How I Got My AI Score Down From 60% to 5%?  

Nobody can tell if you created content with AI. They can only guess. Even AI detector tools are not 100% reliable for detecting AI.Back in 2023, Weber-Wulff et al. evaluated the top 14 detection tools. All of them scored below 80% in accuracy, and only 5 were over 70%.As for humans, we are correct only 54.64% of the time, according to Nexcess.

I always write by myself. Even if I take ideas from AI sometimes, I present them in my own words. But still, AI detectors flag some of my sentences as AI.

That’s because AI checkers try to recognize AI writing patterns in your content. They do look for the origin of your words.

Just like recently, I ran my blog post through an AI detection tool. I wrote it myself and thought it sounded pretty natural. But then, my AI detector hit me with a 60% AI score.

That moment made me pause and think. What else should I do to make it feel like it came from me?

How I got my AI score down from 60% to 5%

I took it as a challenge. Over the next couple of hours, I broke down the process. And rewrote the whole article with the help of some tools.

After that, the same content that scored 60% on the AI content detector came down to just 5% AI. While the core message stood intact.

Following is the process and the tools that I used to humanize my content.

1. Identifying the robotic sentences

The first thing I did was to run the tests again with another sophisticated AI-detector.info tool to check if there was any mistake in the results. It was not; the results were still at 60.74%.

But a good thing was that it highlighted all the sentences with an AI similarity. So it was easier for me to read them and figure out why they were flagged.

The difference between normal and highlighted sentences was their readability and tone. AI sentences were longer, complex, and harder to read. In simple words, robotic.

I even rechecked my assessment through a readability checker by checker-plagiarism.com.

First, I entered the AI part, and the results said that they actually had very low readability. Plus, they have some filler words or fluff in them that make the sentences extra long.

Then I checked the human part, and I was right that its readability was much better.

So, now my job was to rewrite AI sentences in the way we humans normally talk. Means in a natural or conversational way.

2. Rewriting in a natural tone

After pointing out the core problem, I collected all the highlighted text and tried rewriting it with the humanize AI text tool.

This tool from paragraph-generator.com has many different modes, like simple, personalized, PRO mode, and first person.

I tried them all one by one. All of them made my content better, but I liked the default one the most.

Because it made my sentences simple, shorter, and more readable. In simple words, it humanized them.

So I ran the tests again, and this time, my content was much better than before. It came out with only 18.50% AI.

However, more work was required on it. Some sentences were again highlighted for AI. Before rewriting them again, I realized that if I want to make it look more like me, I have to add something of myself: the human touch.

So I started working on that.

3. Adding personal thoughts and transitions

I read the whole passage once again and wrote some sentences from my perspective. I added the knowledge I had on the topic.

See, AI content is always generic, based on public opinion.

But our opinions develop through our personal experience and consciousness, which AI lacks. AI cannot go outside, meet people, and develop an opinion. Only we do that.

So, I added whatever I had learned or discussed during my lifetime about my topic. My topic was militarization, but I am not from the military. My whole experience with it came from reading and discussing with people around me—like troops, military veterans, and army enthusiasts.I even used their views and anecdotes in my content.

That made my content more relevant and opinion-driven. Now it looked even more humanized, so I checked it again.

Again, the results got even better, but not completely how I wanted. Some of my newly written sentences were also highlighted, so I decided to rewrite them in a new style or tone.

4. Improving the tone consistency

Once again, I read the whole content and observed that the new sentences do not match the old ones in tone. Plus, some fluff words were still there.

That was one problem, and the other was some sentences that were constantly being flagged as AI even after humanizing.

So this time, I used a worchanger.net to rewrite with tone consistency. As this tool offers many different rewriting tones, I tested them all, such as professional, fluent, first-person, casual, and academic writing.

The tone that matched my writing style was “fluent.” So I kept that result and replaced the highlighted sentences with them.

After that, I looked for any extra words in my passage. My goal was to remove any word that does not add anything. Removing which does not change the meaning or intent. I looked for them thoroughly and removed every single one of them.Then I ran it again through the AI detector.This time, I got the results I wanted to see. Only 5.03%! That was it. I was happy with that effort and published the article.

Common mistakes people make when trying to lower AI scores

When I was working on reducing my AI score, I noticed one thing: most people mess up because they focus on the wrong fixes. Instead of making their content feel natural, they just try to “trick” the detector. That is not the right approach for both SEO and the target audience.

Here are some common mistakes:

1. Rewriting blindly

Most writers just keep running their AI text through humanizers again and again.

Sure, it changes words, but the tone and structure stay the same. Plus, it may change the intended meaning. Therefore, AI detectors still catch it because the flow doesn’t feel human.

You need to work on your sentence structures.

2. Overloading synonyms

Some writers believe replacing every word with a synonym solves the problem. But it doesn’t.

Too many synonyms make sentences feel unnatural. For example, changing “The idea was good” into “The concept was meritorious” sounds even more robotic.

You need to replace words with their simpler alternatives, only if they do not change the meaning of your sentence.

3. Only caring about the score

That’s the biggest mistake. Your focus should not be only on lowering the AI percentage. Also, take care of readability. Even if your score is 0% AI, but the text is hard to read, your audience won’t care.

At the end of the day, you’re writing for people—not tools. Exactly what Google wants you to do.

4. Ignoring personal input

AI writers are trained not to take sides or offend anyone. Hence, they always stay neutral and narrate general opinions.

For example, ask DeepSeek AI about Taiwan. It will say that it’s a part of China, not a separate country. Because it is trained by the Chinese, and that’s the general opinion there.

On the other hand, we humans have different likes and different perspectives. Because of our unique experiences and environment. We have opinions that might not completely align with those of others. But that’s what makes us unique and our content personal, which in terms of content is called humanized.

Conclusion

So, that’s how I brought my AI score down from a worrying 60% to just 5%.The trick wasn’t just rewriting blindly; it was understanding why the content felt robotic in the first place.Once I made everything flow naturally, I added my personal touches. Then AI detectors started agreeing that the content is human.So, do not panic if you’re also struggling with AI scores as well. Be patient and use the right tools. And do not forget to add your own unique voice. After all, no AI can truly replicate you.