When AI Can Do the Task: Redesigning Product-Based Assessment
Many of us teachers love dreaming up a final project: a data-based research paper, a documentary tied to unit content, a presentation that resembles a real-world policy proposal, or an original poem or reflection. The possibilities are endless.
These are examples of authentic assessments, tasks that represent the work and skills students may encounter beyond school. In the past, evaluating these final products could provide rich insight into students’ learning while helping them develop future-ready skills.
The most meaningful learning has always happened along the way: outlining, researching, drafting, receiving feedback, revising, and struggling through difficult ideas or new skills. As teachers, we know that best practice includes assessing students throughout this process and providing timely, relevant, and actionable feedback. Yet we often rely heavily on assessing students via the final submission. This is understandable. Providing detailed feedback throughout a project is time-consuming.
Now, however, Generative AI has definitively disrupted our ability to treat the final product as sufficient evidence of learning. Large language models can produce polished, professional-looking work in moments. With valid reasons, many teachers are returning to monitored assessments, in-class tasks, and on-demand writing.
I agree that these assessments are a necessary part of an AI-enabled classroom. However, as a passionate supporter of authentic assessment, I mourn the possible loss of real-world tasks that resemble the work students may eventually do as professionals. Recent research highlights an interesting tension: higher education professionals overwhelmingly cite critical thinking and problem solving as their main learning priorities, while employers and learners alike cite practical experience as one the biggest gaps in being prepared for the workplace (Pearson and Amazon Web Services 39). Students regularly cite a desire for learning experiences where they can apply what they know in real-world applications. I think that the classroom is a fantastic place to try out modern skills in an environment that intentionally preserves critical thinking and ethical decision making.
But when students complete their assessments with access to AI tools, how do we know which assessments are still usable? The question is not simply which assessments remain valid. As explored in Blog 3, a better question is:
How do we preserve authentic assessment while ensuring that we are still measuring student learning and growth?
The answer begins with redesigning assessments so that learning remains visible.
Summative Assessments Are Under Pressure
Grading the final product of a well-designed assessment was once common practice. The rise of widespread internet access concerned many teachers, but effective plagiarism-detection software helped address some of those concerns.
Generative AI presents a different challenge. LLMs can follow the directions of previously effective summative assessments and generate grade-appropriate responses in moments. AI detection does not eliminate this issue. In one study, AI-generated submissions earned grades comparable to genuine student submissions. Although an AI-detection tool flagged most experimental submissions as containing some AI-generated material, it identified only 54.8% of the total AI-generated content (Perkins et al., 2024). AI detection also raises serious questions about fairness. Stanford researchers found that seven detectors misclassified 61.22% of TOEFL essays written by nonnative English writers as AI-generated. At least one detector flagged 89 of the 91 essays tested (Zou et al., 2023). Sayonara, AI silver bullet!
In Blog 2, we discussed how trust and transparency can move students away from irresponsible AI use. However, a healthy AI culture must also be paired with deliberate assessment design.
First, Clarify the Assessment’s Purpose
Before overhauling every favorite summative assessment, ask:
Is this assessment primarily of learning, or for and as learning?
An assessment of learning measures independent mastery, often in a higher-stakes setting. It may require safeguards around AI use, such as an in-person assessment, on-demand writing, or a monitored digital environment.
An assessment for or as learning centers growth, reflection, practice, and skill development. It may provide more room for AI integration if students’ thinking remains visible and AI augments rather than replaces learning (Pearson, 2026).
Consider, for example, a U.S. Government course in which the teacher wants to assess students’ understanding of Supreme Court cases involving individual liberties:
Assessment of learning: A proctored, on-demand free-response essay completed on paper or through a secure digital environment.
Assessment for or as learning: A multiphase research project comparing a historical case with one decided within the past five years.
The first demonstrates independent mastery of previously practiced knowledge and skills. The second allows students to acquire and apply knowledge throughout the project. I believe both can contribute to summative evaluation when their purposes, processes, and success criteria are planned intentionally.
Set Clear AI Expectations
Before throwing out the baby with the bathwater, consider which existing assessments leave room for responsible AI use.
Teachers must first decide whether AI would preserve learning, support it, or undermine the task’s purpose. Then, let students into that thought process. Clearly communicate what AI use is acceptable, what is not, and why.
This clarity matters. RAND found that students experience substantial ambiguity around AI use, particularly when rules vary by teacher. The same study found that 67% of students believed increased AI use for schoolwork would harm critical-thinking skills (Schwartz & Diliberti, 2026). College Board similarly found that AI policies remain inconsistent across schools and classrooms, even as student use continues to grow (Adair et al., 2025).
Students are seeking guidance. It is our responsibility to eliminate the guesswork.
Here are two frameworks can help.
