Full article

How to write so it doesn’t show that we used AI. Guide so your text sounds like you

Letters, claims, legal filings, recommendations, etc.

Using artificial intelligence to write is now as normal as using a spellchecker or a calculator. The problem is not the tool, it is the stigma: many people hear “AI” and think of cold, generic, soulless texts. But reality is different: in most cases, the person decides what to say and the AI only helps to execute it.

So, what should you do so that your final text does not sound like AI, but like you?
The answer is simple: it is not about hiding, but about humanizing:

Here are the keys.

1.     You lead, not the tool

The typical mistake is to ask the AI “write X for me” and paste it as is. That produces texts with repetitive patterns. Instead, use AI as a co-pilot in which you define the idea, the tone and the objective. The AI proposes a draft and you turn it into your version, because if you do not add your own criteria, it shows.

2.     Add human details:

examples, context and small real imperfections. Texts

that sound like AI are usually too “perfect” and abstract. To break this effect, you have to add concrete examples that have happened to you or that you know; this removes any suspicion, or add a colloquial phrase of yours, a brief personal observation, an emotional nuance such as “this is frustrating because… the most surprising thing is…”.

AI speaks in general; humans speak in situations.

3.     Break symmetry and the “manual” structure

AI tends to write with overly ordered patterns:
“First… Second… Third…” always the same.
You can: mix long and short sentences, add a rhetorical question, change the rhythm, and cut paragraphs where you would cut them. In essence, naturalness is not disorder, it is simply human rhythm.

4.     Rewrite the opening and the closing in your voice

Even if you leave the body more “assisted”, you change the first 5–10 lines and the final paragraph. That “marks” the text as yours, because the reader feels a person behind it from minute 1 and at the end.

5.     Remove inflated or “AI-neutral” phrase

There are expressions typical of automatic drafts:

“In this current context…”
“It is worth noting that…”
“It is important to point out…”
“In short…” repeated every two paragraphs
“There is no doubt that…”
Replace them with your usual way of speaking, so if a sentence sounds like a brochure, delete it — a simple but useful trick.

6.     Use your won key words

We all have good verbal habits; we often say phrases like:
“Let’s get to the point.” “The reality is that …” “in practice it’s proven that …”

Put some of these phrases throughout the text, not as a trick, but because it is your way of writing.

7.     Verify data and adjust statements

Sometimes what “gives away” an assisted text is not the style, but a strange fact, an overly confident generalization, or something that does not fit your knowledge.
That is why you should review figures, dates and names, correct exaggerations, and add the nuances you consider appropriate.

8.     Avoid a “Wikipedia tone” if your goal is to be close

If you are writing for citizens, remove excessive solemnity.
AI usually sounds like a manual; you sound like a person.
Bring it into everyday language without losing rigor, so that it is less abstract, more direct and above all more practical.

9.     Read it out loud (a foolproof trick)

If when you read it it sounds like a robotic speech, it is.
When a text sound good read out loud, it almost always sounds human.

10.  Very important: AI has its own style in the use of punctuation

Indeed, the way AI organizes, and structures text is very different from a human’s, because it overuses certain punctuation marks, for example replacing commas with dashes; an AI text is full of dashes, parentheses, quotation marks, colons, etc., also of constant classifications with numerations 1,2,3 … or

. the house is characterized
. the house ……
. the house…

Therefore, we always have to ask ourselves this question:
how do I make it clear that it’s me, even if a tool helps me?
If your text has your experience, your criteria, your examples, your rhythm and your words, nobody will think “this was done by a machine.” They will think: “this was written by someone who knows what they are talking about”, which is what really matters.

The key is to understand that AI is a co-pilot, not the author. If you ask, “write me an article” and paste the result, you will get a perfect but generic text. If you first define what you want to say, how you want to say it and with which example, AI only helps you to organize and polish. And that way, the final text sounds like you.
Let’s look at a real example, which is what best shows the difference.

Human text (your text):

“people increasingly turn to social networks and other sources because in Europe the media, as a consequence of subsidies, follow the same editorial line on issues that affect Europe; if we look for example at the European press, it does not explain the two versions of the war in Ukraine and Russia by informing from both sources, so that the citizen can then think and decide; no, it only gives you one partisan version; as a consequence people tend to think the opposite of what is reported in other media, since unanimities are not typical of democracies”

“AI-style” text about the same idea:

“In recent years a clear change has been observed in the information habits of European citizens. More and more people turn to social networks and alternative media because they perceive that much of the traditional press in Europe maintains homogeneous editorial lines, partly due to its dependence on public subsidies. This is especially noticeable in sensitive geopolitical issues such as the war between Ukraine and Russia, where many outlets offer a dominant narrative without presenting in a balanced way the different versions available. As a consequence, part of the population feels that it does not have all the perspectives necessary to form its own judgment and ends up seeking external sources to contrast. Paradoxically, this lack of plurality can generate the opposite effect to what was intended: the more uniform the official discourse is, the more citizens tend to distrust it and to sympathize with alternative readings of the conflict.”

Comparative comment (why AI is more noticeable and how to correct it without “camouflaging” it): First, your text is direct and conversational. You say “if we look at the European press… it doesn’t explain the two versions…”. You speak like a person addressing another person. The AI text, by contrast, starts like a report: “a clear change has been observed in habits…”. It is not wrong, but it sounds academic and distant. For an assisted text to keep your voice, it helps if the opening and tone are more human: “the reality is that…”, “if you compare outlets…”, “you see it clearly in…”.

Second, in your text there is explicit judgment: “one partisan version”. That strong wording, when it is your position, adds personality. AI softens it with phrases like “dominant narrative without balance”. That gives away a machine style because it avoids taking a side. If the text is yours, you don’t have to neutralize it: your criteria are part of its value.

Third, your rhythm is irregular, like real speech. You jump from idea to idea with energy. AI is too symmetrical: long sentences with formal connectors (“as a consequence”, “paradoxically”). That constant perfection is what many people identify as “AI text.” The solution is not to “hide it”, but to add human rhythm: vary sentence length, cut with a full stop when needed, include a rhetorical question.

Fourth, you use an immediate and controversial concrete example (Ukraine–Russia) to support your thesis. AI keeps it but wraps it in manual language. A text with soul adds a more visual detail, like: “if you read a European newspaper and then an alternative channel, it feels like they are talking about two different wars.” That is not camouflage, it is human experience.

The conclusion of the example is clear: what makes a text sound human is not that it has never passed through AI, but that it has your fingerprint. Your fingerprint is: position, real example, direct language, your own rhythm and your words. If you use AI, use it to organize and polish, but then recover your voice with two or three personal touches. That way “the AI” is not noticeable because what is noticeable is the author. And that, in journalism, in letters, in legal filings or in any communication, is what creates credibility.