Supporting teachers to be the best they can be. Quality, agency, and human connection in a world of AI. Create quality, aligned curriculum materials - fast.
Joined March 2023
- Tweets110
- Following62
- Followers786
- Likes102
We’re excited to share that QuestionWell has been selected as one of 26 teams advancing to the next round of the ATS Hub Ideas Challenge, led by Digital Promise and SRI International.
Our selected concept, Research-Grade Quality AI Evaluators, is about helping educators and schools better judge the quality of AI-generated learning materials. That includes assessments, instructional resources, and other classroom content, with a focus on whether those materials are genuinely useful, accurate, and aligned to real learning goals.
We’re glad to be part of this group and to keep building on this work. At QuestionWell, we care about making AI in education more useful, more thoughtful, and more grounded in learning science.
You can read the full announcement from Digital Promise here: buff.ly/5av9FOh
Make video lessons from YouTube quickly and reliably with QuestionWell Chrome Extension. One click on any video, a clean transcript, and questions drafted at the moments that matter.
Get started free: buff.ly/iXEjQ01
Or start from the app: buff.ly/o5FKS6Q
"Reading at a 5th grade level" sounds precise. It isn't.
Most readability formulas just count sentence length and syllables. Different formulas can disagree on the exact same text.
What "reading level" actually means, and how we're rethinking it: buff.ly/dDW3lxR
Summer is for resting, and also when the best teaching ideas find you. Paste a YouTube link into Video Lessons in QuestionWell, and you'll get a draft assignment you can review and refine. Build your fall video library one link at a time: questionwell.org/stop-making…
Your QuestionWell Wrapped is ready 🎁
How many questions did you actually generate this year? How do you stack up against other QuestionWell teachers? Your school year, by the numbers.
See yours: app.questionwell.org/wrapped…
End-of-year budgets are closing. Your school can pay for QuestionWell Premium.
Send admin the Pay for Me page. $70/year, no auto-renew. Unlimited generations, video lessons, standards alignment, vocab tools.
Easiest ask on their desk all week.
app.questionwell.org/pricing…
We've been talking about why multiple choice matters and how we're building better MCQs with AI.
But here's what kicked it all off: we found that one of the most widely cited AI benchmark datasets is riddled with classic item-writing flaws.
Correct answers are consistently longer. Extra spacing gives away the right choice. In some subjects, just picking the longest answer doubles your score.
These are the same mistakes assessment researchers have warned about for decades. And they're baked into the data AI learns from.
If we want AI that writes good questions, we have to start by understanding what makes questions bad.
questionwell.org/blog/item-w…
#AIinEducation #EdTech #Assessment
Last week we made the case for multiple choice.
This week: what happens when you actually try to make them better with AI?
Writing good MCQs is surprisingly hard. Most AI tools just generate questions and hope for the best. We took a different approach: we measured quality using known item-writing flaws from the assessment research literature, then trained a model against those signals.
The results are encouraging, and we're sharing the data.
Read the full breakdown on our blog.
questionwell.org/blog/we-tra…
#AIinEducation #EdTech #Assessment
Stop making EdPuzzles by hand.
Find a video. Rewatch it. Pause. Write questions. Adjust timestamps. Repeat. Sound familiar?
QuestionWell does the hard part for you: finds the right moments, generates high-quality questions, and matches them to accurate timestamps.
Create a video lesson in minutes, not hours.
questionwell.org/stop-making…
Teachers are designers, not users.
Every AI tool in education embeds beliefs about how learning works, whose knowledge matters, and what "good" teaching looks like. Most of those beliefs are invisible.
We think they should be visible, adjustable, and under teacher control.
We wrote a manifesto for what AI in education should look like: transparency over obscurity, amplification over automation, design over default.
questionwell.org/blog/a-mani…
#AIinEducation #EdTech #TeacherAgency
Multiple choice questions get a bad rap. "Too easy." "Only tests recall." "The lazy option."
