Echoes & Algorithms is a mini-series delving into the evolving relationship between grassroots journalism and artificial intelligence (AI), exploring how AI influences narrative authenticity and sparking necessary discussions about the future direction of grassroots journalism organizations like Weave News amidst technological advancements.

As this series draws to a close, we must resist the urge to tie it all up neatly. There is nothing neat about the collision between grassroots journalism and artificial intelligence (AI). Instead, we end with a rupture, a break in narrative ease, to ask: What kind of future is being constructed in our name? And who profits from its manufactured inevitability?

The dominant discourse tells us that AI is simply a tool. That it is neutral. That it is here to make our lives more efficient. But this is a rhetorical sleight of hand. Tools are never neutral. And efficiency, in the hands of capital, is never benign. As AI becomes embedded in editorial pipelines, content curation, image generation, and even the cadence of our language, we must examine not just what it does, but what it demands of us. What forms of knowing, labor, and attention does it reward? And what does it erase?

When we dig into these questions, it becomes clear that AI is not just automating tasks; it is automating ideology.

Extraction Disguised as Innovation

The vectoralist class (a term coined by McKenzie Wark to describe those who control not just production but the vectors of information) has quietly become the most powerful editorial board on the planet. They do not write stories, but they structure the feeds. They do not speak, but they decide who gets heard. And in the case of generative AI, they profit from the recombination of cultural labor they did not create.

What we call "training data" in the AI context is often little more than the unpaid archives of writers, artists, and communities whose work is scraped, flattened, and recontextualized under the banner of "innovation." This is not merely a legal or ethical issue. It is also an economic one. We are witnessing the enclosure of the cultural commons: dead labor repackaged as futuristic progress.

“To embrace AI without interrogation is to accept the terms of a system that was never built for us. Optimization, in this paradigm, is not about doing better; it is about doing faster, cheaper, and with less friction. But friction is where meaning lives. Journalism, at its most radical, slows down the story. It introduces context. It complicates. Grassroots media must reclaim the right to be inefficient. To prioritize voice over virality. To tell stories that algorithms cannot predict because they emerge not from data but from relationship. This is not a technophobic stance. We can imagine AI otherwise.”

Under these conditions, grassroots journalism, already operating on the margins of capital, finds itself in a double bind: Use these tools and risk narrative homogenization, or reject them and risk irrelevance in an algorithmically optimized landscape.

But this binary is a trap. The question is not whether to use AI, but under whose terms?

The Crisis of Voice

AI systems are trained on what has already been said, published, or otherwise deemed legible by dominant systems. In doing so, they reinforce the epistemologies of the already-visible. This is not an accident. It is a structural limitation whose implications become clear when we consider the fate of voices shaped by experiences like trauma, diaspora, or refusal. When the machine cannot parse the urgency of such voices, it renders them anomalous: statistical outliers to be smoothed out.

The result is a journalism that sounds increasingly the same. Flattened tone. Predictable cadences. Stories optimized for readability rather than resonance. This is not just about aesthetics; it is about erasing entire ways of knowing and being in the world. The language of resistance is awkward, stuttering, contradictory, and embodied. When it confronts the AI creation of the vectoralist class, it gets lost in translation.

For grassroots journalists, whose authority often stems not from institutional credentialing but from lived experience, this presents a profound threat. The threat is not just to style, but to legibility itself. As Tricia Hersey reminds us, rest, refusal, and slowness are forms of resistance. So, too, is the refusal to be streamlined.

Against Optimization

To embrace AI without interrogation is to accept the terms of a system that was never built for us. Optimization, in this paradigm, is not about doing better; it is about doing faster, cheaper, and with less friction.

But friction is where meaning lives. Journalism, at its most radical, slows down the story. It introduces context. It complicates. Grassroots media must reclaim the right to be inefficient. To prioritize voice over virality. To tell stories that algorithms cannot predict because they emerge not from data but from relationship. This is not a technophobic stance. We can imagine AI otherwise.

Speculative Infrastructure: What If We Built Differently?

What would AI look like if it were built by and for communities? Imagine systems that prioritize care over prediction. Imagine language models trained not on scraped data, but on collectively contributed oral histories, local knowledge, and cultural practices, with consent, attribution, and reciprocity.

Imagine refusal as a feature, not a bug.

There are precedents: the Data Detox Kit, Indigenous AI protocols, feminist server movements. These are not just "alternatives"; they are blueprints for liberation infrastructures. They reframe technology as a site of relationship, not extraction. A terrain of negotiation, not domination.

Grassroots journalism must align itself with these movements. It can do so by creating resources itself, like our colleagues at Project Censored have done with the Algorithmic Literacy for Journalists (alfj.org) initiative. Not just to survive, but to reimagine what journalism can be when it is no longer tethered to the demands of platform capitalism.

Conclusion: What Are We Training?

We often ask what AI will do to journalism. But we forget that journalism is also training AI. Every time a grassroots outlet adopts an AI tool, it sends a signal: this is what matters, this is what truth sounds like. If we allow our work to be shaped by tools built for speed, predictability, and consensus, then that is the world we are reinforcing.

But if we resist - if we tell jagged stories, cite uncitable sources, build local infrastructures - we introduce noise into the system. And maybe, just maybe, that noise is what keeps the future open.

Let us not be merely users of AI. Let us be disturbers of it. Let us teach it to hear voices it was never meant to understand.