What It Really Takes to Write with AI: Ethics, Edits, and Human Discernment

Sarah Downey
Sep 4, 2025
4 min read

Updated: Dec 19, 2025

Yes, I used AI. No, that doesn’t mean I didn’t work hard. Here’s what actually goes into it.

“Did you get AI to write that for you?” gets tossed around like a disqualifier.

Often, it comes from someone who hasn’t worked with the tools and doesn’t see what’s involved.

For me, AI changes the process. It feels more co-creative. Less work? I don’t think so.

When I use AI well, I engage more deeply, not less. The work is different. It’s partnered.

I become editor, ethicist, evidence-checker, and accountability author. The model speeds mechanics. I carry meaning, care, and consequence.

What about you? If you use AI for writing, what are the human parts you hold on to?

The work I do now when I write

1) Clarify the task

(Intent, audience, constraints)

Before I touch any tool, I get clear on my message. What do I want to say? Who is it for? What must be included or avoided? If I’m not crystal clear, I might ask AI to help clarify my thinking before I start. Clarity up front prevents drift and keeps me in the lead.

2) Lateral reading, always

Lateral reading means verifying claims by consulting multiple credible sources and is a habit borrowed from professional fact-checkers.

We live in an age of easy misinformation. Using AI has strengthened my commitment to verification. I leave my draft often. I open credible tabs, compare what they say, and only then adopt a claim. I bring a specific citation back so someone else can trace my path.

3) Fact-check and attribute

AI can sound confident and still be wrong. I confirm numbers, dates, names, and statistics against reliable sources. I attribute where it helps the reader. If a quote is edited or anonymized, I say so. Transparency builds trust.

4) Bias and dignity review

I know AI tools reproduce human conditioning—what decolonial scholars name Implicit Conditioning: the systemic programming that encodes stereotypes and dominance. I use bias-checking tools, scan for stereotypes, ask whose voice is missing, and rewrite until the language holds dignity and agency.

5) Relationship check (Consent and privacy)

If I’m writing about real people, I pause. Do we have consent? Could details identify someone who didn’t agree to be named? If a person cannot reasonably opt out, I choose not to include it.

6) Accessibility and plain language

Short sentences. Familiar words. Clear headings. I write for stressed brains, screen readers, and neurodivergent processing styles. Care for the reader comes first. I’m writing to share ideas, not to prove vocabulary. If English isn’t their first language, the nuance of a complex word may be lost anyway.

7) Reflection on tool effectiveness and updates

AI tool use for me is iterative, evaluative, and quietly strategic. I like the collaborative nature, even if it sometimes takes me longer. I ask: Did this help? Did it distract? Did I use a prompt worth saving? How could this tool have given me the result I was looking for?

I go into the backend of custom tools. I update instructions. I shape the tool to stay in charge of outcome. This part of the process—quiet, thoughtful, behind-the-scenes—requires deep reasoning no one sees.

8) Final accountability

I read, re-read, and revise. Sometimes there are seven drafts. I draft, AI polishes, I make it mine, AI checks for clarity, I revise again. I’m the one signing my name. Tools don’t carry consequences. People do. If I wouldn’t stand behind it in public, it isn’t ready.

The elephant in the room

If the elephant in the room is that “everyone is using AI and no one is talking about it,” then maybe the way through is naming what we are doing - and learning to do it with care. Practice varies. Self-assessment has limits. That’s why we need open conversations, and in some contexts, shared norms of practice.

If you’re using AI, name your process. Hold accountability. This is how we move AI from shortcut culture into a culture of care.

So when someone asks, “Did AI write that?” I don’t defend the tool. I explain the work: the brief I wrote before I began, the tabs I opened to check what was real, the language I revised so a person felt seen and not scrutinized. The multiple revisions and bias checking. The choice to anonymize, to ask consent. And that when I sign my name, I take responsibility.

Drafting can be accelerated. Discernment cannot.

I do not think this is less work.

I think it is different work, and it matters.