Language Learning
I Analyzed My Own 9 Articles to See If My English Actually Improved
I measured word count, vocabulary, and grammar errors across my first 9 articles to find out whether my English actually improved. The results surprised me.
I Analyzed My Own 9 Articles to See If My English Actually Improved
By Jinwon Jang
10/7/2026
Why I Stopped Trusting My Feeling of "Getting Better"
I grabbed my coffee and lay down on my bed. I opened my laptop to write my article. I had carefully chosen the topic and created an outline. But when I started writing my 10th article, I immediately closed my laptop.
My brain was filled with uncertainty about my English progress. No teacher. No outside feedback. No school. I had reached the point where trusting my own feelings to measure my progress no longer worked. I needed real feedback from the public.
I stopped trusting my feeling of "getting better" because, unlike other subjects, English progress is hard to measure. In writing especially, subjective assumptions don't give accurate data. So I decided to go back and check real evidence from my previous articles instead of just assuming.
The Method
Before I started this ESL project, I was skeptical and uncertain about my writing. Could AI realistically improve it? Would I still need a real class? This project isn't a formal study, and I'm doing it completely independently, so I worried that my own subjective judgment would get in the way of my progress.
I opened my previous articles on my website. I didn't want to rely on my feelings to measure my progress, so I had AI help me measure three things: word count, vocabulary level (the share of longer words, 7 letters or more), and grammar errors (tense, articles, agreement, and spelling typos).
There was one problem. Every published article had already been polished, so the published text couldn't show my real progress. So I split the work in two. I measured word count and vocabulary on my published articles, and I counted grammar errors in my raw drafts, before any polishing.
What the Data Actually Showed
My word count fell from 1,383 words to 826. That was disappointing, because I want my articles to get longer over time. Writing good-quality sentences matters, but so does holding a clear thought across a long paragraph, and that is what I was aiming for. My best explanation is that I was focused on publishing more articles instead of writing better ones. I can't prove that with this data, but my three shortest pieces are my three most recent (#7, #8, and #9). It's a fixable problem, and I now know what to work on: writing more sustained paragraphs.
Next, I measured vocabulary level. Plain, simple words deliver a message best, but I still catch myself wanting to sound more professional and sophisticated, which is the exact habit this series is about breaking. The result was basically flat: long words made up 27.8% of my first article and 28.8% of my ninth. Because every published article was already polished, this number says little about my unaided vocabulary.
The last measure was sentence length, which rose from about 11 words to about 15. The jump came right when the polishing started (10.7 in #4, 17.8 in #5), so part of the rise probably comes from the polishing, not from my own growth. I can't tell how much.
The most encouraging result came from my raw drafts. Their grammar errors fell from roughly 11 per 100 words to roughly 2 to 5. These are rough hand counts from small samples, so I treat them as a direction, not a precise number.
Looking at the data, instead of trusting my feelings, showed me which parts of my writing have improved and which still need work.
What the Data Can't Tell Me
Across the nine articles, the data showed some progress and some problems. The errors in my raw drafts clearly dropped. But my word count fell, my vocabulary stayed flat, and the rise in sentence length may come from the polishing. These measurements are useful, but some things can't be counted. Does my audience find my writing interesting?
I believe a writer's job is to make the writing enjoyable enough that the audience doesn't leave. That is a subjective measure, because everyone has different preferences for voice and ideas. This is where I ask AI for feedback. I ask it to compare my writing with professional articles from The New York Times or The Economist, which shows me room for improvement and styles I might borrow.
This is the hardest part of writing. Today, anyone can produce error-free writing with AI tools, but developing your own voice is still hard. So I treat AI's feedback as a guideline, not a second opinion. It has told me several times that I'm improving, but my audience is human, not an AI.
That is why I keep publishing on my website: I want feedback from real readers. Sadly, no one has read my articles yet, and I haven't received any comments.
What's Next
The data changed what I need to aim for in my next articles. Measuring my writing showed me which areas to work on.
First, I need to make my articles longer. Short, high-quality writing usually beats long writing with no meaningful message. But at some point, a piece that is too short can't hold the message clearly. From now on, I'm aiming for 1,200 to 1,500 words in every article.
Second, I need to focus on the quality of my articles, not the quantity. I was so focused on publishing a lot that I felt pressure to keep writing, and that pressure drained my motivation and my enjoyment. Unlike my school assignments, my articles have no deadline. My solution is to write fewer articles with more depth.
Honestly, I got exhausted and stopped writing. That is why I went back to my earlier articles and checked my progress: I wanted my motivation back.
After checking, I still don't know whether I should trust the data more than my feelings. As an ESL learner, my ultimate goal is to express my thoughts in my own voice, not just to write a perfectly clean paper. Error-free grammar matters, because it prevents the wrong message from reaching the audience, but it isn't the reason I started this project. I'll keep tracking my progress and see whether I can reach that goal. I'll run the same measurements again after my next three articles.
FAQ
Can you really measure writing progress with numbers?
Yes and no. Numbers can't tell me whether my voice is strong or whether readers enjoy my writing. But at my current level, they give me something concrete to check instead of relying only on my feelings.
Is word count a good measure of improvement?
It depends on the topic. For my portfolio, it matters, because a longer piece shows that I can sustain my message long enough to develop it fully. For a short, clear message, fewer words can be better. But my word count fell while I was rushing to publish more articles, so for me it works as a warning sign that I am cutting my ideas short.
Why did you measure your raw drafts instead of your published articles?
Because anyone can polish writing with AI. My raw drafts show my unaided writing ability.
Can AI feedback replace human readers?
No, because my real audience is human. But AI gives solid feedback, and it has helped me a lot. It catches my grammar errors and shows me how professional writers work, but it can't tell me whether a real person finds my writing interesting. Only readers can do that, and I'm still waiting for my first comment.