A media planner I work with ran the same prompt into ChatGPT, Perplexity and Google AI Overviews last week. Her client, a mid-tier financial services brand, appeared in zero of the three answers. Two competitors, both with smaller media budgets, appeared in all three.
That is the shift. Discovery has moved upstream of the click.
For two decades, the search results page was the front door. Brands paid for position, optimised for keywords, and measured success in blue links. The page still exists, but a growing share of intent never reaches it. The user asks a model. The model answers. The brands cited in that answer win the consideration set before the user has typed a second query.
Agencies have noticed. Generative Engine Optimisation moved from conference panel to budgeted retainer this quarter. Profound, Goodie and Peec AI are selling share-of-voice tracking inside LLM outputs. Major holding companies have spun up GEO practices with their own audit frameworks. The work itself is unglamorous. Schema markup. Entity disambiguation. Source authority. Citation hygiene across the corpus the models actually crawl.
It is not SEO with a fresh coat of paint. SEO optimises for a ranking algorithm that returns a list. GEO optimises for a generative system that returns a synthesis. The unit of value is no longer the click-through, it is the citation. And citations accrue to brands whose underlying entity data is clean, consistent, and machine-legible across the open web.
This is where most brands are exposed. Their product pages were written for humans and lightly tagged for Google. Their PR coverage lives behind paywalls the models cannot ingest. Their Wikipedia entries are thin or contested. Their knowledge graph entries on the major platforms are stale. When a model tries to answer "who are the leading providers of X in market Y", it reaches for the cleanest available structured data. If your competitors have invested in that scaffolding and you have not, you are invisible. Not down-ranked. Invisible.
The publishers we partner with figured this out faster than most brands. Newsrooms that adopted structured content disciplines years ago, with proper entity tagging on every person, place, organisation and product, are now over-represented in AI answers. The Brief is a case in point. The investment in clean taxonomy was made for editorial reasons. It pays off in a discovery layer nobody had designed for.
The implication for brands is uncomfortable. GEO is not a campaign deliverable, it is an operating discipline. It sits across content, PR, product marketing, and developer relations. It requires someone to own the entity graph the way a CFO owns the chart of accounts. Most marketing organisations have no such role. The agencies that win the next cycle will be the ones that fill it.
Inside Broadbrand, GEO sits in the Create pillar alongside rich media, social and SEO. We treat it as a structured-content practice with measurable outputs. Citation share inside named models. Entity coverage across the open knowledge layer. Source authority scores on the domains that feed the crawlers. The work is closer to information architecture than to copywriting, and the talent profile reflects that. Editors who understand schema. Strategists who can read a knowledge graph. Producers who think in entities, not assets.
The measurement question is the one clients ask first. The honest answer is that GEO measurement is where SEO measurement was in 2008. The tooling is improving weekly. Share-of-voice inside LLM outputs is now trackable with reasonable confidence across the major engines. Attribution from AI citation to downstream conversion is harder, but the proxy metrics are stabilising. Brands that wait for perfect measurement will be late. The competitors moving now are building citation share that compounds, because models trained on a corpus where you are already cited will keep citing you.
The deeper point is about who feeds the models. The open web is being rewritten as training data and retrieval input for systems most users will interact with daily. Brands and publishers that treat their content as structured, authoritative, machine-legible source material will be foundational to that layer. Those that treat content as disposable campaign output will be excluded from it. The asymmetry between the two postures will widen quickly.
The front door has moved. The brands walking through it first are the ones who built the entity graph before they needed to.