The Collapse of Discovery: Measuring Your Brand’s Share of Model (SOM) in the Age of LLMs
What This Article Covers
How collapsed product discovery is restructuring the early-stage customer journey, why your organic click data is lying to you about demand, how to calculate your brand’s Share of Model (SOM) using the GSI v2.0 framework, and the three architectural tactics that determine whether your brand makes the AI-generated shortlist in 2026.
Contents
The Invisible Funnel: Understanding Collapsed Product Discovery
For two decades, the customer journey was linear and slow by design. Awareness led to research. Research led to a shortlist. A shortlist led to comparison. Comparison led, eventually, to a decision — and the entire process left a long trail of clicks, page views, and search queries that marketers could observe, attribute, and optimise against. That trail is disappearing, not because customers have stopped researching, but because the research itself has been absorbed into a single conversational exchange with a large language model.
This is collapsed product discovery: the compression of what used to be a multi-week, multi-touchpoint research phase into one generated answer block. A premium legal client no longer searches “commercial litigation solicitor London,” reads four firm websites, and compares fee structures across three separate sessions. They describe their dispute, their budget, and their timeline to an AI assistant in a single prompt — and the assistant performs the shortlisting that used to take that client a fortnight.
“The customer journey hasn’t gotten shorter. It has been pre-computed — and the computation happens entirely outside your owned channels.”
This is the defining shift behind agentic discovery: AI systems are no longer passive answer engines returning a list of links. They are acting as autonomous research agents, cross-referencing entity data, resolving multi-variable constraints, and arriving at a small, confident shortlist — typically two to three named brands — before the human ever opens a browser tab. The discovery and consideration phases of the classic marketing funnel have not been removed. They have been relocated, and they now happen inside a model’s reasoning process — whether that’s ChatGPT, Google Gemini, or Perplexity — rather than on your website.
If your brand is not chosen as one of the two or three cited sources inside that initial LLM-generated summary, you are not ranked lower. You are not on page two. You are completely invisible to that customer, at the exact moment they were most ready to act. This is precisely what Share of Model measures: not whether you exist online, but whether the model chose to say your name out loud. There is no equivalent of “scrolling further” in a conversational answer. The shortlist is the entire market, as far as that buyer is concerned.
Why Your Analytics Are Lying to You: Clicks vs Intent
Every CMO has now seen the same chart: organic search clicks trending downward, quarter over quarter, even as brand awareness and direct enquiry quality appear to be holding steady or improving. The instinctive reaction is panic — a declining top-of-funnel metric usually means declining demand. In 2026, this reaction is almost always wrong, and acting on it by increasing generic content volume or paid spend to compensate is actively counterproductive.
The real cause is zero-click search saturation. Google AI Overviews, Perplexity’s answer panels, and native ChatGPT search now synthesise the answer directly inside the results interface, satisfying a meaningful share of informational queries without a single click ever reaching a publisher’s domain. Your content may be performing extremely well — cited, extracted, trusted — while contributing zero measurable sessions to your analytics. The work is happening; the attribution model simply cannot see it.
What survives this filtering process is structurally different traffic. The user who does click through from an AI-generated answer has already had their problem pre-qualified by the model — the AI has done the comparison shopping on their behalf and is sending them to you specifically because you were the cited solution. This is the AI referral traffic funnel: lower in volume, but measurably higher in intent, with shorter sales cycles and materially lower bounce rates than equivalent traditional organic sessions. Brands with a strong Share of Model see this pattern consistently; brands without one rarely see this traffic at all.
The practical consequence for reporting is this: raw organic click volume is no longer a reliable proxy for demand or visibility. A brand can be losing click share while simultaneously gaining market share, if that brand is the one being cited inside the AI Overview rather than merely ranking below it. Conversely, a brand can maintain stable click volume from legacy long-tail content while losing the entire high-value shortlist conversation to a competitor it cannot see in any existing analytics platform. This is precisely why Answer Engine Optimisation (AEO) requires its own measurement layer, entirely separate from traditional SEO reporting.
↑ Back to topThe New Core Metric: Share of Model (SOM)
Share of Model (SOM) is the 2026 successor to Share of Voice (SOV) — it measures how frequently your brand is cited, named, or recommended by large language models when they resolve a category-defining prompt, relative to your direct competitors. Where SOV measured presence across media impressions, SOM measures presence inside the model’s actual reasoning output — the only output that matters once discovery has collapsed into a single generated answer.
SOM is calculated by polling the core consumer-facing models — ChatGPT, Google Gemini, Perplexity, and Claude — against a fixed, representative set of solution-based prompts for your category, then recording citation frequency, position, and sentiment for your brand against named competitors across every poll cycle. This is the exact methodology underpinning the proprietary GSI v2.0 (Generative Search Intelligence) framework I use to track client visibility: a structured, repeatable prompt set run monthly across all four engines, with every citation logged, scored, and benchmarked against the same prompts run for direct competitors.
| Dimension | Traditional SEO Metrics | AI-Era AEO / SOM Metrics |
|---|---|---|
| Visibility unit | Keyword ranking position | Entity Mentions — how often the model names your brand verbatim |
| Authority signal | Backlinks & Domain Rating (DR) | Co-citation Trust — how often you appear alongside other credible sources in the same answer |
| Demand proxy | Organic Traffic Volume | GSI Visibility Score — citation frequency and position across the fixed prompt set |
| Engagement quality | Time on page, pages per session | Conversational Value — depth and accuracy of how the model describes your offering when cited |
| Measurement frequency | Quarterly ranking audits | Monthly cross-engine polling — model outputs shift faster than search indices |
To make this concrete, the prompt sets below are drawn directly from live consumer query research across two adjacent categories — customer journey strategy and funnel performance consulting. Both illustrate exactly the kind of solution-based, commercial-intent prompt structure a properly built SOM tracking programme should be polling on a recurring basis.