The Stoplight Model
The Stoplight Model communicates one of three expectations for each assignment:
Red: AI is not permitted.
Yellow: AI is allowed with permission or explanation.
Green: AI use is encouraged within defined ethical expectations.
Mormando recommends including AI expectations directly in assignment instructions and explaining why a task falls within a particular category (Mormando, 2023).
The AI Assessment Scale
The AI Assessment Scale offers five levels:
No AI
AI planning
AI collaboration
Full AI
AI exploration
The appropriate level depends on the learning outcomes being assessed. The framework is designed to create greater clarity and transparency while recognizing that AI may be appropriate for some assessments but not others (Perkins et al., 2024).
Balance Independent Mastery and Authentic Assessment
No-AI assessments still have an important role. In-person tests and on-demand writing can provide reliable snapshots of what students know and can do independently. They should also be supplemented with other evidence of learning.
Not every student demonstrates understanding best through a traditional test, and these formats may limit collaboration and authenticity and increase performance anxiety for some students. Real-world work rarely requires people to operate entirely alone, without tools, feedback, or revision.
This brings us back to Blog 2’s themes of trust and transparency. Rather than forcing students to hide AI use, we can explain when independent mastery is necessary and when technology, collaboration, and revision are appropriate. The goal is not to choose between authentic and traditional assessment. Strong assessment in the AI era requires both. When students understand each task’s purpose and AI expectations, teachers can preserve trust while gathering richer evidence of learning.
Looking for more ways to future-proof your assessments? Follow along for the final blog in this series where we define and exemplify six high impact shifts to future-proof our assessments!
References
Adair, A., Howell, J., Jacklin, A., & Radford, A. W. (2025). U.S. high school students’ use of generative artificial intelligence: New evidence from high school students, parents, and educators. College Board. https://research.collegeboard.org/media/pdf/ai-research-brief-1_vf.pdf
Mormando, S. (2023, November 9). A stoplight model for guiding student AI usage. Edutopia. https://www.edutopia.org/article/creating-ai-usage-guidelines-students
Pearson. (2026). School educator guide: Redefining formative assessment in a generative AI era. https://plc.pearson.com/sites/pearson-corp/files/2025-12/school-educator-guide-pearson.pdf
Pearson and Amazon Web Services. AI Readiness: Building the Bridge from Higher Education to Work. Apr. 2026, https://www.pearson.com/content/dam/global-store/global/resources/ai-readiness/AI-Readiness-Report-2026.pdf.
Perkins, M., Roe, J., Postma, D., McGaughran, J., & Hickerson, D. (2024a). Detection of GPT-4 generated text in higher education: Combining academic judgement and software to identify generative AI tool misuse. Journal of Academic Ethics, 22(1), 89–113. https://doi.org/10.1007/s10805-023-09492-6
Perkins, M., Furze, L., Roe, J., & MacVaugh, J. (2024b). The Artificial Intelligence Assessment Scale: A framework for ethical integration of generative AI in educational assessment. Journal of University Teaching & Learning Practice, 21(6). https://doi.org/10.53761/q3azde36
Schwartz, H. L., & Diliberti, M. K. (2026). More students use AI for homework, and more believe it harms critical thinking: Selected findings from the RAND American Youth Panel. RAND Corporation. https://www.rand.org/pubs/research_reports/RRA4742-1.html
Zou, J., Liang, W., Yuksekgonul, M., Mao, Y., & Wu, E. (2023, May). James Zou et al. warn on the objectivity of AI detectors. Stanford Electrical Engineering. https://ee.stanford.edu/james-zou-et-al-warn-objectivity-ai-detectors
About the author
Morgan Lieske
Morgan Lieske holds a master's degree in Curriculum and Instruction and brings a decade of experience teaching, coaching, and leading curriculum initiatives in college preparatory K-12 education. As an educator, department chair, and instructional leader, Morgan has designed engaging learning experiences that foster critical thinking, inquiry, and student ownership of learning. Her work spans social studies and interdisciplinary studies, where she developed a passion for creating dynamic curriculum that connects learning across disciplines and prepares students for success beyond the classroom.
Throughout her career, Morgan has championed student-centered learning practices that promote collaboration, authentic learning, and higher-order thinking. As an Advanced Placement educator and facilitator of school-wide programs, including National History Day, she has helped students engage deeply with content while developing the skills needed for college, career, and civic life. She also has experience leading curriculum alignment efforts and instructional improvement initiatives, supporting educators in creating rigorous, engaging, and technology-enhanced learning environments.
A Fulbright Scholar and lifelong learner, Morgan enjoys exploring new ideas both professionally and personally. She lives in Minneapolis, where she can often be found attending live music events, biking, waterskiing, or painting.
