When AI Starts Writing Our Emails, What Happens to Our Thinking?

Google has just taken another step toward integrating generative AI into one of the most popular communication tools in the world: email. Its Gmail AI Inbox aims to transform how we manage our emails. Instead of a list ordered by date, it provides an interface where AI filters, synthesizes and suggests what it considers relevant in our digital lives, and most notably, writes draft emails for us.

It’s a tempting proposition: less time wasted navigating endless threads, less stress from not knowing what to read first, less effort in writing clear and effective answers. These features, automatic thread summaries, the assistant that writes drafts and suggestions, along with task prioritization, are already part of the new Gmail experience, and will soon be available to its more than 2.5 billion users. But beyond the immediate utility, this technological wave raises deeper questions about what it means to outsource an activity as personal and habitual as written communication.

Human communication is more than a simple exchange of information. It is a cognitive exercise that structures thought, clarifies ideas and constructs meaning. Writing an email requires choosing words, organizing arguments, anticipating reactions and modulating tone and style. Each of these steps is part of the ongoing exercise of our linguistic and social ability. What the new Gmail proposes is shortcutting these steps, skipping them or delegating them to a model that synthesizes and generates emails for us. This change may make sense for routine emails, but when normalized, what remains of our writing, clear-thinking or critical judgment skills?

From a psychological and communicative perspective, the presence of an automated mediator alters the very nature of writing. Research into AI-mediated communication not only analyzes efficiency, but also how these tools reconfigure human relationships, attention strategies and personal expression. Artificial intelligence is not just a neutral filter: it also modifies what we perceive as important, what we prioritize, and how we represent our intentions to others.

Automating this can lead to a subtle but profound phenomenon: the atrophy of fundamental communicative skills. By letting a model synthesize the content of an email or compose coherent responses, we delegate a complex cognitive function. These are not mechanical tasks, they require language skills such as thinking, choosing or articulating, which make up the core of our reflective thinking. When we stop practicing these skills, they stop developing. We lose the ability to condense long ideas into clear syntheses, to detect important nuances or to mediate conflicts through well-chosen words. Can you do it satisfactorily well? Perhaps, but it is not us, nor does it have the nuances that, on many occasions, we want or need to contribute. It’s something else. It’s as if I stopped writing my articles: my regular readers would know immediately: quite simply it wouldn’t be the same, even if they were perfectly well written.

Beyond individual communicative competence, there are social and ethical factors at play here: the way we perceive those who write to us. If a message is not the product of the mind of the interlocutor but of a generative model, how do we interpret their sincerity or intention? The ethics of human communication are based on authorship and intention: by introducing AI as a mediator, we change the foundations of that social contract. Some studies suggest that AI-generated or AI-assisted messages alter how we perceive the sender’s character, weakening the indicators we use to infer human traits from their words. Reading the messages we usually receive from people we know and comparing them with automated ones, it seems difficult to doubt it.

Of course, there is a valid argument in favor of using AI in administrative or repetitive tasks, and we all understand it: freeing up time for activities of greater cognitive, creative or interpersonal value. But this gain in efficiency should not blind us to the fact that outsourcing complex cognitive activities has an invisible cost. The question is not simply whether AI can do it, which it surely will sooner or later, but whether we should substitute a task that largely defines our professional and social life. When we stop practicing a skill, we soon lose it. Reading carefully, synthesizing an argument or responding with a minimum of empathy are human capacities that we should not simply give away to an algorithm.

The introduction of the AI Inbox in Gmail is a cultural turning point. We can choose to integrate artificial intelligence as a support tool, as a kind of co-pilot that assists us without replacing our judgment, or we can allow it to speak for us. The first path recognizes that technology expands our capabilities, the second runs the risk of turning an entire generation into passive users of other people’s ideas, unable to deal with the complexity of our own. And that is the real question that should concern us.

—

This post was previously published on Enrique Dans’ blog.

***