How AI decides who to trust — and who to recommend.
For over a year I have run independent, observational research into how large language models and answer engines form opinions about organisations: how they resolve entities, infer trust, and decide which sources to surface. This page collects that work, its methods and what it means for businesses navigating AI-led discovery.
Four questions at the centre of the work.
Each engagement and study returns to the same underlying problem: what happens between a user's question and an AI system's answer, and how does a business end up included, excluded or misrepresented along the way.
What makes a model trust a source?
Observing which signals appear to drive whether a source is surfaced, hedged or ignored — corroboration, consistency and credibility over any single ranking factor.
How do models know who you are?
How systems disambiguate an organisation from similarly named entities, and what causes them to merge, confuse or mislabel a business.
Where do answers actually come from?
Mapping the sources retrieval-augmented systems reach for, and how the shape of your digital presence changes whether you are found.
Why does AI recommend one brand over another?
Tracing how a model moves from "aware of you" to "actively recommends you" — and the gaps that quietly keep businesses out of the conversation.
Featured paper · ResearchGate & SSRN
How Do Generative AI Systems Infer Source Trust in Informational Responses?
An observational study of trust signalling in AI-led information discovery. Using structured prompt testing across leading platforms, it examines what models appear to rely on when deciding which sources to trust, cite and surface — and where those judgements can be shaped or corrected.
- Structured, repeatable prompt testing across multiple AI platforms.
- Focus on corroboration, consistency and entity clarity as trust signals.
- Practical implications for how organisations present themselves online.
Observed behaviour, not assumptions.
The research is grounded in what AI systems actually do when asked real questions — captured through a consistent, transparent process rather than guesswork about how the models work internally.
Structured prompting
Repeatable prompt sets ask models the questions real customers ask, in controlled, comparable conditions.
Cross-model comparison
Responses are compared across platforms to separate model-specific quirks from consistent patterns.
Signal analysis
Patterns are traced back to the underlying signals — clarity, corroboration, freshness and entity resolution.
From research to a practical guide.
The research underpins my book — a concise guide to how generative systems construct brand narratives, infer trust and decide which organisations to surface, and what businesses can actually do about it.
It translates the same observed behaviour into language and steps that business leaders, marketers and founders can act on without a technical background.
GEO, AEO and AI visibility, explained.
What is Generative Engine Optimisation (GEO)?
GEO is the practice of improving how clearly and accurately generative AI systems — such as ChatGPT, Gemini and Perplexity — understand, represent and recommend an organisation when answering user questions. Unlike traditional SEO, the goal is not a blue link on a results page but an accurate, favourable mention inside the AI's answer itself.
How is AEO different from GEO?
Answer Engine Optimisation (AEO) focuses on being the source a system draws on to directly answer a question. GEO is broader — it covers how you are described, compared and recommended across generative experiences. In practice the two overlap and reinforce each other.
How do AI systems decide which sources to trust?
Observational research suggests models rely on a combination of signals — consistency across sources, corroboration between them, the clarity of your entity, and the credibility of where information appears — rather than any single ranking factor. When those signals agree, a model is more confident; when they conflict, it hedges or omits.
Why does AI visibility matter for businesses now?
AI systems increasingly form a first impression of an organisation before a customer reaches its website. If that impression is inaccurate, outdated or missing, it shapes decisions you never see. Understanding how it forms is the first step to influencing it.
Want this applied to your organisation?
The same research powers every AwarenessAI engagement — turning observed AI behaviour into a clear plan for how models represent you.
Founder of AwarenessAI, author of How Does AI Talk About Your Brand? and an independent researcher in AI visibility and Generative Engine Optimisation (GEO).