Where It Helps Most
Writing is where most people first try AI, and where the results are most uneven. Used one way it removes the friction of starting and leaves the thinking to you. Used another way it produces competent, forgettable text that readers increasingly recognize and discount. The difference is mostly about where in the process you bring the tool in.
The blank page is a genuine obstacle, and AI is very good at removing it. So is the "I know what I want to say but not how to open" problem, and the "this paragraph is fine but four sentences too long" problem.
What it does not do is know what you think. If you have not worked out your position, AI will supply a reasonable-sounding one, and you will end up defending an argument you did not make. That is the failure mode to watch for — not obvious plagiarism, but quietly outsourcing the judgment.
Practical Approaches
- Ask for openings, not the document. "Give me five different ways to open a message about a schedule change" gets you moving without committing you to a structure.
- Talk it out first. Describe what you are trying to say in messy, unstructured terms and ask for it back organized. This works remarkably well, because the thinking was already yours.
- Ask for an outline and argue with it. The disagreement is productive — realizing "no, the second point should come first" is you doing the work.
- Dictate and clean up. Speak the rough version, paste it in, ask for it tightened. Faster than typing for many people, and it preserves your phrasing.
Bring your own draft and ask for specific changes rather than general improvement:
- "Cut this by a third without losing the second point."
- "This is too formal for an email to colleagues I know well."
- "Where does the logic jump?"
- "Rewrite this for someone who has not read the earlier message."
- "Which sentence is doing the least work?"
"Make this better" produces generic polish that sands off your voice. Specific instructions produce specific improvements.
The same content frequently needs three versions: one for students, one for colleagues, one for administrators. Writing all three is tedious; adapting one is fast.
"Rewrite this announcement for students rather than faculty." "Give me a two-sentence version for a newsletter." "Make this readable for someone whose first language is not English."
Check the result. Adaptation sometimes changes meaning, particularly when simplifying anything with conditions or exceptions attached.
Default AI prose has a recognizable texture: balanced, hedged, structurally tidy, fond of "it is important to note," and lightly enthusiastic about everything. Readers notice it. In professional contexts it reads as a lack of care.
To avoid it:
- Give it samples. Paste three things you have written and ask it to match how you write. Far more effective than describing your style.
- Name what to avoid. "No bullet lists." "Do not start sentences with Additionally." "Do not hedge." "Contractions are fine."
- Read it aloud. If it does not sound like you talking, it will not read like you wrote it.
- Put your own sentences back in. The line you would actually say is usually better than the smooth version.
Images and Other Media
Image generation raises questions that text generation does not, and they are genuinely unsettled.
These tools learned from very large collections of existing artwork, photography, and design. Whether that use required permission or compensation, and what it means for people who make a living from that work, is actively disputed — in the field, in the courts, and among faculty and students in our own programs. Some creative professionals decline to use these tools on principle; others have built them into their practice. This site does not resolve that for you, and Discipline-Specific Skills has questions worth taking up in a course.
Practical points that are not in dispute, if you do use them:
- Copyright status is unsettled. Under current U.S. Copyright Office guidance, purely AI-generated images generally cannot be registered for copyright. If something needs to be protected, this matters.
- Do not generate real people. Not colleagues, not students, not public figures.
- Say so. If an AI-generated image appears in college material, label it.
- Accessibility still applies. Every image needs alt text, and AI-drafted alt text needs a human check — it describes what is visually present, not what the image is doing in context.
- Check before official use. Anything carrying Maricopa branding goes through Marketing regardless of how it was made.
A word on academic work
For students: everything above describes tools, not permissions. What you are allowed to use depends on your instructor and your assignment, and those rules differ by course. Check the syllabus, and ask if it is unclear — asking is never the wrong move.
The reliable test: if the assignment exists so that you will practice a skill, having AI do it removes the reason for doing it. Using AI to understand feedback on your draft is different from using AI to write the draft, even though both involve the same tool.