SEO in AI search has collapsed 60-80% in the past two years.
A BCG study published this year surveyed 9,000 consumers across nine countries and found that 33% now discover brands through AI agents. One in three of your customers is not searching. They are asking.
They are typing questions into ChatGPT, Gemini, Claude, or Perplexity, and those platforms are generating answers. Recommendations. Brand comparisons. "What is the best option for X" queries that used to end with a Google results page now end with a paragraph written by a language model.
The problem is not that this is happening. The problem is that you have no instrument to measure it.
Traditional measurement tools track rankings, impressions, and traffic. They are built for a system where content competes to appear in a list. AI engines do not produce lists. They produce synthesized answers, and the content that influences those answers is drawn from a source layer almost entirely different from the one that drives search rankings. Your current stack is measuring the wrong surface.
This post covers what a measurement discipline for that surface looks like, what the data actually shows when you run it, and what the prescription layer produces once you have a score.
---
The Structural Shift No Marketing Stack Accounts For
Search engines and AI engines are not the same system, and the distinction matters more than most brand teams currently treat it.
When a consumer searches Google for "best baby clothing brand for newborns," the ranking algorithm surfaces pages based on authority signals, backlinks, structured data, and content relevance. Your SEO investment is designed to influence those signals.
When that same consumer asks ChatGPT the same question, the model generates an answer synthesized from training data, retrieval sources, and in some cases live web content. The content that influences that answer is not your page rank. It is whether your brand appears in the source types those models draw from: review platforms, community discussions, editorial coverage, structured databases, and authoritative third-party references.
Those two source layers barely overlap.
The BCG finding is not a trend indicator. It is a distribution problem. If 33% of discovery is now happening in AI surfaces and your measurement stack has no visibility into those surfaces, you have a structural blind spot that compounds every quarter as that share grows.
The correct response is not to optimize faster for search. The correct response is to measure what is actually happening in the systems where your customers are asking questions.
---
What the Measurement Looks Like in Practice
We built Modern Discovery to answer a question that no existing tool could: where does a brand actually stand inside AI-generated answers?
The first thing we did was run it on ourselves.
Our AVI score came back lower than we anticipated. AVI stands for AI Visibility Index. It measures your brand's share of presence across generative AI answers, weighted for mention rate, citation rate, sentiment, and authority by model. We expected to perform reasonably well. We are an AI company that publishes regularly and operates in a clearly defined category. The score told us otherwise.
More revealing was the ASOV breakdown. ASOV stands for Agentic Share of Voice. It breaks your AI presence down platform by platform: ChatGPT, Gemini, Claude, Perplexity. When we looked at our ASOV, the results were wildly inconsistent across platforms. One model mentioned us regularly. Another had never heard of us. Same brand. Same query set. Same time window.
That split is not a quirk. It reflects the fact that different AI platforms pull from different source layers. A model that had encountered our brand through certain content types and sources gave us credit. One that had not indexed or referenced those sources treated us as if we did not exist. An aggregate score alone would not have told us which platform to fix first or why.
That is the operational difference between AVI and ASOV. AVI tells you how present you are. ASOV tells you where you are present and where you are invisible.
Then we ran it on other brands. Well-known brands. Household names across multiple product categories.
The results followed the same pattern. Brands with decades of market presence, massive advertising budgets, and dominant search rankings were underperforming, missing, or completely absent from AI-generated answers in their own categories. These are not startups with no brand equity. These are category leaders that had no idea they were invisible in the channel where one in three of their customers now makes decisions.
Think about that for a moment. A major brand spending millions on marketing has no view into what a third of its customers see when they ask an AI engine "what is the best option" in that brand's own category. The AI platform generates an answer. The brand may or may not appear in it. And nobody on the marketing team can tell you either way, because the instruments to measure it have not existed.
That is the gap Modern Discovery was built to close.
---
The Gap Between Being Mentioned and Being Cited
The most important distinction this measurement discipline surfaces is not between presence and absence. It is between being mentioned and being cited.
When we ran the pipeline on a well-known consumer brand, we found a pattern that has repeated in every category we have measured: mention rates that look healthy on the surface, and citation rates that are close to zero.
A mention means the model has heard of you. When someone asks a general question in your category, your brand name appears somewhere in the answer. A citation means the model references you as a specific, authoritative source in a direct answer to a purchase or recommendation query. In this brand's case, the distance between those two numbers was enormous.
That gap is where revenue lives.
A strong mention rate looks good if you have never seen a citation rate next to it. It stops looking good the moment you understand that mentions happen passively, baked into training data from years of brand presence, and citations happen when AI platforms have a reason to treat your brand as a credible, structured, retrievable source for a specific question. The content types that drive citations are not the same content types that drive mentions. They are more specific, more authoritative, and more often housed on platforms that AI models use as reference anchors.
