Your Brand Is Invisible to AI Search

By Linda Cereda, former Global VP Marketing Data, Nike

Eighty percent of Google searches in 2026 now end without a single click. The query is asked, an AI engine answers, and the user moves on. They never visit your site. They may never know your brand was an option.

That number has jumped from 60% in 2024. The acceleration is not slowing down. About 58% of consumers have already replaced traditional search with AI tools for product recommendations, up from 25% in 2023. AI-powered shopping journeys convert at roughly four times the rate of traditional ones, but only for brands the AI chooses to cite.

For most retail and consumer brands, this is a discovery problem they have not yet diagnosed. SEO is still funded. Paid search is still measured. But a growing share of the traffic that should be flowing through those channels is being intercepted upstream - at the answer layer - and most marketing organisations have no visibility into how their brand performs there.

This is not an argument for stopping SEO. SEO remains the foundational layer that AI models rely on to retrieve and evaluate sources. Strong domain authority, quality backlinks, and well-structured content still matter; in fact, they matter more than ever, because they are the prerequisite for even being considered. The point is that winning in SEO no longer guarantees winning in AI search. A brand can be the top organic result and still not appear in the answer. The two are not the same game.

The discipline that addresses this gap is called Generative Engine Optimisation, or GEO. The brands that build this capability in the next 12 months will own AI-mediated discovery for the rest of the decade. The ones that treat it as someone else's problem will be invisible inside the answer, regardless of how well they rank in the blue links.

The mechanism is different

A traditional search engine indexes pages, ranks them by relevance and authority, and presents a list. The user picks the result that looks most useful. The brand wins or loses on the click.

A generative engine does not present a list. It synthesises an answer from sources it considers most credible, most clearly written, and most directly relevant to the question. The user reads the answer and moves on. The brand wins or loses inside the synthesis, before the click ever happens.

Two consequences follow. First, ranking on a results page no longer guarantees inclusion in the answer. Second, the signals that drive citation are different from the signals that drive ranking. SEO foundations (domain trust, authoritative backlinks, clean site architecture) are still the entry ticket. But the signals that determine whether a brand gets cited inside the answer go beyond them. Understanding those signals is where most brands have a gap.

What drives citation: the four pillars

When a user asks a generative engine a brand-relevant question - "What are the best running shoes for flat feet?" or "Which loyalty programme has the best rewards for frequent travellers?" - the engine retrieves candidate sources, evaluates them, and synthesises. The brands that get cited consistently score well across four dimensions.

Relevance is about being the exact answer to the right question at the right moment. That means mapping the questions your category owns, not just the branded ones, but the generic "best of" and comparison queries your customers actually type. Each page should lead with a direct, 30-60 word answer to the question it targets. FAQ formats, comparison tables, and product specifications all help. LLMs are looking for content they can extract cleanly and reuse inside an answer, not brand voice writing designed to seduce a browsing human.

Trust is where SEO foundations and new signals overlap. Strong E-E-A-T signals, reputable backlinks, and cited sources remain important. But AI engines also weigh a broader set of proof signals. Original research adds significant credibility - e.g. a product efficacy study run with a university partner, a clinical trial, proprietary consumer data published with methodology - because it gives models a primary source to cite rather than relying on secondary coverage. Expert quotes in industry publications, consumer reviews, Reddit threads where your category is discussed, and peer recommendation sites all contribute to the same effect: your claims appearing across multiple independent, trusted contexts. Content freshness matters too; pages refreshed regularly with current data are weighted more heavily than those that have not been touched in two years.

Readability is about making content easy for a bot to scan, parse, and reuse, which is different from making it beautiful for a human to browse. Descriptive headings, short paragraphs, structured tables, and FAQ schema all improve citation rates. Gated PDFs, image-heavy pages, and long walls of unbroken text are largely invisible to models. A category page with five paragraphs of brand voice and a product grid is excellent for a returning customer. It is nearly useless to a model trying to answer a comparative question.

Identity is the most underestimated pillar. If an AI engine is uncertain about who you are, what category you play in, or what your brand actually stands for, it will default to a competitor it is more confident about. A consistent brand presence across the web, a strong Knowledge Panel entry, Wikipedia presence, robust author pages linking experts to their content, consistent naming and positioning across directories, gives models the clarity they need. In regulated or high-trust categories, this is where digital passports and structured entity data become critical.

Third-party content drives citation more than owned content

Most brand GEO programmes start with owned content: rewriting website pages, adding FAQ sections, restructuring product descriptions. That is the right place to start. It is not where citation authority is built.

Third-party sources drive approximately 90% of AI citations in unbranded queries - e.g. the "best running shoes for beginners" type of search - and around 60% even in branded queries, where the user is explicitly asking about your brand. Models are trained to weight sources they encounter across many different contexts. A brand that appears only on its own domain, even with excellent content, does not carry the same citation weight as one that is referenced across industry publications, expert roundups, review platforms, and community discussions.

The practical implication is a shift in where effort goes. Earned media in trusted industry publications, expert contributions to relevant editorial, presence in Reddit and Quora threads where your category is actively discussed, structured reviews on category-specific platforms; these are not PR activities separate from GEO. They are GEO. Brands that separate owned optimisation from third-party seeding will see partial results. The ones building both in parallel are the ones moving citation share.

Where to start: build your knowledge forest

The first step is not creative. It is strategic.

Map the territory your brand needs to own. Start with your trunks - the two or three category topics that are essential to your business (e.g. Running Shoes). Under each trunk, define the branches, the specific questions consumers are asking and that AI engines are synthesising answers for (best shoes for beginners with knee injuries, how to choose running shoes for overpronation, running shoe comparison: stability vs neutral). Under each branch, build the leaves, the content assets that answer those questions directly: product pages, blog posts, FAQs, expert contributions, earned media, reviews.

Then run the diagnostic before you create anything. Leveraging a GEO tool, put your top 3-5 brand-relevant queries through ChatGPT, Perplexity, Gemini. Document what each engine says, which sources it cites, and where your brand appears or does not. That audit tells you exactly what kind of gap you have. If your brand appears with weak framing, you have a content and relevance problem. If your brand does not appear at all, you have a corpus visibility problem, and the fix starts with third-party presence, not page rewrites.

In one recent programme for a global company in a regulated category, this audit identified a priority topic where the brand was completely absent from AI-generated answers. Six months of structured GEO work - diagnostic audit, priority question mapping, content rebuild, and cross-functional operating model - moved the brand to the top position on AI search in the focus market, measured on share of citation across ChatGPT and Gemini. The working model was fully adopted post-handover and has continued to compound.

This is not an SEO project

The final point is the most important one for how this gets set up internally.

GEO is not an SEO team initiative. It is a cross-functional programme. The SEO team handles the technical foundations - crawlability, schema, site architecture - without which nothing else works. Brand decides which category stories matter and what the non-negotiables are on positioning and tone. Content creates the answer-first assets. PR and comms secure the third-party presence that drives citation authority. Analytics tracks citation share and AI share of voice over time. Legal reviews claims in high-risk categories.

When one team owns this alone, the brand gets partial results. The brands generating real AI citation share have broken the silo and run it as a company-wide capability - the same way they would treat any meaningful shift in how customers find them.

SEO moved sites up the page. GEO moves brands inside the answer. The transition is underway. The question is whether your brand shows up in it.

Linda Cereda leads The Convened's AI Visibility & GEO offering within Capability Infrastructure. To discuss a GEO assessment or transformation engagement, contact us.