A client asked me something recently that I think about a lot. She wanted to know why a smaller, less established competitor kept getting recommended by ChatGPT ahead of her business, despite her company being larger, older, and better resourced.
It’s a genuinely fair question, and the answer tells you almost everything you need to know about how AI recommendation actually works, because it has very little to do with company size and everything to do with trust signals most businesses have never deliberately built.
I run Essheo, a search marketing agency working across the UK and US, and understanding exactly how AI platforms decide who to recommend is central to what we do for clients. I want to walk you through the actual mechanics here, because once you understand what these systems are genuinely evaluating, the path to earning a recommendation becomes considerably clearer.
What Is Actually Happening When an AI Recommends a Brand?
The first thing I always clarify with clients is that AI systems aren’t ranking brands the way Google ranks web pages. They’re doing something closer to reputation synthesis. Rather than matching keywords and counting backlinks, a large language model reconstructs a picture of your brand’s authority and trustworthiness fresh, on every single query, by drawing on two distinct sources of information.
What Are the Two Clocks Every AI Model Runs On?
I think of this as two clocks running at different speeds, and it’s a genuinely useful way to explain it to clients. The first clock is the model’s pretraining data, the enormous body of text it learned from before its knowledge cutoff, which shapes its baseline understanding of who you are and what your brand represents.
The second clock is real-time retrieval, the specific web pages the system pulls into context the moment it answers a live query. A brand’s authority in any given answer is the joint product of both clocks working together.
This matters practically because it means you’re managing two separate things simultaneously. You can’t fix a negative baseline impression instantly through fresh content alone, because that impression is partly baked into the model’s training. But you can influence what gets retrieved live, which is exactly why ongoing content and authority building compounds over time rather than working as a single quick fix.
Does the Model Form an Opinion the Way a Person Would?
Not exactly, but the effect on the reader is remarkably similar. Research into this describes it as pattern synthesis rather than genuine opinion formation. The model weaves together patterns from product reviews, news coverage, social media discussion and technical documentation to form a contextual understanding of your brand, then generates new descriptions carrying implicit sentiment and authority based on those patterns.
If your brand consistently appears alongside language like “reliable,” “industry leading” or “well reviewed” across authoritative sources, the model learns and repeats that association. If it mostly appears in complaint threads or negative reviews, that becomes the baseline too.
I think this is genuinely one of the more powerful, underappreciated aspects of AI recommendation. A positive mention inside an AI response carries an implicit third-party validation, because the AI appears to the user as a neutral, disinterested source simply sharing what it knows, not a business trying to sell anything.
Which Trust Signals Actually Determine Whether You Get Recommended?
Research across this space consistently converges on a similar structure, and I’ve found it genuinely useful to explain to clients as three connected categories of trust.
What Are Website Trust Signals?
These are the fundamentals sitting on your own domain: clear, unambiguous explanations of what your brand does, consistent positioning that doesn’t contradict itself between your homepage and your service pages, and structured data that makes your business legible to a machine rather than just persuasive to a human reader.
If an AI model can confidently explain your brand in a single sentence without hedging, that’s a strong signal your website trust signals are working.
What Are Inbound Trust Signals and Why Do They Matter More Than People Expect?
Inbound trust signals are everything outside your own website that guides potential customers, and models, toward you: media coverage, genuine customer reviews, industry directory listings, Wikipedia and Wikidata entries, analyst citations, and organic social proof of every kind.
Research into this consistently identifies third-party citation density, meaning how many independent, authoritative sources mention your brand in a relevant context, as one of the single most consistent predictors of whether a model recalls and recommends you.
This is precisely why LLM training datasets favour authoritative, frequently cited sources. High-authority content is simply more likely to surface in outputs than content that only exists on a brand’s own site. I’d stress to any business reading this that a brand relying purely on its own website content, however well written, is addressing only one of three trust categories that actually determine AI recommendation.
How Do SEO Trust Signals Fit Into This?
Traditional SEO fundamentals still contribute meaningfully, not because rankings themselves are the goal any more, but because the same signals that earn strong organic rankings, technical crawlability, structured data, genuine topical depth, also make your content easier for an AI system to retrieve, parse and trust in the first place.
This is exactly why I never advise clients to abandon core SEO work in favour of GEO. The disciplines reinforce each other rather than competing for the same budget.
How Do Sentiment and Mention Frequency Actually Shape Recommendations?
This is the layer I think gets least attention from businesses, and it’s arguably the most important one to actively manage.
Does It Matter How Often Your Brand Is Mentioned, or Just Where?
Both, and research consistently ranks mention frequency as one of five core factors deciding which brands get recommended, alongside source authority, review sentiment, query fit and structured data.
Brands mentioned more frequently across the data a model has trained on and retrieved are simply more likely to surface when a relevant query comes in. But frequency without quality doesn’t compound the way businesses often assume. A brand mentioned constantly in low-authority, spammy contexts builds considerably less trust than one mentioned occasionally in a respected industry publication or a genuine comparison article.
I’d also point out that authority in this context is topic-specific rather than domain-wide. A smaller brand that repeatedly and clearly owns a specific niche can outperform a much larger, more generalist competitor, provided its association with that niche is consistent and well documented across the web.