Here's what most people miss: bad multiple choice is worse than open response. But good multiple choice can actually be better.
Three reasons:
• Instant grading gives teachers 2+ hours back per assignment
• Well-crafted answer choices reveal specific misconceptions, not just right vs. wrong
• They cut right to the heart of student thinking in ways open response often can't
The problem was never the format. It was how we used it.
questionwell.org/blog/the-ca…
#AIinEducation #EdTech #Assessment
Your AI committee built a checklist. But does it go beyond "does this look good?"
6 practical steps to evaluate AI tools for your school, from content quality to what happens when you stop paying.
questionwell.org/blog/6-step…
#AIinEducation #EdTech #SchoolLeadership
AI lesson plans were cool. In 2023.
Now the question isn't "can AI generate materials?" — it's "are they any good?"
Some studies are finding AI isn't saving teachers time at all. Because sorting through low-quality output is its own kind of work.
We're building AI that helps teachers evaluate and refine, not just generate.
questionwell.org/blog/genai-…
#AIinEducation #EdTech
AI tools often promise to “adjust text to grade level”… but how often do we actually verify that claim?
This post is a great reminder: readability isn’t just a setting you toggle — it’s something you have to check. Many AI tools can rewrite text, but that doesn’t guarantee the output truly matches the intended reading level or preserves meaning.
buff.ly/AZrKA8J
What if the best way to help students learn with AI is to take something away?
Our founder, Maya Bialik, just published a new piece with Solution Tree exploring a "balance bike" approach to AI in education: start from day one with the hardest human skills. Navigating uncertainty, making sound judgments, and staying in the driver's seat.
AI can take students further than ever. But only if we design learning that keeps them balancing, not coasting.
Read Maya’s full post here: buff.ly/FBBi1aL
#AIinEducation #StudentAgency #TeacherLed #EdTech #QuestionWell
At #NAISAC 2026?
Join Maya Bialik (QuestionWell AI) + Peter Nilsson for:
What Learning Science Teaches Us About AI
🗓 Feb 27 | 1:15–2:15 PM
🎯 AI for Learning & Leading
When should students use AI and when shouldn’t they? This session explores research-based answers grounded in learning science.
Book: irreplaceable.info
Workshop: buff.ly/fRxkCT4
Headed to #NCCE in Seattle? We’d love to see you at Greg Bagby’s session featuring QuestionWell! 🎉
Greg is sharing practical AI tools you can use right away — including how QuestionWell helps cut quiz-writing and grading time with fast exports.
📍 The AI Trio: QuestionWell, KIRA, and IBM SkillsBuild
🗓 Friday, Feb 27 | 12:00–12:50 PM
🎯 Focused Learning Session (AI)
View the session:
buff.ly/V4cVk60
Presenting at a conference and want to feature QuestionWell?
Email [email protected] with subject line: conference:___
With @Gregbagby 👏
#NCCE26 #EdTech #AIinEducation #QuestionWell
New from QuestionWell: Clever Roster Sync — Teachers can now import their class sections from Clever with one click, assign quizzes and interactive video assignments, and automatically group student results by section for easier analysis. A smoother workflow + better insights = more time for teaching.
Read more: buff.ly/3FtvMjZ
Try QuestionWell: buff.ly/IWsrnCI
Discover how QuestionWell's AI Quality Check tools can boost teacher agency and instructional judgement...try it out! 💡📘#AI #pedtech #teachers #lessonplanning questionwell.org/blog/genai-…
Writing good multiple-choice questions is surprisingly hard—and most AI evaluations don’t really measure that.
At QuestionWell AI, we've been researching how to produce better multiple-choice questions with language models. Instead of eyeballing outputs, we measured quality using known multiple choice question writing flaws from the assessment literature—and trained against those signals.
There's nothing magical here, but the results are encouraging and validate our choices. We'll be releasing this new model as a Beta for all users, and improving it over time.
If you're interested in assessments, education, or practical LLM evaluation, check out our blog.
