What should a district ask a curriculum vendor about AI?

Start with what the AI does, who reviews its output before students see it, whether it can be turned off, and how the district will be notified of changes. Then ask for evidence specific to the AI feature and what any badge or certification actually demonstrates.

The work only humans can do

AI will keep showing up in the materials students use every day. Some of it will help. Some of it won’t. None of it should arrive without a human deciding it belongs there.

Adoption used to be the decision. Now it’s the beginning of one.

So here’s my question for you: if a teacher in your district asked tomorrow what AI is active in their core curriculum, could anyone answer?

Sources

• Study: EdTech Is Rushing AI Integration Before Proving It Works — EdSurge, September 29, 2026
• Can Schools Trust AI Features in Edtech Products? (podcast) — EdSurge, September 30, 2026
• AI in K-12 Instructional Materials: What We’re Seeing — EdReports, September 8, 2026 (covered by EdSurge this week)