One of the most helpful mental models I’ve picked up is simple: we’re moving from a world of links to a world of answers. And in that world, the win condition changes.
When your audience asks a question in an LLM interface, they aren’t met with ten blue links. They’re met with a single synthesis, often confident, often tidy, sometimes wrong, but always influential. The new challenge isn’t “can people find you?” It’s “does the machine include you?”
This is why I keep coming back to the phrase: a citation economy. If answers are the currency, citations are how that currency is backed. Your brand’s authority is increasingly expressed through what gets cited, repeated, and reinforced across the sources these models rely on.
And if you take nothing else from this piece, take this: LLMs reward clarity and usefulness, not polish and boilerplate. When content is written in the traditional PR “announcement” style – heavy on adjectives, light on concrete facts and figures – it can be strangely invisible to systems built to extract answers.
The “Human Algorithm”: three things communicators now need to do
I describe this as a “Human Algorithm”, a formula for optimising human connection in an LLM world. It expands our craft to include tech, while keeping the human outcome at the centre
- If audiences are talking to machines, we have to learn to talk to machines
This is where concepts like Generative Engine Optimisation (GEO) come in. Not as a gimmick, but as a recognition that content now needs to be legible to systems that synthesise and cite.
A lot of what we used to treat as “nice-to-have” content structure becomes strategic. Formatting key information as Q&As, fact sheets and listicles; making your claims explicit; and ensuring your owned platforms are machine readable, not just beautifully designed.
And crucially, earned media shifts too. LLMs often prioritise relevance over sheer size, which means subject-specific domains, specialist trades, journals and niche credibility can matter more than a single “big” hit.
- We still have to protect human-to-human trust
The most underrated risk in this space isn’t that machines exist. It’s that people may stop interrogating sources because the answer sounds plausible.
LLMs can cite accurately and they can also hallucinate, flatten nuance, or lift an outdated framing into the present. That creates reputational risk and an opportunity for communicators to do what we’ve always done best: build credible narratives anchored in truth.
For me, this is where communications leaders need a new kind of literacy: understanding how authority signals are formed and repeated, while holding the line on accuracy, ethics and context.
- We need to create authentic experiences worth choosing
If AI compresses information into summaries, then the brands that win won’t just be the ones that are mentioned, they’ll be the ones people still feel something about.
The more mediated discovery becomes, the more valuable genuinely human experiences are. We’re not optimising for the machine because we want to please machines. We’re learning the machine layer so that we can better earn trust on the human layer.
What I’m learning (in practical terms)
First: treat each model like a different editorial lens. Outputs differ. Citations differ. The same question can produce very different framing across tools, which is why testing across more than one has become a discipline, not a curiosity.
Second: build content that answers real questions. Not what we want to announce, but what people actually ask when they’re trying to decide. Specialist earned coverage, community platforms, and strong owned content all matter because they influence where answers are sourced.
Third: invest in algorithmic capital and understand the nuances of language. Not only does each model differ, but when you run analysis in Arabic and English, for example, the results will differ. This means reframing content as an asset: content that is findable, understandable, and citeable by machines. The point isn’t to game anything. The point is to make sure the truth about you is structured well enough to travel.
And yes, community platforms matter. In an AI-driven world, credibility is shaped by shared, experience-led answers and what’s repeated, debated and agreed upon at scale. When so much of generative output is shaped by what’s considered helpful and conversational, the places where people trade real answers end up disproportionately influential.
What this means for communicators (and for leadership)
I’ve led communications teams through plenty of shifts. This one is different because it touches the mechanism of discovery itself.
Leaders don’t all need to become technical specialists, but they do need enough fluency to ask better questions:
- Where is the machine getting its information?
- Which sources does it trust?
- Where are we absent, outdated, or misrepresented?
- If our reputation is being summarised, would we like the summary? Personally, learning this space has meant slowing down, asking basic questions, getting things wrong, and pressure-testing what’s useful versus what just sounds impressive. It’s been humbling in the best way, because it brings you back to first principles: clarity, evidence, credibility and empathy. And that’s why I’m optimistic.
Because if there’s one thing communications people know how to do, it’s build trust in complex environments. The tools are changing, but the job remains beautifully familiar: help people understand what’s true, what matters, and what to do next.
By Sophie Simpson, Managing Director of Ruder Finn Atteline – MENA.
