Something changed in how people find information, and most marketing departments are still measuring it with the wrong ruler. A growing share of searches now end without a click at all. The user asks a question, an AI system answers it directly, and the brands that used to compete for the top ten blue links are increasingly competing for something narrower: a mention inside a single generated paragraph.
That shift has a name, Generative Engine Optimization, or GEO, and a growing body of evidence behind it. OpenAI said in 2025 that ChatGPT users were sending more than 2.5 billion messages to the platform each day. Bain & Company’s research found that about 80% of consumers rely on zero-click results for at least 40% of their searches, while about 60% of searches end without the user progressing to another destination. Those figures help explain why marketing teams that spent a decade mastering search rankings are suddenly being asked a different question: does the AI even know we exist, and if it does, does it recommend us or somebody else?
Search visibility used to mean one thing. Now it means two.
Ranking on page one and being named inside an AI answer are no longer the same job, and treating them as one is the mistake many brands are making. A traditional search result gives a user several options and lets them choose. An AI generated response, by contrast, often hands the user a synthesized answer, with a smaller number of brands or sources appearing directly inside that response.
The stakes of that compression are real. Bain’s research found that around 80% of consumers rely on zero-click results for at least 40% of their searches, while roughly 60% of searches now end without the user moving on to another destination. Adobe’s data adds another layer: visitors arriving at websites from AI channels convert at 31% higher rates than visitors from other channels. Read those findings together and the picture is uncomfortable for anyone who still treats organic traffic as the only scoreboard. Fewer searches are necessarily ending in a click, but the people who do arrive through AI referrals can be further along in their decision making.
GEO doesn’t replace SEO. It adds a second exam.
Generative Engine Optimization is not simply a rebrand of SEO, and it isn’t a replacement for it either. It is better understood as a parallel discipline with a different end goal. Where classic SEO focuses heavily on helping a page rank for a query, GEO is concerned with making information clear, authoritative, useful and easy for AI systems to understand and use when constructing an answer.
Adobe’s recent work on GEO makes the distinction clear. Brands need to monitor how they appear across AI surfaces, improve the content behind those appearances, and treat optimization as an ongoing practice rather than a one off campaign.
That doesn’t mean traditional search is disappearing. Search advertising, rankings, technical SEO and strong websites still matter. The difference is that search is being joined by another battlefield where the currency isn’t only clicks. It is also mentions, citations, recommendations and whether a brand becomes one of the sources an AI system chooses to use.
The four metrics that actually matter now
Ask most marketing teams how their brand performs in ChatGPT, Gemini, Perplexity, or Copilot, and you may still get a shrug. Adobe’s research found that 80% of companies have significant gaps in how their content surfaces in LLM results, and many don’t know what AI systems are actually saying about them. That’s a measurement problem, but it is one that can be addressed once teams stop trying to force AI visibility into an old SEO dashboard.
Four metrics are becoming particularly useful.
Citation rate tracks how often a brand’s content is directly referenced in an AI answer. It provides a practical signal of how frequently that content is being selected as supporting material.
Brand mentions capture the broader version of that signal: a company can be named even when an AI response doesn’t provide a direct link. That still matters because the language used to describe a brand can influence how it is associated with a product category or topic.
Share of voice looks at how often a brand appears compared with competitors when people ask category level questions. This becomes especially interesting in an AI environment because a generated answer may surface only a handful of names instead of a full page of results.
And then there is AI referral traffic, which tracks visits that actually arrive from AI driven platforms. Adobe describes these metrics as a way to understand not only whether a brand is visible, but how it is being interpreted and recommended.
None of this is purely theoretical. Adobe says that when it applied GEO practices to Adobe.com, it recorded a fivefold increase in citations for Firefly, a 200% increase in LLM visibility for Acrobat, and a 41% increase in LLM referral traffic within weeks. Adobe also reports that General Motors saw AI visibility rise by 23% and citations increase by 35% after developing LLM friendly content pages.
Why low effort content backfires here
There’s a temptation, understandable given how cheap AI writing tools have made content production, to simply publish more. That instinct is unlikely to solve a GEO problem on its own.
The more useful question is whether the content gives an AI system something worth using. Generic articles repeating the same points as hundreds of other websites do not offer much distinctive information. Content built around original research, first party data, expert commentary, useful examples and clearly sourced information has a much stronger case for being referenced.
Adobe makes a similar argument in its GEO research: content needs to be differentiated, authoritative and useful rather than simply abundant.
The inverse is also true, and it’s where original research, first party data and genuine expert commentary earn their keep. A model synthesizing an answer about commercial security systems, financial software or supply chain technology has more reason to draw from a source that offers a specific data point, a documented case study or expert perspective than from a page repeating conventional wisdom.
That is an old publishing principle meeting a new distribution system. Strong reporting has always needed something distinctive to say. The difference is that an AI system may now be the first reader deciding whether that information is worth passing on.
What brands should actually be doing this quarter
Practically, the work breaks into a few concrete steps rather than a vague mandate to “embrace AI search.”
Start with a real audit. Put 20 to 50 realistic buyer questions into ChatGPT, Perplexity, Gemini and other relevant AI platforms. Record whether the brand appears, how it is described, which competitors are mentioned and which sources are being cited.
Brands can do this work internally, although specialist AI visibility services now also offer ways to benchmark mentions, citations and competitor visibility across AI surfaces. AI visibility auditing and citation services can be useful when teams need a more systematic view of where a brand is appearing and where it is missing.
From there, the priorities are fairly practical. Write in a way that answers a question early rather than making the reader dig for the point. Keep important informational pages current. Build pages around real questions customers ask. Publish original information when you have it. Make authorship and sources clear. And don’t rely entirely on owned media when third party editorial coverage can provide additional evidence about a company’s expertise and relevance.
The important part is consistency. A company can have an excellent website and still be poorly represented across AI systems if the wider web says something different about it.
Publishers face the same test, with different stakes
For publishers, the calculation is a little different but no less urgent.
A publication’s business model has traditionally depended on being the destination, not just the source. When an AI system summarizes a news story without sending a reader back to the original article, the traffic that once supported that reporting can disappear.
The realistic response isn’t simply to fight the shift. Publishers also need to make sure that when their reporting is used by an AI system, the original publication is clearly identifiable as the source.
That makes the basics more important, not less: clean article structures, clear sourcing, recognizable bylines, original reporting, distinctive analysis and information that cannot easily be replaced by a dozen copied summaries.
Aggregation can still have a place. But a publisher that only rearranges information available elsewhere gives an AI system little reason to point back to the original outlet.
The visibility fight has quietly moved
None of this replaces the fundamentals of good content, good journalism or good SEO. If anything, it raises the price of mediocrity.
When an AI system generates an answer, it has to decide which pieces of information are useful enough to include and which sources are worth drawing from. Brands and publishers therefore have another visibility problem to solve alongside rankings: becoming a source that is consistently present when the right questions are asked.
The companies that understand that shift early will not stop doing SEO. They will simply stop treating rankings as the only definition of being visible.
Visual idea: A horizontal flow diagram titled “The New Search Journey”, showing six connected stages: Search Query → AI Retrieval → Source Evaluation → Brand Mention → Citation → User Action. A secondary branch beneath Source Evaluation should show four trust signals: authority, entity clarity, content structure, and source diversity. The visual should look like a clean editorial infographic rather than a crowded marketing dashboard.
