Illinois Just Drew the Line on AI in Teacher Evaluations — and Left the Paperwork On the Right Side of It

By Anthony Neely, Ph.D. — co-founder, EvalScribe

My work — and EvalScribe's whole reason for existing — has been about the ethical integration of AI into teaching and learning: not keeping it out, not letting it in everywhere, but being deliberate about where a tool belongs and where a human has to stay. In teacher evaluation, that principle draws a clear line. AI has no business rendering the judgment — but it is genuinely good at the write-up, the slow translation of evidence into rubric-aligned language. Point AI at the watching and you've outsourced the one thing that requires a human in the room. Point it at the paperwork and you've handed an administrator back their Sunday.

This month, Illinois wrote that line into law.

What the law actually says

On July 2026, Illinois enacted Senate Bill 2909, now Public Act 104-0565, amending the state's teacher-evaluation statute (105 ILCS 5/24A-5). The headlines called it a "ban on AI in teacher evaluations," and if you stop at the headline you'll hear something the statute does not say. Here is what it says, in its own words:

"An evaluator is prohibited from using an artificial intelligence tool to assign a numerical score or qualitative rating, such as 'excellent', 'proficient', 'need improvement', or 'unsatisfactory', for any component of a teacher's evaluation or any evaluation task that requires professional judgment. However, an artificial intelligence tool may be used to support the evaluator in administrative tasks."

Read the whole sentence, because most of the coverage stopped at the comma. The prohibition is narrow and it is exactly right: a machine may not assign the rating or perform the professional judgment. The permission that follows is just as deliberate: AI may support the evaluator in the administrative work. The law didn't ban AI from evaluation. It banned AI from doing the one part a present, accountable human is supposed to do — and expressly protected its use for the part that was never about judgment in the first place.

The statute draws the same line a second time for teachers: a teacher may not use AI to generate the evidence of practice that an evaluator will judge, but may use AI to support their own administrative tasks. And it adds a transparency rule worth pausing on — evaluators and teachers must disclose the name and purpose of any AI tool used in the process. Hold onto that one; it matters more than it looks.

This is the distinction, now with a statute behind it

Every teacher evaluation is two acts stitched together. The observation is judgment: a trained human reading a room in ways a transcript never captures — the difference between a quiet class that's disengaged and a quiet class that's deep in thought, the lesson that went sideways for reasons no model outside the room could know. The write-up is translation: taking the evidence a human already gathered and rendering it into the language of a rubric — Danielson, your state framework, whatever instrument your district adopted — so the feedback is specific, aligned, and defensible.

Illinois just legislated the boundary between them. The score and the professional judgment stay human. The administrative translation is fair game for a tool. That is not a compromise the law backed into; it is the line the statute draws on purpose, in plain language, twice.

For a district leader in any state — not just Illinois — that turns a philosophical debate into a usable rule of thumb. Point AI at the paperwork, never at the professional judgment. Illinois is the first to make it black-letter law, but the logic isn't local, and it is a safe bet other states read this and follow.

The permission most administrators didn't know they had

Here's the reframe underneath the whole thing. If the phrase "AI in teacher evaluations" has made you nervous, that instinct was protecting you against the wrong half. You were right to refuse to let a model rate your teachers' craft. You were never obligated to keep doing the write-up by hand — or, worse, inside a consumer chatbot that was never built for personnel records.

You are allowed to use AI for the paperwork. Illinois just said so out loud. The question that remains is the one that actually determines your risk: which tool does the paperwork.

Why the disclosure rule quietly favors a purpose-built tool

Return to that transparency clause — disclose the name and purpose of any AI tool used. Think about what that asks of the common workaround, which is an evaluator pasting a Danielson rubric into a general chatbot and asking for suggestions. What's the name? "ChatGPT." What's the purpose? It's… whatever that session happened to be. The general tool has no defined, evaluation-specific purpose, no record built for disclosure, and a different behavior every time you open it. Compliance with a naming-and-purpose rule is awkward precisely because a general chatbot has no fixed role in your process.

A purpose-built tool answers the disclosure question by design. Its name is stable, its purpose is singular and nameable — translate the evaluator's gathered evidence into framework-aligned language — and it does the same job the same way every time. That's the posture EvalScribe was built for.

It has your framework built in. A general chatbot doesn't know your instrument, so you re-supply the rubric and the scoring every observation and hope it holds the thread — quietly turning every evaluator in your building into a part-time prompt engineer. EvalScribe supports the frameworks already in use across all 50 states — Danielson, Marzano, Stronge, CEL 5D+, NIET, and state-specific instruments like Florida's FEAPs and Tennessee's TEAM — with the rubric hardcoded in, so there's no prompt engineering and no framework drift.

EvalScribe empphasizes "human in the loop". That means, the judgment and the rating stay with you. This is the part that matters most under a law like this one. EvalScribe's job is the administrative side the statute protects: it lets an evaluator capture the way they actually work — type, dictate, or scan handwritten notes on the iPhone, iPad, or Mac they already carry into the classroom — and turns those fragmented notes into evaluation-ready, framework-aligned language, and drafts post-conference coaching scripts. The evidence is yours. The professional judgment is yours. The rating you assign is yours. What changes is that a sixty-minute write-up drops to ten or less — on the order of 750 hours a year given back across a typical district — without a machine ever making the call the law reserves for you.