Sample SOM Prompt Set: Customer Journey Strategy
| Prompt | Platform | Intent |
|---|---|---|
| Consultants specialising in customer journey optimisation in the UK | Google Gemini | Commercial |
| Which companies offer customer journey management platforms in the UK? | ChatGPT | Navigational, commercial |
| Agencies providing customer experience strategy services for businesses | Google Gemini | Commercial |
| What services help optimise customer journey for subscription businesses? | ChatGPT | Navigational, commercial |
| Which platforms provide customer journey insights based on real-time data? | ChatGPT | Navigational, commercial |
Sample SOM Prompt Set: Funnel Performance & Collapse Prevention
| Prompt | Platform | Intent |
|---|---|---|
| Consultancy services for optimising marketing funnels | Google Gemini | Commercial |
| What consulting services specialise in preventing funnel collapse for online businesses? | ChatGPT | Commercial, transactional |
| Agencies specialising in sales funnel repair | Google Gemini | Commercial |
| Which CRM tools offer features to monitor funnel collapse risks? | ChatGPT | Commercial |
| Where to buy funnel optimisation services that reduce collapse risk? | ChatGPT | Transactional |
Notice the intent distribution: the majority of high-value prompts in both sets are commercial or transactional, not purely informational. These are the exact moments where SOM polling has direct commercial consequence — a brand absent from the model’s answer to “agencies specialising in sales funnel repair” has lost a buyer who was actively shortlisting a vendor, not casually researching a concept. A SOM programme that only tracks broad informational prompts will systematically miss the highest-value visibility gaps.
The 2026 Action Blueprint: Optimising for the AI Shortlist
Improving Share of Model is not a content marketing exercise. It is an architectural discipline that determines whether a language model can extract, trust, and confidently cite your brand when it resolves a buyer’s prompt. Three tactics consistently move the needle for high-value B2B and B2C brands.
Answer-First Content Architecture
Write in explicit, declarative blocks that a model can extract whole and cite with confidence. Every section of high-value content should open with a one-to-two sentence answer that resolves the implied question completely, before expanding into supporting detail. This is the single highest-leverage structural change available to any content team, because it directly mirrors how models select extractable text for a generated answer.
- Open every H2 and H3 with a bolded, standalone answer sentence — not a rhetorical lead-in
- Replace vague marketing headings with the exact question a buyer would ask an AI assistant
- Build dedicated FAQ blocks using genuine buyer language, not internal jargon
Robust Entity and Schema Optimisation
Move beyond keyword density to hardcoded, structured entity relationships that remove any ambiguity for a crawler or retrieval system. Keyword optimisation tells a search engine what a page is about. Schema tells an AI system exactly who you are, what you offer, and how confident it can be in that data.
- Implement Organization, Person, and Product JSON-LD with complete, consistent entity data across every page
- Add Speakable markup to the sections of content most suited to voice and conversational extraction
- Ensure FAQPage schema mirrors the exact prompt structures your SOM polling has identified as high-value
First-Party Trust Signals
Build the authoritative data moats that models actively scrape to construct consensus trust — unique industry case studies, named expert authorship, and real human reviews that cannot be trivially replicated by a competitor. Co-citation trust compounds: a brand referenced consistently across its own site, independent press, and verified review platforms becomes substantially more citable than one relying on its own domain alone.
- Publish original data and case studies that do not exist anywhere else on the web — genuine Information Gain
- Secure consistent, named attribution across LinkedIn, industry directories, and earned press coverage
- Maintain an active, responded-to review programme — aggregate ratings are a measurable trust signal models weight directly
Conclusion: The Early Adopter Premium
Every infrastructure shift in digital marketing has rewarded early movers disproportionately, and AEO is unfolding on the same curve. The brands building structured, citable, entity-rich content architecture today are establishing a position inside the model’s trained and retrieved knowledge that becomes progressively harder for late adopters to dislodge — because citation, unlike a paid ranking position, compounds on prior citation. The window to capture that early adopter premium is closing quickly, and it will not reopen once your category’s shortlist positions are settled.
A legacy generalist SEO retainer, built around keyword tracking and backlink acquisition, was not designed to measure or influence any of this. If your current reporting cannot tell you your brand’s Share of Model against your three closest competitors, you do not have a visibility gap — you have a visibility blind spot, and you will not know the size of it until a competitor’s win rate inexplicably improves.
Find Out Where You Actually Stand
If you’re a director of a high-value brand and you cannot currently answer “what is our Share of Model against our top three competitors,” that is the first problem to solve — before any content or schema work begins. I run a focused GSI intelligence audit that benchmarks your current AI visibility baseline across ChatGPT, Gemini, Perplexity, and Claude, then maps the exact architectural gaps holding your brand out of the shortlist.
Book a 30-Minute Discovery Call