We think we can recognize artificial intelligence. And that's exactly the problem

Today, artificial intelligence does more than just help with searching for information or translating texts. It writes articles, creates advertising campaigns, and generates photos, videos, and social media posts. It can come across as professional, convincing, and often surprisingly human.

The biggest change that AI has brought to media and marketing, however, may not be the technology itself. Rather, it is the loss of the assumption that a human being is behind every piece of text, image, or video. We are no longer automatically aware of who created the content we read, watch, and share every day. And with that comes a question that is more important than any new tool: Who is behind the content, and who is responsible for it?

Research conducted at the Faculty of Mass Media Communication at UCM in Trnava shows that the answer is not as simple as most people think. In an interview, Dr. Michal Kubovics discussed the findings of a project titled The Creation and Reception of Artificial Intelligence-Generated Content by Target Groups in Media Studies.

Can we still reliably distinguish between what was created by humans and what was created by artificial intelligence?

Honestly, without clear labeling, we can’t do it in most cases. People feel like they can spot AI-generated content right away. They’ll say, “This is too polished,” or, conversely, “This sounds strange—it must have been written by a machine.” However, research has shown that when people relied solely on intuition, their success rate in recognizing AI-generated content was approximately 36 percent. This isn’t a failure on the part of individuals. It’s proof that our intuition was shaped in a different digital world.

A few years ago, automatically generated texts could often be identified by their clunky style, repetitive phrasing, or illogical sentences. That’s no longer the case today. Generative AI can create linguistically sound, well-structured, and professional-sounding content. It can mimic technical language, a marketing tone, formal communication, and even emotional tone. And that is precisely the problem. We naturally associate a professional appearance with credibility. But a text can be error-free and perfectly structured and still be inaccurate, incomplete, or misleading.

So what helps people the most?

Transparency. When content was clearly labeled as having been created with the help of artificial intelligence, people’s success rate in recognizing its origin increased to approximately 97 percent. This shows that labeling AI-generated content is not just a formality tucked away somewhere beneath a post. It serves as a guide. People don’t need to be told what to believe. They need to be able to make informed decisions. And to do that, they need to know how the content was created, who was involved in its creation, and who is responsible for its publication.

Transparency, however, does not mean that everything created with the help of AI is automatically bad or inferior. Artificial intelligence can be a useful tool. It can save time, assist with research, analyze data, generate proposals, or speed up routine work. The problem does not lie in the use of the technology itself. The problem arises when someone begins to hide behind it.

Does labeling content as "AI" mean that people automatically trust it less?

Not always. And this is where an interesting paradox emerges. People often rate texts created by artificial intelligence as professional, expert-sounding, and well-written. Sometimes they may even appear to be of higher quality than texts written by humans. At the same time, however, people perceive them as less engaging, less personal, and less memorable. They are less inclined to read them all the way through, and they resonate less emotionally with them.

AI can create content that looks high-quality. However, it doesn’t always create content that leaves a lasting impression on people. Professionalism, after all, is not the same as persuasiveness. A powerful text isn’t created solely by correct sentences, the right tone, and good structure. It also comes from experience, observation, perspective, the courage to say something unexpected, and the details that a specific person has noticed.

Artificial intelligence can process language. However, it has no memory of its own, no personal history, and no real experience of the world. And although readers may not always be able to put their finger on it, they often sense this absence.

What does this mean for marketing and the media?

AI shouldn't replace creative thinking. It should support it. The worst-case scenario is when an organization starts producing more content but less value. More posts, more campaigns, more text—but at the same time, less originality, fewer ideas, less trust, and less accountability.

Artificial intelligence is excellent for initial drafts, research, summarization, data analysis, or routine tasks. It can help eliminate blank pages in a document, speed up the preparation of supporting materials, or suggest alternative phrasing.

However, humans must focus on what matters most: selecting topics, interpreting information, fact-checking, editing, deciding on the tone of communication, and, above all, assessing whether it makes sense to publish the content at all. AI can prepare an initial draft. Humans must provide a reason why it should be of interest to anyone.

People are concerned about misinformation. Does this fear also influence how they perceive specific content?

Not as directly as we might expect. Many people perceive AI-related disinformation as a serious threat. They are concerned about fake news, manipulated videos, artificially created photos, and fraudulent messages sent in the name of well-known figures. At the same time, however, it appears that the mere fear of misinformation does not automatically lead to greater caution when encountering a specific post, article, or advertisement.

When people evaluate content, they often base their decisions on other cues: whether they recognize the source; whether the text comes across as professional; whether it aligns with their own experiences; and whether it confirms something they already believe. That’s why it’s not enough to simply warn against misinformation. It is more important to build an environment where trust is evident directly in the communication—an environment where it is clear who is speaking, why they are speaking, what sources they are drawing on, and who bears the consequences for errors or misleading context.

Isn't labeling AI-generated content just a partial solution?

It is. Labeling alone won’t solve everything. It cannot replace critical thinking, quality journalism, media literacy, brand ethics, or the responsibility of authors. However, it is of fundamental importance because it provides people with information that they often cannot reliably obtain on their own without such labeling.

The label should not serve as a warning: “Beware, this was created by AI.” It should serve as a fundamental marker of fair communication: “This content was created with the help of artificial intelligence, but someone reviewed it, approved it, and takes responsibility for it.” That is the difference between technology we fear and technology we can reasonably trust.

How do people who work with AI every day view it?

Interestingly, experienced AI users are not typically uncritical enthusiasts. The more they work with the technology, the better they understand both its benefits and its limitations. They know that it can save them time and assist with design, research, analysis, and repetitive tasks. At the same time, however, they realize that without rules, transparency, and human oversight, the technology can undermine trust in the entire communication environment.

So reality isn't divided into two camps: those who want AI everywhere and those who would prefer to ban it. Many people want to use AI. However, they want to know where its limits lie and who takes responsibility when something goes wrong.

Why must the last word always belong to a human being?

Because content isn’t just text, an image, or a video. Content can influence purchasing decisions, a person’s reputation, their relationship with a brand, trust in a public institution, and views on social and political issues. AI can generate compelling output, but it does not take responsibility for any errors, inaccuracies, insensitivity, or harm it may cause. The person or organization that publishes the content must bear responsibility for it.​​​​​​​

For the media, this means fact-checking and editorial oversight. For brands, it means taking responsibility for the tone and accuracy of their communications. For public institutions, it means not using AI as an excuse along the lines of, “The system generated it that way.” Trust is built when it’s clear who stands behind the words and who is prepared to vouch for them.​​​​​​​

What is the main message of your research?

This isn’t a battle between humans and artificial intelligence. It’s a matter of rules, trust, and responsibility in a world where it will become increasingly easy to create convincing content and increasingly difficult to verify its origin. The future of media and marketing will not belong to those who can produce the most text, images, and videos. It will belong to those who can create content that people understand, can trust, and know who is behind it. After all, the strongest currency of the future will not be the quantity of content. It will be trust.

This article was produced as part of Project No. 09I03-03-V04-00367, supported by the European Union through NextGenerationEU funding as part of the Recovery and Resilience Plan of the Slovak Republic.