There's a gap between the content that marketing teams publish and the content that the audience actually wants to read. It's not a problem of technical quality or keyword optimization. It's a problem of perspective: we write from inside the company, not from the customer's head.
The traditional solution is audience research. Interviews, surveys, review analysis, listening to sales calls. All of this is still valid and necessary.
The problem is that it doesn't scale. You can't do an interview every time you need to review a paragraph, validate an angle, or decide what tone to use in a new piece.
In-person GPTs don't replace that research. But they do allow you to consult it in real time, without waiting for someone to be available on the calendar.
A persona GPT is a language model configured with information from your buyer persona.
In practice, it's a version of ChatGPT where you upload everything you know about your audience, their objectives, their frustrations, how they make decisions, what language they use, what questions they ask, and then you can ask them questions as if you were talking directly to that customer profile.
It's not magic or an oracle. It is a tool for outsourcing the user's point of view in a quick and consultable way.
When you're editing an article at eleven o'clock in the evening and you want to know if the introduction connects to someone who has the problem described in the article, you can't call a customer. But you can ask the GPT in person.
The real value is in the iteration speed. You can try five versions of a headline, a different content structure, or a different CTA, and get a reaction from your audience's perspective in minutes.
A personal GPT is only as good as the information you upload to it. If the person is built on the assumptions of the marketing team, the GPT will reproduce those assumptions. If built on real data, the GPT will reflect real audience signals.
Before creating the GPT, you need to have the person well-founded. There are several data sources that work well for this.
Review mining is one of the most accessible. Reviews of your product or service, and those of competitors, contain real language from real customers describing their problems, expectations, and frustrations. What people write in a three-star review about what a product lacked is high-quality audience information that most teams aren't using systematically.
Sales calls are another underrated source. The questions that prospects ask before they buy, the objections they raise, the vocabulary they use to describe their problem: all of that is direct material for building a person with real depth. If your team records calls, reviewing them periodically should be part of the process of building and updating people.
Audience Research Tools like SparkToro, they allow you to explore what sites your audience visits, what topics they are interested in, what vocabulary they use on social networks. It is especially useful for understanding the information ecosystem in which your client operates, what sources he consults and to whom he gives authority.
With that information, the person should include at least: a basic demographic and professional profile, specific objectives, specific frustrations, the decision-making process, the language they use to describe their problem, and the questions asked at each stage of the buying process.
The technical process is simple. In ChatGPT, go to “Explore GPTs” in the side menu and click “Create” in the upper right corner. This opens the GPT constructor.
What makes a personal GPT work well is not the technical configuration, but the quality of the initial prompt. The prompt is where you explain to the model who is the person who is going to embody.
A well-constructed prompt for an in-person GPT should include:
You can paste screenshots of reviews, snippets of transcribed sales calls, or data from audience tools directly into the builder. The more real material you include, the more useful GPT will be.
In the “Configure” tab, you can define initial questions that make it easier to consult the GPT consistently. Some useful initiators for an SEO-oriented personal GPT:
Having these predefined initiators speeds up the workflow and makes GPT easier to use for people on the team who aren't used to prompting.
Before starting an article, you can describe to the GPT the topic and angle you have in mind and ask them if that connects with their real concerns. It's a quick check that can save you from typing a thousand words in the wrong direction.
The introduction is where we most often lose the reader. You can paste the introduction to the GPT and ask him if the problem he describes resonates with what he's experiencing, if the language feels familiar or distant, and if there's anything I would expect to see in the first few paragraphs that isn't there.
Once you have a draft, you can ask the GPT to read it and tell you what questions it had left unanswered. This usually reveals gaps that the author doesn't see because he already knows too much about the subject.
Calls to action are difficult to write because they require understanding what matters to the audience at that point in the process. You can show two or three versions of a CTA to the GPT and ask it which one would move it to act and why.
There is an important limit that is worth clarifying. In-person GPT doesn't replace contact with real customers, and their answers aren't evidence.
If the GPT tells you that an introduction doesn't connect, that's a signal to review, not a conclusion. If it tells you that a CTA works well, that doesn't validate that it's going to convert. It's a simulated perspective based on the information you uploaded to it, and that information may have biases or gaps.
The way to use it responsibly is to ask them to base their answers on the information you gave them. If he says that something doesn't connect, ask him what part of the person that conclusion comes from. If he can't cite a source within the material you uploaded, his answer is more speculation than analysis.
It can also hallucinate. If an answer sounds strange or contradictory to what you know about your audience, question it directly. The GPT can be corrected when an inconsistency is pointed out to it.
A person who doesn't update ages poorly. The same thing happens with the person's GPT.
Every time you learn something new about your audience, whether through an interview, a change in the market, sales feedback, or behavioral data on the site, it's worth adding that information to the GPT. The process is simple: go back to “My GPTs”, enter the person you want to update, and add the new information in the Configure tab.
A GPT of a person who feeds on real learning on an ongoing basis is significantly more useful than one that was built once and never touched.
SEO is moving toward a model where understanding the real intent behind a search matters more than the density of Keywords or the number of backlinks. Google and AI systems are increasingly evaluating whether content actually solves the problem of the person searching.
Producing that type of content requires an in-depth understanding of that person. And most marketing teams don't have the time or resources to do audience research every time they create or update a piece.
A well-built personal GPT doesn't solve that problem completely. But it makes the knowledge you already have about your audience searchable in real time, at the moment you're making editorial decisions.
That's enough to improve the quality of the content consistently, without adding too much time to the process.
.png){kind=small}
Consultoría SEO personalizada para escalar tráfico, posicionar tu marca, aumentar el revenue y dominar el mindshare en Google y modelos de IA como ChatGPT.
%402x.svg){kind=icon}