This brand's low citation rate was not bad luck. It was a source footprint problem. The brand had broad general awareness baked into the models' training data, but almost no presence in the structured, third-party reference sources those platforms pull from when they generate specific recommendations.
We saw similar patterns in every category we measured. Even brands actively publishing content had citation voids across query categories where their products were directly relevant. The AI platforms returned answers with no mention of them at all. Not a low score. Absence.
That is the non-obvious thing about this problem: you can be producing content consistently and still have a citation void, because the content you are producing is not landing in the source types AI platforms use to answer specific questions. The measurement tells you which categories the void exists in. That is where the diagnosis starts.
---
What the Prescription Actually Says
The score is not the product. What comes after the score is.
After the measurement cycle runs, the Diagnose layer produces three things. First, it maps your citation voids by query category: awareness queries, comparison queries, recommendation queries, authority queries, use-case queries. The void in a recommendation category is a different problem than a void in a comparison category, and the fix is different. Second, it identifies competitive deltas: where a competitor is being cited in answers where your brand is absent. Third, it surfaces consensus gaps: query areas where no brand in your category is being cited with consistency.
That diagnostic output is what makes the prescription specific rather than generic.
For the well-known brand with the citation gap, the Diagnose layer identified the source types those platforms were pulling from when they generated recommendations in that product category. The brand was not present in them. The prescription named the specific source categories: structured third-party review platforms where the brand had no profile, editorial guides on publications the models reference as authority sources, community discussions on platforms indexed by those models, and structured data that would allow the models to retrieve specific product details when answering use-case queries.
The output was not "you need more content." It was: these are the specific source types you are missing from, these are the query categories where each gap creates a citation void, and this is the sequence to close them, ordered by the platform weight of each gap.
For our own brand, the prescription output told us which specific content types to produce and where they needed to live so that AI platforms would have a reason to reference us in the query categories where we had voids. The content we are building is not SEO content. It is citation-source content, built for the platforms and source types that AI models actually pull from.
That is what a prescription looks like in practice. Not a list of general recommendations. A gap-by-gap closure map with platform specificity and sequencing.
---
AI Competitive Intelligence as a Measurement Discipline
It is worth being precise about what this is and what it is not, because the vocabulary around AI and brand presence is still loose.
GEO (Generative Engine Optimization) and AI SEO are optimization tactics. They tell you what to produce and how to structure it to improve your AI presence. They are useful.
AI Competitive Intelligence is a measurement discipline. It tells you where you stand relative to your competitive set, what the gap looks like platform by platform, and what specifically must change. The instruments are different. The diagnostic layer is different. The output is different.
An SEO consultant can tell you to produce FAQ content optimized for AI surfaces. AI Competitive Intelligence tells you whether that FAQ content is producing citations, which query categories it is missing, which platforms are still not referencing it and why, and how your citation share compares to the brands competing in the same answer space.
That distinction matters for how you staff, budget, and measure the work. An optimization tactic is a project with a start and an end. A measurement discipline is something you run every quarter, with instruments that track whether the last quarter's actions actually moved the number.
The market for AI search tools is growing fast. The more immediate point for a marketing leader right now is that 33% of consumer brand discovery is already happening in surfaces your current stack cannot see. That is not a 2030 problem. It is a 2026 problem.
Modern Discovery is the platform we built to run this measurement discipline. The instruments are AVI and ASOV. The cycle is Measure, Diagnose, Prescribe. We built the pipeline, ran it on our own brand, ran it on brands across multiple industries and product categories, and refined it against what those runs actually produced. This is not a demo environment with sample data. The output has been consistent: every brand we measured had gaps they did not know existed, and the prescription layer gave them a specific, ordered plan to close them.
---
What to Do If Your Brand Has This Problem
Most brands that run the measurement cycle for the first time find the same thing: their AI presence is inconsistent across platforms, their citation rate is far lower than their mention rate, and they have citation voids in query categories they assumed were covered because they had been publishing content in those areas.
The awareness that you have this problem is the start of the discipline, not the solution.
[Get your free AI Shortlist Score](https://modernai.io/ai-shortlist-score) to see where your brand stands across AI platforms today. If you want the full measurement cycle and the prescription layer, you can [learn more about Modern Discovery](/discovery) or reach out to info@modernai.io.
The brands that measure and act on this now will have specific knowledge their competitors do not. That advantage does not compound forever. AI platforms are already being used at scale in discovery. The window to build a presence in those surfaces before your competitors do is not wide.
---
Matt Kott is the founder of Modern AI, a company building AI Competitive Intelligence infrastructure for enterprise marketing teams. Modern Discovery is the company's AI visibility measurement and prescription platform.---
> Results referenced are based on observed patterns in available data and are not guarantees of performance. Individual outcomes vary based on data quality, implementation, and market conditions. Modern AI recommends independent validation of any metric before business decisions are made.
---