If the model reads first: GEO as a double optimization

These days, the first reader of a corporate text is often a machine. Anyone who wants to be featured in AI responses must cater to two audiences at once: people and the models that now act as intermediaries between communications teams and their stakeholders.

This article was written by: Martin Regnet

Open your latest press release and ask yourself a simple question: Who read it first? Before October 2022, the answer was: “A human!”—a journalist, an analyst, or a customer. Today, the first reader is most likely a crawler, and the second is a language model that decides whether and how your text will ever reach a human. These days, your target audience often turns first to Perplexity, ChatGPT, Claude, or Gemini—and only then, if at all, ends up on your website.

This goes to the very heart of communication. When the primary reader is no longer a human being, a second question is added to the old one—“How do I convince my audience?”—“Will I be understood by the entity that stands between me and my stakeholders?”

The majority of readers are already automated

The figures are uncomfortably clear for advocates of the status quo. In June 2026, 57.4 percent of all global HTTP requests for HTML content came from bots, not humans (source: Cloudflare). For the first time, the majority of the online audience consists of machines—and this happened about eighteen months earlier than some widely cited forecasts from early 2026 had predicted, which had estimated the timeline for late 2027. A significant portion of this growth is likely attributable to Agentic AI. Traffic from these autonomously operating systems has grown by 7,851 percent year-over-year (source: HUMAN Security), albeit from a naturally small base, as Agentic AI still played a minor role in 2024.

These rapidly emerging trends underscore the fact that a growing portion of what happens to your content takes place before a human even sees it. This machine-driven process does not replace the human reader; rather, it curates and influences the reader’s perception. At first, this may feel like a loss of control for communications teams, but those who take it seriously will regain control.

Four Ways to Connect with Your Audience

Until now, communication has followed one dominant path: person-to-person. The press conference, the phone call, the off-the-record conversation with a journalist. That path hasn’t disappeared, but it’s now just one of four.

Face-to-face (H2H) interaction remains essential, but it is now less common as the first point of contact.

From Person to Agent (H2A): Your stakeholder checks a model first, not your IR page. The initial contact takes place at a level you don’t control.

Agent-to-Human (A2H): A model summarizes, weights, and makes recommendations before handing off to a human. Synthesis replaces search; curation replaces the channel.

Agent-to-Agent (A2A): Models pass information on to other models without any human intervention.

These days, there’s an increasing likelihood that at least one automated system stands between you and your stakeholders. It needs to be served first before a human sees your content.

The model plays a role in the decision

A channel conveys information neutrally. The newspaper prints what it prints; the newsletter sends out what you write. A language model is not a channel, simply because it selects, condenses, rephrases, and synthesizes. It decides which source to cite, which aspect to emphasize, and which to omit.

This means that a machine is taking on a role that used to be the responsibility of editorial teams: that of the gatekeeper. This gatekeeper does not follow any editorial line that is familiar or can be addressed. It uses probabilistic mechanisms and follows a selection logic based on credibility, consistency, and structural clarity. Anyone who ignores this logic risks being omitted from the AI’s response, even if not a single word is incorrect.

Two Readers, One Text

The discussions held during the CommTech Summer School on July 8, 2026, revealed a widespread reaction: The more AI takes over, the louder the call to return to what is human becomes—authenticity, creativity, attitude, and a distinctive personal style, rather than AI-generated drivel and uniformity. This reaction is absolutely correct. These qualities will actually become even more important in the future because they are the only things that cannot be automated. However, the discussion often obscures the fact that while this return to human values affects the communications team as the sender, it affects only one of the two categories of recipients.

What’s new is the second target audience. Today, corporate communications are aimed at two audiences simultaneously, each with demands that are sometimes contradictory: people who evaluate content based on emotional, social, and contextual factors; and LLMs and AI agents that algorithmically select, condense, weigh, and disseminate information.

Both follow their own logic. People respond to storytelling, emotion, credibility, and identification. Models need clarity: structured, consistent, quotable information and clear signals regarding reputation and authority. Therefore, something that is easy to read is not automatically easy to extract. An elegant, allusive paragraph may be brilliant to a human but useless to a model because no clear, quotable statement can be derived from it.

However, these two logics are not clearly separated. The very signals that demonstrate human authority also carry significant weight in the models: thought leadership articles, industry rankings, citations in the media and studies, and speaking engagements at relevant events. When a person collects such evidence, they build credibility, and that same evidence feeds the model with clear signals of authority. A reputation built by humans is thus also a machine-readable ranking signal. The picture is therefore more complex than two strictly separate audiences: What convinces one makes you quotable to the other—at least if the signals can be clearly interpreted.

For GEO, this means a twofold optimization. In the future, successful communication must work for both: human perception and machine interpretation. It is not enough to focus solely on SEO, nor is it enough to rely solely on human creativity. We need content that convinces both audiences at the same time: human enough to build trust and foster identification, and machine-readable enough to be presented in the first place. Both audiences want to read the same text, each in their own way.

What this means for communicationmeans for the communications team means

The profession is expanding. For decades, the job was to convince people. Now a second task has been added: to be understood by the machines that, in turn, convince people. This second task falls under the realm of communication, not IT. It’s about meaning, wording, and narrative—not code.

For decades, the central question in communication was: How do I reach my target audience? While this question remains valid, it is no longer the top priority. Today, a second question takes precedence: Am I understood by the intermediary who stands between me and my target audience?

Anyone who takes this seriously will continue to write for people while also making sure the algorithm lets their content reach them. The best time to start doing this was with the last model update. The second-best time is the next piece you publish.

GEO workshop at the CommTech Academy

In the CommTech Academy’s four-hour GEO workshop, Martin Regnet explores how to effectively manage your reputation in the age of AI. The next workshop will take place on August 12 . For more information, click here.

MORE INFORMATION HERE

OTHER BLOG ARTICLES