Where the notes go — the question a personnel record forces

There's a second reason the chatbot workaround should stop you cold, and it has nothing to do with Illinois. Pause on what gets pasted into a general chatbot during an evaluation: candid, identifying notes about a named teacher's performance. Consumer AI tools, by default, may retain those inputs and use them to improve future models. For a personnel record, that should be disqualifying.

EvalScribe's posture is built for districts rather than retrofitted. In its own words, on the site and in its privacy policy: all data resides physically on your device, with no central cloud database; your notes are sent securely to the engine, processed in memory, and returned; neither the notes nor the evaluation are stored on its servers; and its Azure instance is configured to opt out of all model training. The compliance certifications an IT director will ask about — SOC 1, 2 & 3, ISO 27001, FedRAMP High, HIPAA — belong to the Microsoft Azure OpenAI infrastructure it runs on, the same enterprise-grade platform used by major banks and healthcare systems. A Data Processing Agreement and a signed FERPA-aligned attestation are available on request. The tool is built to evaluate teachers — adults — not students, does not maintain student records, and carries a standing recommendation to keep full student names out of notes.

That's the difference between a workflow you can defend to your board and your teachers' association, and one you'd rather no one asked about.

Built by evaluators

EvalScribe wasn't built by a general AI company that discovered schools as a market. It was co-founded by Anthony Neely, Ph.D., a district Teacher of the Year, and Andrea Neely, Ph.D., with a combined thirty years in classrooms — people who wrote these evaluations by hand and know precisely which half of the job is worth automating. The rollout is designed to be painless — no IT integration, no rostering, one access code — so adopting it doesn't become an implementation project. You can watch the demo, and if you want to feel the write-up side on a real observation, try three evaluations free — no card required.

The line was always there. Now it's the law.

Illinois didn't invent the boundary between watching and writing — it just made the boundary official. The score stays human. The judgment stays human. The administrative translation is a tool's job, and the statute says so in as many words. The districts that get this right won't be the ones that used AI the most, or the least. They'll be the ones that were clear about which half — and picked an instrument that keeps the judgment where the law, and common sense, insist it belongs.

See the framework library for your instrument at evalscribe.com/frameworks, size the time your team is spending with the district capacity calculator, or reach the team directly at [email protected].

Frequently asked questions

Does Illinois's new law (SB 2909 / Public Act 104-0565) ban AI in teacher evaluations? Not broadly. It amends 105 ILCS 5/24A-5 to prohibit an evaluator from using an AI tool to assign a numerical score or qualitative rating, or to perform any evaluation task that requires professional judgment. It expressly permits an AI tool to support the evaluator in administrative tasks. It also bars a teacher from using AI to generate the evidence of practice an evaluator will judge, while allowing AI for the teacher's own administrative tasks, and it requires disclosure of the name and purpose of any AI tool used.

Can administrators still use AI to help write up teacher evaluations in Illinois? Yes — the statute explicitly allows an AI tool "to support the evaluator in administrative tasks." What it prohibits is using AI to make the professional judgment or to assign the score or rating. Translating evidence you gathered into framework-aligned language is administrative support; rendering the rating is not.

How does EvalScribe fit a law like this? EvalScribe is built for the administrative, write-up side the law protects: it turns an evaluator's own captured notes into rubric-aligned language while the evaluator keeps the evidence, the judgment, and the rating. Because it has a single, nameable purpose and a framework built in, it also fits the law's requirement to disclose the name and purpose of any AI tool used — in a way an ad-hoc chatbot session does not.

Is EvalScribe secure? What happens to my evaluation data? EvalScribe is local-first: all data resides on your device, not in a central cloud database, and neither the notes nor the generated evaluation are stored on its servers. Observations are processed in memory on Microsoft Azure and returned, and its Azure instance is configured to opt out of all model training. The compliance certifications (SOC 1/2/3, ISO 27001, FedRAMP High, HIPAA) belong to the underlying Azure OpenAI infrastructure. A Data Processing Agreement and a signed FERPA-aligned attestation are available on request.

Does AI replace the administrator in a teacher evaluation? No — and under the Illinois law it legally cannot render the judgment. The administrator observes, exercises judgment, and assigns the rating. AI assists with the write-up: aligning evidence to the framework and keeping language consistent.

Will other states follow Illinois? It's a reasonable expectation. The observation-versus-write-up distinction Illinois codified isn't specific to Illinois, and states routinely borrow one another's education-policy language. The practical takeaway travels regardless of where you are: point AI at the paperwork, never at the professional judgment.

Primary source: Illinois SB 2909 / Public Act 104-0565 (amending 105 ILCS 5/24A-5); signing coverage: Daily Herald, "Pritzker signs 31 new laws, including ban on AI teacher evaluations" (July 13, 2026). Origin of the observation-vs-write-up frame: GovTech, "AI for Teacher Evaluations: Major Time-Saver, or Premature?".