My high school English teacher had a habit that used to frustrate me enormously. She would hand back our essays covered in questions. Not corrections. Not rewrites. Just questions, written in red pen in the margins. “What do you mean here?” “Why does this follow from the previous point?” “Is this the strongest example you could have chosen?” I remember thinking she was making her job easier by making mine harder. It took me years to understand she was doing the opposite of making it easier. She was doing the most demanding thing a writing teacher can do: she was actually trying to make us better.

That memory comes back to me now whenever I think about where AI writing tools are headed and what most of them are getting wrong.

Automation Is Tempting for Obvious Reasons

Nobody needs to be convinced that automated writing tools are appealing. The promise is simple: less time, less effort, usable output. For a content team under deadline pressure, for a professional who needs a quick report, for someone who genuinely dislikes writing and sees it as an obstacle rather than a skill worth developing, automation makes complete sense. I am not dismissing that. Those are real needs and real constraints.

But somewhere along the way, the logic of automation started bleeding into contexts where it does not belong. Into student essays. Into professional development. Into the work of authors, editors, researchers, and communicators for whom the quality of their thinking, expressed through writing, is literally what they are paid for. In those contexts, automating the writing does not solve a problem. It removes the work that creates the value.

The more conversations I have with teachers and seasoned authors on the topic, the more convinced I become that the last thing anyone needs is an automated writing service that can generate content for them. What they need is an application that will help them see their work objectively and improve upon it.

What Good Feedback Actually Does

Think about the best feedback you have ever received on something you wrote. Not the most flattering: the most useful. My guess is it did not tell you that your writing was fine. It pointed at something specific. A section where your argument went soft. A paragraph that buried the most important sentence at the end. A word you kept using that was doing too much heavy lifting. Good feedback is uncomfortable in a productive way. It shows you something about your own thinking that you could not see from inside it.

That is what revision is for. Not polishing: genuine rethinking. The kind of rethinking that happens when someone you respect points at a paragraph and says, honestly, I am not sure what you are arguing here. That moment of discomfort is not a failure. It is the mechanism through which writers improve. It forces you to go back in, understand why the paragraph is unclear, and figure out what you were actually trying to say. By the end of that process, you understand your own argument better than you did before. You have become a slightly better writer.

Automated generation produces none of this. It gives you a paragraph that sounds clear without requiring you to think clearly. The clarity belongs to the tool, not to you. And next time you need to write something, you are in exactly the same position you were in before.

The Long Game of Writing Development

Here is what I think gets lost in most conversations about AI and writing: people focus on the individual document rather than the trajectory of the person producing it. Whether this particular essay is good enough matters less, over a lifetime, than whether the person writing it is getting better at writing. The document is temporary. The skill is cumulative.

The student who receives honest and clear feedback on her writing skills for four years at school ends up in a completely different place from the one who figured out her own way during the same amount of time. She has gained something along the way, a more intimate understanding of her own mind and a better understanding of whether an idea is valid or not. The second student has a collection of documents that were never quite theirs.

This is true beyond school as well. Professionals who write regularly and get genuine feedback on that writing, from editors, from colleagues, or from any honest reader, develop a kind of writing intelligence that shows up everywhere: in emails, in presentations, and in the way they frame problems and communicate decisions. That intelligence does not develop through automation. It develops through the cycle of writing, receiving feedback, revising, and writing again.

The ChatGPT Comparison Worth Making

Most people by now have a reasonably clear picture of what ChatGPT does with writing. You give it a prompt, it gives you text. The text is often impressive, sometimes surprisingly good, and almost always faster than writing it yourself. For many tasks, that is genuinely useful. I use AI tools myself for certain things, and I do not think there is anything wrong with that.

But if you are a writer, a student, an editor, a researcher, or anyone who depends on the quality of your own thinking expressed in language, there is a real question about what ChatGPT is doing for your development versus what it is doing to it. If you are always generating rather than writing, you are not building the skill. You are just producing output.

The more interesting comparison is not ChatGPT versus doing nothing. It is ChatGPT versus a tool that actually engages with your writing as a reader would. If you want to understand that distinction in concrete terms, the comparison that Thanis draws against ChatGPT is worth reading. The core difference is not about features. It is about what each tool assumes you need: a finished product or a better process.

Feedback as Respect

I want to go back to my high school English teacher for a moment. What she was doing with those margin questions was actually a form of respect. She was treating us as people capable of figuring things out, not as people who needed to be handed correct answers. That assumption, that the writer is capable and that the job of a good reader is to help them realize more of that capability, is the foundation of serious writing instruction.

It is also the foundation of what a genuinely useful AI writing feedback tool should do. Not rewrite your sentences. Not generate cleaner paragraphs on your behalf. Read what you wrote, engage with it honestly, and tell you where it is working and where it is not, in a way that helps you go back in and make it better yourself.

That is what Thanis is designed around. This is the idea that your work should be taken seriously, your thought processes should be developed further, and AI’s proper role in all of this is that of a good critic, one that does not simply write for you but rather one who asks, “What do you mean by this?” Why does this follow? Is this really the best you can do?

The Harder but More Honest Path

Automation is easier to sell than feedback. It removes friction. Feedback adds it, deliberately, in the places where friction produces growth. That is a harder value proposition to explain, especially in a market that has become obsessed with speed and effortlessness.

But for the authors, teachers, and professionals concerned with the endgame, this isn’t news. These individuals have experienced firsthand the difference between the effort they’ve put into writing and revising something until it’s perfect and writing something just to write it and never think of it again. One of those experiences leaves something behind. The other does not.

Feedback is slower than automation. It is also, for anyone who actually wants to improve, worth considerably more.