Generative AI has changed a basic assumption behind written assessment.

For years, instructors could look at a submitted essay, report, reflection, or take-home assignment and reasonably assume that the final document provided a meaningful representation of the student's writing process.

Today, the final document tells only part of the story.

Students can use AI to brainstorm ideas, generate outlines, clarify concepts, revise paragraphs, improve grammar, or produce complete responses. Those uses are not equivalent, and a finished document alone may not reveal how the work was developed.

Recent research reflects this challenge. Zhang et al. (2026) argue that in generative AI-affected writing environments, final texts alone provide limited insight into how students produced, revised, and justified their work. Their classroom-based study instead examined the value of writing-process evidence and the procedures used to review it.

That creates a more useful question for educators than simply:

"Did this student use AI?"

A better question is:

"How was this work developed, and what role did AI play?"

That is the idea behind authorship transparency.
‍

AI Detection and AI Transparency Ask Different Questions

Traditional AI detection generally evaluates a finished piece of writing and estimates whether some of the text may have been generated by AI.

Authorship transparency takes a different approach.

Instead of looking only at the finished submission, it gives instructors more context about the process that produced it.

VisibleAI records the writing process inside its editor, including typing, revisions, pasted content, AI-generated or AI-assisted text, and conversations with the built-in AI assistant.

Instructors can review students' conversations with the built-in assistant alongside the document's revision history.

This additional context can help instructors better understand how an assignment developed rather than relying only on the final submission.

The broader value of process evidence is also reflected in recent research. In an exploratory study of four academic-writing classes at a university in China, Zhang et al. (2026) asked students to provide evidence such as outlines, drafts, revision traces, and literature-search records. The study suggests that examining the process behind a final text can provide a richer basis for reviewing student work in GenAI-affected writing environments.

VisibleAI also provides AI-related percentages, but these should be interpreted as indicators rather than verdicts. AI-generated content percentages are probability estimates, and content created or revised in outside tools and later pasted into the VisibleAI editor can affect authorship insights.

The purpose is not to make the academic judgment for the instructor.

It is to provide more evidence for the instructor to consider.
‍

Transparency Is About More Than Collecting Evidence

Process evidence alone is not enough. How that evidence is interpreted matters too.

This distinction is particularly important in Zhang et al.'s (2026) study.

Researchers compared two approaches to handling GenAI-related writing concerns. One emphasized explanation, student response, and opportunities for correction. The other relied more heavily on detector thresholds, warnings, and routine follow-up.

Students in the more procedural approach showed higher end-of-term writing-process transparency than those in the technology-deterrence approach (Hedges' g = 0.48). The study also found higher perceptions of procedural fairness (g = 0.59) and lower anxiety about being falsely accused (g = -0.41).

Because this was an exploratory classroom-based study in one university context, these findings should not be treated as proof that the same effects will occur in every course or institution. However, they highlight an important principle:

Transparency works best when evidence supports a fair conversation rather than automatically producing a judgment.

The study did not evaluate VisibleAI. Its relevance is in the broader approach it supports: looking beyond a final-text score, considering evidence from the writing process, allowing students to explain their work, and keeping academic judgment with educators.
‍

Faculty Still Decide What Appropriate AI Use Looks Like

There is no single AI policy that works for every course or every assignment.

Appropriate AI use depends on what students are supposed to learn.

For example, an instructor might want students to complete one assignment with little or no AI-generated content so they can demonstrate what they understand independently.

On another assignment, the instructor may permit AI for brainstorming or clarification while expecting the central argument and analysis to remain the student's own.

A course focused specifically on AI literacy might allow greater AI use while requiring students to evaluate, revise, verify, and take responsibility for what the technology produces.

VisibleAI supports this flexibility through assignment-level settings.

Instructors can set limits for AI-generated content and pasted content, including turning either option off entirely.

They can also choose how the built-in AI assistant responds.

In Guiding mode, the assistant explains ideas and asks guiding questions designed to encourage students to work through the problem rather than simply receiving a direct answer.

In Standard mode, the assistant responds using the course and assignment context with its default behavior.

Instructors can also choose whether students complete a short reflection on how they used AI.

The result is an AI policy that becomes part of the assignment itself rather than something students encounter only after a concern arises.
‍

Transparency Can Change the Conversation

When instructors have more context around how work developed, conversations about AI can become more specific.

Instead of asking only:

"Did you use AI?"

an instructor can ask:

"What were you trying to accomplish when you used AI?"

"Which parts of the work represent your own reasoning?"

"What did you change after receiving AI assistance?"

"How did you decide whether the AI response was accurate or useful?"

"Can you walk me through how this part of the assignment developed?"

These questions place the focus back on student thinking.

They also recognize an important reality: using AI and outsourcing thinking to AI are not necessarily the same thing.

A student who asks AI to clarify a difficult concept, evaluates the response, verifies the information, and develops an independent argument has engaged with AI very differently from a student who submits an AI-generated response without understanding it.

Authorship transparency gives instructors more context for making that distinction.

This approach is consistent with Zhang et al.'s (2026) finding that writing-process transparency is shaped not only by the evidence students provide but also by the procedures through which that evidence is reviewed.
‍

What Faculty Are Saying About VisibleAI

Faculty at Point Loma Nazarene University have described VisibleAI as useful for creating a more transparent environment for students working with generative AI.

