We Still Don't Know How to Optimize for AI — And That's Okay

Optimizing for AI? Take it easy.

There is a lot of talk—definitely too much—about how to “optimize content for AI systems.” Every week, new formulas and supposed best practices emerge for appearing in the responses generated by large language models or AI-integrated search engines.

But let's be honest: this anxiety to teach something that has barely come into existence sounds premature, to say the least—not to mention presumptuous.

As Aleyda Solís, an international SEO authority, rightly noted, everything is still very new. We are facing something that is constantly taking shape. In this kind of landscape, any definitive formula risks being outdated by next month.

The mistake lies in trying to control what we don't yet understand

The rush to find a new optimization “manual” ignores one simple fact: the AI systems themselves are still in the process of learning and adapting.

Trying to anticipate the behavior of tools that aren't yet stable, and whose parameters frequently change, is like trying to predict the shape of the clouds two days from now based on today's wind.

Generative search systems, for example, are learning from human interactions. They change their display criteria, their models, and their answers. The guidelines are still vague—and often even contradictory.

What actually still works

If there’s anything solid in all of this, it’s what we should have been doing for years: Useful, relevant, well-structured content that clearly and empathetically answers your audience's questions. This withstands cycles, algorithm updates, and even platform shifts.

If you publish consistently, answer real questions, understand your audience's language, solve concrete problems, and deliver value clearly, then you are already on the right track. You aren't behind. You don't need to completely reinvent yourself with every new headline-making update.

It’s not the time to rush — it’s the time to observe

Adapting to new technologies shouldn't be done out of anxiety. It requires observation, active listening, and continuous analysis. Yes, look at the data. Track search behaviors, consumption patterns, and the signals your audience is sending.

It is important to monitor clicks, conversions, sales, time on page, engagement, and what your competitors are doing—just as we always have. But do it without falling into the trap of blindly following every new trend.

What AI (still) doesn't know how to do

Even as language models evolve, something remains out of reach for automation: the authentic human experience.

• Knowing what to say when someone is afraid.

• Explaining something complex with simplicity, without talking down to the reader.

• Adapting an explanation to the local context, the historical moment, or the audience's culture.

These subtleties still escape pre-trained models. And they are what generate trust, empathy, and true connection—the kind of things readers (and customers) don't forget.

Arrive slowly. But arrive right.

As Pacote—a caricatured but highly popular figure on Instagram and TikTok—would say, in a language that echoes more between the lines than in algorithms: "arrive in slippers" (take it slow and steady).

The internet, and now AI systems, have always favored what is consistent, clear, and useful. What doesn't force its way in. What arrives slowly, but stays.

There is no urgent need to rush to become “AI-optimized.” There is, however, an urgent need to continue being useful. To be easily understood. To keep a watchful, but calm, eye on things.

Because, in the end, what survives updates and outlasts cycles is whatever remains necessary and useful.