How to Use AI in Your Marketing Without Sounding Like a Robot

Mano escribiendo a pluma en un cuaderno junto a una laptop con formas luminosas abstractas

Comparte este artículo

We’ve all read that text: it opens with “in today’s world,” lists three benefits with the same structure, and closes with a sentence that says nothing. Nobody set out to write it that way. It’s what comes out when you ask an AI to “write a post about my business” and publish it as is. The tool isn’t the problem; the starting point is.

At graphemics.net we’ve published 579 articles (figure from the site’s API as of November 16, 2026), and AI plays a part in much of the work. Here’s what we’ve learned so that it helps without the machine showing through.

1. First write how you talk, then ask it for anything

The most common mistake is starting with the request. The first step is to give it material of your own: five texts that sound like you (an email to a client, a post that worked, the explanation you give out loud) and ask it to describe how you write: sentence length, words you repeat, what you avoid, whether you use a casual or formal register.

That description is your voice guide. Save it and paste it at the start of every request. Without it, the AI writes with the internet’s average; with it, it has a reference point that isn’t the average.

2. Give it what only you know

The AI knows what’s published everywhere. It doesn’t know your business. What makes a text recognizable is exactly what it lacks:

  • A specific client case, without names if needed.
  • A figure of your own, verifiable.
  • An opinion someone in your industry would argue with.
  • A mistake you made and what changed afterward.

If your request doesn’t include at least one of these four elements, the result will be generic by definition. Two paragraphs of your own data are worth more than twenty lines of style instructions.

3. Ask for drafts of parts, not the whole text

An article requested in one go comes out with the most predictable structure possible. It works better to split the work:

  1. You decide the central idea and the order.
  2. It proposes three ways to open; you pick one or write your own.
  3. It develops one section at a time, with your data pasted in.
  4. You rewrite the opening and the closing, which are what show the most.

Whoever keeps the decision about what gets said and in what order keeps the authorship. The rest is drafting, and drafting can be delegated.

4. Use it to think, not just to write

Where AI delivers the most in marketing isn’t in the final text. It’s before and after:

  • Before: ask it to challenge your idea, to say what objections a skeptical customer would raise or which questions remain unanswered.
  • After: ask it to read your draft and flag where it loses the reader, which sentence is unnecessary and which claim needs proof.
  • With data: paste in a table of campaign results and ask for patterns. You verify the numbers; the AI saves you the first read-through.

With that kind of use there’s no machine-written text to detect, because you wrote the text.

5. Run a filter before publishing

Even with good inputs, traces remain. Check for these before hitting publish:

  • Textbook openings: “in an increasingly digital world,” “now more than ever.”
  • Lists that always have three items, with sentences of the same length.
  • Closings that repeat the title in other words.
  • Big adjectives with no data to back them up: “incredible,” “revolutionary.”
  • Figures you can’t trace to a source. If you don’t know where they come from, cut them or turn them into a range with a disclaimer.

The last one is the most serious. An AI can write a statistic with complete confidence and be wrong. Publishing it under your name is your responsibility, not the tool’s.

6. Read it out loud

It’s the cheapest and most reliable test. What you wouldn’t say in front of a client shows when you say it: sentences that run too long, connectors nobody uses in conversation, a politeness that isn’t yours. If you stumble or it sounds like an ad, rewrite that stretch in your own words. Two or three changes per paragraph are enough for the text to sound like a person again.

A before-and-after example

Weak request: “write a post about the importance of SEO for restaurants.” Result: three paragraphs interchangeable with those of any other business.

Request with material: the voice guide pasted above, plus three data points (“70% of our reservations came through the map, not the website,” “we uploaded the menu photos and got more calls that week,” “a customer told us they couldn’t find us by the neighborhood name”) and the instruction to open with the third. Result: a text that already has a story, a voice and something to say. You adjust it in ten minutes instead of rewriting it from scratch.

The numbers in that example are illustrative: in your text, use your own, verified.

If a whole team uses it

When several people produce content, the risk is that each one uses AI in their own way and the brand sounds different on every channel. Three agreements are enough: a single shared voice guide, a short list of banned phrases and openings, and one person responsible for the final read before publishing. That avoids the catalog tone and keeps the speed.

What you shouldn’t delegate

There are parts of marketing where AI helps little and the risk is high:

  • Testimonials and reviews. Making them up deceives the reader, and it’s also easy to spot.
  • Replies to upset customers. That’s where it shows whether there’s someone on the other end.
  • Positioning. What makes you different is a business decision, not a text.
  • Any legal, tax or medical information without review by someone who knows the field.

A minimal workflow to start this week

  1. Gather five of your own texts and put together your voice guide on one page.
  2. Pick a single type of content (for example, the weekly post) and work only with that for two weeks.
  3. For each piece, write down beforehand three data points of your own that must appear.
  4. Run the filter from point 5 and read it out loud.
  5. Note how long it took. If you saved no time and the result didn’t improve, adjust the voice guide, not the tool.

In summary

AI sounds like a robot when you use it as a substitute for your judgment. It sounds like you when you use it as an assistant that drafts, questions and organizes, with your voice and your data as raw material. The difference between the two isn’t in the tool you choose but in what you put in at the start and what you review at the end.

If you’d like us to review how your marketing team is using AI, or to build your voice guide with you, send us a message and tell us about your case. And if this article was useful, share it with someone else who produces content at your company.

Versión en español: Cómo usar IA en tu marketing sin sonar a robot

Suscribete a nuestro Blog

Actualización de contenido de notícias

Más Notícias de Marketing, Inteligencia artificial y automación

¿Necesitas una Concultoría?

Envíanos un mensaje que agendamos una cita gratis

CTA