Christine Tokunaga, Program Director at Point Loma Nazarene University, described VisibleAI as providing a:

"safe and transparent way for my students to explore the power of generative AI tools."

That perspective frames transparency as more than an academic integrity mechanism.

Students can explore AI while knowing that the process remains visible to their instructor.

Catherine Walter, also a Program Director at Point Loma Nazarene University, highlighted another faculty perspective:

"I feel like VisibleAI has been really helpful for my students, especially for supporting students in their classrooms who are learning English as a second language."

This should not be interpreted as evidence that VisibleAI improves language-learning outcomes. Rather, it illustrates one way a faculty member sees transparent AI support fitting into the needs of her students.

VisibleAI gives instructors the flexibility to decide where AI assistance supports the learning goal and where students need to work independently.

Internal publication note: Confirm written permission to publish the Point Loma quotes before the blog goes live.
‍

What Students Reported About VisibleAI

A Winter 2025 survey included 164 undergraduate and graduate students using VisibleAI.

One indicator of continued interest was that:

69% said they wanted VisibleAI in a few or all of their other courses.

The survey also found:

82% used VisibleAI for their assignment.

45% said VisibleAI helped build trust with their teachers.

Students also commented positively on being able to distinguish their own work from AI-generated content and better understand where text in an assignment came from.

These results represent student perceptions, not evidence that VisibleAI caused a particular learning outcome.

But they do provide insight into how students experienced a more transparent approach to AI use.

Rather than transparency necessarily feeling punitive, many students indicated that they would like to experience the tool in additional courses.
‍

A Flag Should Start a Conversation, Not Make the Decision

Academic integrity decisions should remain human decisions.

VisibleAI provides evidence about how work developed.

The instructor decides what that evidence means.

A percentage, an unusual writing pattern, or a high level of pasted content may give an instructor a reason to look more closely at an assignment. It should not automatically determine whether misconduct occurred.

This distinction is especially important given the procedural findings reported by Zhang et al. (2026). Their study found that an approach emphasizing explanation, opportunities for student response, and correctability was associated with stronger perceptions of procedural justice and lower false-accusation anxiety than an approach centered on detector thresholds.

In practical terms, evidence should help instructors investigate and ask better questions, not replace those questions.

With access to writing history and other authorship insights, an instructor can review the broader context and speak with the student before reaching a conclusion.

That creates space for questions such as:

• Was the student's AI use consistent with the assignment policy?
• Does the writing history provide additional context?
• Was outside material pasted appropriately and cited?
• Can the student explain the reasoning behind the work?
• Does the student understand what they submitted?
• Can the student explain how and why AI was used?

Technology can help surface those questions.

The academic judgment still belongs to the instructor.
‍

Responsible AI Use Requires More Than a Good Prompt

Students are entering workplaces where generative AI will increasingly be part of everyday work.

Preparing them for that environment requires more than teaching them how to prompt an AI system.

Students need to learn:

• When AI is appropriate
• When AI should be questioned
• How to evaluate an AI response
• What information needs to be verified
• How to maintain ownership of their work
• How to explain the decisions they made
• How to communicate transparently about AI assistance

These skills are difficult to teach when AI use remains hidden.

Authorship transparency gives instructors an opportunity to make AI use part of the learning conversation rather than something investigated only when misconduct is suspected.

The goal is not simply to expose AI use.

It is to make the student's decision-making, revision, reasoning, and responsibility more visible.
‍

VisibleAI Can Work Without Peer Assessment

VisibleAI does not need to be paired with a peer assessment activity.

Instructors can create a stand-alone Individual Authorship Activity that contains only the Create stage.

Students complete their work inside the VisibleAI editor, while the instructor can use authorship tracking, the built-in AI assistant, citation features, AI policy settings, and optional reflection.

After the Create stage ends, the instructor reviews and grades the submissions directly.

This format can work particularly well for:

Journals, reflections, written assignments, take-home assessments, and other work where the instructor wants greater visibility into the creation process without adding peer evaluation.
‍

From Detection to Evidence

AI has changed the way students write.

Assessment needs to adapt with it.

Emerging research suggests that one promising direction is to look beyond the final text and consider evidence of how students planned, drafted, revised, and developed their work. Just as importantly, that evidence should be reviewed through processes that allow explanation and preserve instructor judgment (Zhang et al., 2026).

VisibleAI does not tell instructors what to conclude.

It shows them more of how the work was written, so they can make decisions using evidence.

Instead of relying entirely on a score applied to a final submission, instructors can see more of the process that produced it and establish clearer expectations for AI use from the beginning.

That creates a stronger foundation for responsible AI use, academic integrity, and more productive conversations about student learning.

The question becomes less about simply determining whether AI was present and more about something educators ultimately care about much more:

Did the student still do the thinking the assignment was designed to develop?
‍

Explore VisibleAI

Use VisibleAI as a stand-alone authorship activity to make the writing process more transparent, establish clearer expectations for responsible AI use, and give instructors more evidence about how student work develops.

CTA: Explore VisibleAI

‍

‍