This is genuinely good news for smaller, specialist businesses competing against larger, better-resourced rivals, and it’s a dynamic I’ve seen play out directly in client accounts.
Does the Sentiment Around Your Brand Genuinely Affect What the AI Says?
Yes, and this is where I think most businesses have the biggest blind spot. AI sentiment analysis doesn’t simply count positive versus negative keywords. It evaluates comparative language, qualifying statements, where your brand sits within a recommendation hierarchy, and whether endorsing language is present or conspicuously absent.
Sentiment can be drawn from two distinct places: web grounding, meaning the real-time sources an AI cites for a specific answer, and training data, meaning the general historical associations baked into the model itself.
Practically, this means a cluster of negative Reddit threads or unresolved review complaints doesn’t just sit quietly on that platform. It can genuinely shape how confidently, or cautiously, an AI model describes your brand to a potential customer months or years later . I’d recommend any business genuinely monitor what’s being said about them across Reddit, review platforms and forums specifically, since AI training data carries that sentiment forward in a way traditional SEO monitoring was never built to catch.
How Confident Is a Model in Recommending You If Signals Conflict?
Considerably less confident, and the model’s hedging language will often show it. Research into B2B brand trust signals found that when independent signals align, meaning a brand appears consistently across reputable sources with stable positioning, clear expertise association, and evidence linked to specific outcomes, model confidence rises.
When those signals conflict, or when a brand’s positioning shifts noticeably between different sources, the resulting recommendation weakens, becomes vaguer, or gets dropped in favour of a competitor with a cleaner, more consistent footprint.
Are Consumers Actually Trusting These AI Recommendations?
I think this question matters as much as the mechanics themselves, because it determines how much genuinely rides on getting this right.
How Widespread Is Consumer Trust in AI Product Recommendations?
Considerable, and rising quickly, particularly in the UK. EY’s research found that 58% of UK consumers are now comfortable receiving AI-generated product suggestions, with 38% willing to add a recommended item directly to their basket based on that suggestion alone.
Consumers are increasingly using AI search specifically to compare products against their own budget and requirements, assess brand reputation, and evaluate alternatives before ever visiting a website directly.
EY’s research also describes a genuine shift in how AI systems retrieve information, moving from deterministic keyword matching toward what it calls semantic retrieval, where content gets selected based on how probable it is to actually answer the query well, rather than how closely its keywords match.
This reinforces exactly the point I’ve been making throughout: it’s genuine trust, clarity and consistency that earn a recommendation now, not keyword density or paid placement alone.
What Happens When AI Gets Your Brand Wrong?
This is a genuine risk businesses need to take seriously. AI systems can misinterpret brand-owned content, user-generated commentary, or simply outdated information, leading to inaccurate or misleading representations of your brand that you may never even see unless you’re actively monitoring for them.
As AI increasingly shapes consumer perception at scale through these recommendations and comparisons, maintaining some control over your brand narrative becomes a genuinely commercial concern, not just a reputational one.
This is exactly why proactive monitoring of how AI platforms describe your brand needs to become a standing part of any serious marketing operation, rather than an occasional check. It’s precisely the kind of ongoing tracking we build into every client engagement at Essheo, because a recommendation you’re earning today can quietly shift if negative sentiment accumulates somewhere you’re not watching.
How Should Your Business Actually Build These Trust Signals?
Bringing all of this together, I’d focus deliberately on strengthening each of the three trust categories rather than treating any single tactic as a complete fix.
On your own site, make sure your positioning is genuinely consistent and unambiguous, and that structured data clearly communicates who you help, what problem you solve, and how you’re different, exactly the kind of clarity a model needs to confidently summarise you in one sentence.
Beyond your site, invest deliberately in earning coverage, reviews and mentions across genuinely authoritative, independent sources, since this inbound trust category is consistently the strongest predictor of recommendation across the research I’ve reviewed.
And monitor sentiment actively across review platforms, forums and social discussion, because that sentiment feeds directly into how confidently AI systems describe you, whether you’re watching or not.
This is precisely the search everywhere approach we take with every client at Essheo. We don’t treat your website, your third-party coverage and your platform presence as separate projects. We build them as one connected trust signal, because that’s genuinely how these AI systems evaluate you.
Ready to Find Out What AI Is Actually Saying About Your Brand?
If a competitor keeps appearing in AI recommendations ahead of your business despite being smaller or less established, it’s very rarely bad luck. It’s almost always a gap in one of the trust signals covered here, and those gaps are entirely fixable once you know where they sit.
At Essheo, every practitioner on our team carries 8 plus years of experience building genuine authority and trust in hard, competitive sectors against household name rivals, and we work across Google, LLMs, YouTube and social platforms because recommendation trust today gets built across all of them at once, not on a single channel.
Our clients have generated over £45 million in combined revenue over the last two years through the systems we’ve designed and implemented, built on real operational results rather than theory.
We’ll start with an honest audit of exactly how AI platforms currently describe your brand against your competitors, then build a tailored roadmap to close the trust gaps that matter most.
Book a strategy call with me, and let’s find out exactly what ChatGPT, Gemini and Perplexity are telling your customers about you right now.