LLMs, GEO, and the New Rules of Influence

A Princeton, Georgia Tech, and IIT Delhi study on generative engine optimisation found that adding citations, statistics, and quotations to content could lift its visibility in AI-generated answers by up to 40 percent. This means that getting found by an LLM is not the same game as getting found by Google, and brands are now in the process of adapting their marketing strategies to reflect that.

Why third-party validation matters more, not less

The biggest shift for brands is not technical, it's reputational. LLMs place a heavier weighting on what other people say about a brand than on what the brand says about itself. That changes the economics of PR and comms — a discipline that had been quietly waning in recent years — back into a genuinely critical function.

What that means practically is straightforward: it doesn't matter how much a brand says about itself on its own website; claims need external verification. Saying "we're experts in X" only works if a piece of independent industry press exists to back it up. The same logic runs through consumer products, where recommendations, referrals and reviews now function as legitimacy signals, not just reputation management. LLMs are, in effect, forcing brands to prove what they've always merely claimed. Good for their customers.

The search journey was already changing

LLM-mediated discovery is a continuation of a trend that was already well underway. Organic search results have been pushed further and further down the page for years, and targeting has become so precise that the old behaviour of scrolling through page after page of results, comparing multiple sites, had mostly disappeared before LLMs entered the picture. What has already been working — hyper-targeting and social proof, particularly the influence of people others aspire to be — has been doing more and more of the persuasive work in recent years.

What does that mean for where brands put their effort? Inside an LLM conversation, the experience today functions like a faster, more capable search assistant: find this, find three versions of it, compare the prices. It is informed discovery with less friction, not a new form of persuasion. Arguably the most impactful source of persuasion, at least for now, is still social platforms and the people on them, including the growing presence of AI-generated influencers. Whether people ever follow an LLM's recommendation in the way they follow an influencer's — seeing it as aspirational — is an open question.

Measuring visibility is still catching up

Until recently, understanding a brand's visibility inside LLM answers was largely manual and haphazard: guessing at prompts a customer might type and checking, one by one, whether the brand showed up. That's changing. A new category of GEO measurement tools — including Profound, new Semrush functionality, and AthenaHQ — now track brand mentions, citations, sentiment, and share of voice across ChatGPT, Gemini, and Perplexity, functioning as something close to the SEO rank-tracking of the previous era. The Princeton research specifically credited citations, statistics, and credible quotes as the strongest levers — consistent with what brands are already learning the hard way: LLMs reward content that can be checked against outside sources, not content that simply asserts.

Not every brand should chase AI influence the same way

Whether a brand should adopt AI influencers, or any AI-mediated persuasion at all, comes down to a familiar old question in a new context: what does the brand actually stand for? Brands built around efficiency, innovation, and a digital, avant-garde identity have a natural affinity with AI influencers and the audiences who already live in that space. Gaming is an obvious example. Brands built on craftsmanship and heritage, while often employing a large army of human social influencers, do not feel like such a fit. A luxury brand selling the provenance of hand-selected leather, tanned and stitched by people with a point of view, is selling something explicitly human — and pushing that brand toward AI-generated influence works against the thing customers are actually paying for.

The gatekeeper question is overblown, for now

There's a reasonable worry that a handful of AI model makers could become the new gatekeepers of commerce, deciding what gets recommended and what doesn't. That concern is currently overblown. Google held that position for a long time already, then Meta built a comparable one through Facebook and Instagram advertising. Multiple LLMs exist today, and consumers can choose between them. The real question is whether the market consolidates. Gemini's position inside Google is worth watching in that regard, but that concentration hasn't happened yet.

What does good advertising inside an LLM look like?

Paid advertising inside LLM conversations is still at an early stage. Before considering whether ads inside LLM conversations are intrusive or not, it's worth being precise about what "intrusive" actually means. Intrusiveness is not really about whether an advert is present. A billboard on a motorway is obviously an advert, yet isn't inherently intrusive — unless the nature of the content itself is. The early days of search advertising went through exactly this maturing process, with rules emerging about what counted as reasonable, non-manipulative content.

What makes advertising genuinely annoying is irrelevance and poor tonal judgement: an advert that has nothing to do with what someone is currently doing or needing, or one that ignores basic context — a gambling advert appearing next to a search for children's products, or humour that tips into being offensive rather than funny. The best advertising, in an LLM context or anywhere else, understands the person and serves them something that fits who they are in that moment. AI's ability to actually personalise that, at the level of the individual rather than a broad segment, is one of the more interesting opportunities in the category, if it's built with restraint.

The upside for consumers

The conclusion across all of this is that LLMs are quietly raising the standard for brand honesty. Because models weight independent validation so heavily, brands can no longer claim things about themselves that nobody else is willing to confirm. That is, on balance, a good outcome for consumers — and it's worth agencies and brands treating it that way rather than as a threat to be managed.