Interview with Mathieu Trepanier, managing director at PharosGraph, a narrative engine optimization platform from the USA.
Nice to meet you, Mathieu. Could you share a bit about your background and your current role and responsibilities at PharosGraph?
I am the founder and CEO of PharosGraph. My background is at the intersection of economics, politics, and society. I’m interested in how we can leverage advanced computational linguistics and next-gen AI to understand how consumers relate to brands, voters to candidates and societal issues, and stakeholders more generally to reputation challenges. I’ve founded several companies in this space.
I hold a PhD in managerial economics and strategy and studied applied economics and business administration. Over my career, I’ve had the pleasure of working with renowned global clients such as Nestlé, Procter & Gamble, The Boston Consulting Group, Novartis, and Pernod Ricard, among others.
How does PharosGraph position itself differently in the market? What is your competitive advantage?
PharosGraph is designed as an AI-native Narrative Engine Optimization platform – referred to as NEO among peers and friends. We map emotional and semantic patterns while detecting blame attribution in global discourse, that is, real conversations people are having all over the world and not just the keywords or mentions they use.
We measure “what” is being said, but mostly “how” and “why” it’s gaining traction across media and public narrative ecosystems. This enables clients to not only see how their brand, policy, or campaign registers in AI models but to understand the causal influence and latent shifts in public perception.
Our edge lies in blending advanced computational linguistics, behavioral sciences, and advanced AI to generate high-resolution narrative tracking.
PharosGraph is using next-gen AI – a field that’s rapidly evolving. How do you see it developing over the next 5–10 years, and how will that reshape the needs of companies?
LLMs (Large Language Models) in AI have already become central to how people explore ideas, compare options, and form opinions. If your brand, candidate, or cause is not positively represented in these outputs, you risk becoming irrelevant in the market.
In the coming decade, AI will shift even more from being an analytic tool to an interpretive collaborator. Models will evolve further from detecting perception to decoding social meaning, intent, and long-range reputational impact. For our space – media, brand, campaign, and cause narrative intelligence – this means clients will expect systems that can forecast perception cascades and adapt messaging in near real time. PharosGraph is architected to meet that trajectory, with infrastructure for emergent behavioral modeling already built in.
What are the most common misconceptions that clients have when it comes to AI within media analytics?
Many assume AI can deliver “truth” or final judgment on perception, when in fact it reflects patterns of perception that are often culturally and linguistically biased. Others think of AI as a black box that automates insight. The true value is in helping people understand more by revealing patterns, not by taking over their thinking. There is also an overreliance on sentiment polarity. If you label things as merely positive, negative, or neutral, you miss the deeper, nuanced emotions and the true message of a story – also known as narrative function (what role something plays in a story – like setting tension, revealing conflict, creating empathy, etc.). PharosGraph’s NEO is designed to pick up on those subtle layers that basic sentiment analysis overlooks.
Can you share concrete examples of how global sentiment insights, like blame, praise, or empathy, have influenced consumer behavior in ways relevant to your clients’ businesses?
In a highly polarized electoral environment, candidates in Colorado used NEO Battleground to navigate localized shifts in perception and the risks in narrative framing. Real-time insights revealed whether candidates were being cast as reformers or obstructionists. With trust maps, we were able to show that suburban voters were starting to lose confidence in climate-related messages, while a Low Clarity Score among rural youth flagged the need for simpler healthcare language.
Campaign teams could leverage NEO’s live dashboards and automated playbooks to adapt messaging, strengthen weak zones, and track how messages are being framed and shifting over time across different digital platforms.
With our other system, NEO Brand, we’ve been exploring the explosive growth and public scrutiny around GLP-1 drugs, used for example for type 2 diabetes. Specifically, how narratives around weight-loss drugs were evolving.
NEO Brand found that people were split: some saw it as a luxury shortcut to medical necessity, others as something medically important. It also spotted cultural tensions about fairness, especially among Gen Z. People explained their views in two main ways: either as a big scientific win or as an unfair cheat.
Having surfaced these findings, we used NEO’s recommendation engine and reputation heatmaps to recalibrate messaging and align local physicians, which allows brand teams to manage influencer discourse in real time.
In your experience, what are the most common blind spots or assumptions companies make when trying to interpret emotion or intent through data analytics?
Companies often assume emotional intent is uniformly expressed across languages and cultures – which isn’t true. Or they treat engagement metrics, such as likes and shares, as proxies for meaning, missing deeper narrative cues like irony, skepticism, or passive dissent. NEO is built specifically to surface those deeper cues.
Another blind spot is underestimating how fast narrative context shifts. Data may be current, but feelings can invert within days if not tracked at narrative scale.
PharosGraph’s NEO is built to understand how a brand or candidate is framed in AI. Being visible is one thing, but being framed as a hero, villain, victim, or neutral, is power. It determines how audiences trust, support, or reject your message.
Which features or capabilities of your product do you believe have untapped potential, but haven't yet seen widespread adoption among your clients?
Our dynamic narrative maps, which visualize emerging clusters of perception and semantic influence, are powerful but underutilized. So is our emotion-by-topic crosswalk – a feature in PharosGraph that maps specific emotions, such as blame, admiration, empathy, fear, or shame, to topics or entities over time and across media contexts. This lets clients track how feelings like shame, hope, or admiration attach to specific concepts over time.
These tools offer early-warning signals, but many clients are still adapting their workflows to integrate predictive narrative intelligence, not just retrospective reporting. And they are critical today because context is everything: Audience dynamics are increasingly fragmented by geography, issue salience, and cultural perspective. An effective business strategy is one that matches narrative insights with an understanding of the local environment.
What role do you see AI playing in how we understand and influence narratives in the future?
We believe the future of AI in media and narrative intelligence is about meaning. As narratives become the battleground of trust and reputation, platforms like PharosGraph NEO can help brands, institutions, and even nations make sense of the emotional currents shaping global discourse.
You cannot simply automate judgment. We give our clients the tools to see what’s beneath the surface today for better outcomes tomorrow. NEO metrics are not passive indicators. They are built to trigger playbooks, align teams, and direct action. The goal is not to observe the narrative, but to shape it.
By Anna Roos van Wijngaarden
Twingly offers News APIs with over 3 million daily news articles from 170,000 active global sources. It gives you access to noise-free news data for timely updates and comparisons. As a complement, please also have a look at our Blogs API, where we monitor 3 million active blogs worldwide.
