The AI Search Framework: Visibility → Citability → Retrievability

Pepper’s operating model for winning in AI search – and what’s breaking most brands at every stage.

Your Map Through the Machine: What This Article Covers

• The shift that broke traditional search strategy
• The V→C→R Framework – defined
• Stage 1: Visibility – Can LLMs even see you?
• Stage 2: Citability – Do LLMs trust you enough to quote you?
• Stage 3: Retrievability – Can LLMs actually use your content?
• The benchmark: What 110 enterprise companies got wrong
• The LLM Retrieval Score formula
• Industry Updates: What’s changing right now
• The V→C→R Diagnostic Checklist
• FAQ
• YouTube Script: 3-Minute Video

The Game Has Changed. Most Brands Haven’t.

Your buyers aren’t typing keywords anymore. They’re asking ChatGPT which enterprise CDP to buy. They’re asking Perplexity which SaaS tool handles compliance best. And they’re getting a confident, sourced answer in under ten seconds.

That answer either has your brand in it – or it doesn’t.

Here’s what most CMOs don’t yet know: winning AI search isn’t a single problem – it’s a three-stage pipeline. You need to be visible to LLMs, trusted enough to be cited by them, and technically structured so they can actually retrieve your content. Miss any one stage, and you’re invisible.

Pepper calls this the V→C→R Framework: Visibility → Citability → Retrievability. It’s the operating model behind Search Everywhere Optimization – Pepper’s approach to making your brand the answer across every AI surface where a buyer might find you.

What Is the V→C→R Framework?

Think of it as a funnel – but unlike a marketing funnel, the stages are sequential and non-negotiable. You can’t be cited without being visible. You can’t be retrieved without being citable. Every brand’s AI search problem lives somewhere in this pipeline.

Stage 1: Visibility – Can LLMs Even See You?

Visibility is the most misunderstood stage. Most marketing teams assume that because they rank on Google, they’re visible to AI. They’re wrong.

LLMs build their knowledge from two sources: training data (what the model learned before deployment) and retrieval-time data (what it finds when it searches the web in real-time via RAG). You need presence in both.

Training-Time Visibility: Being Known Before the Query

Training data favours Wikipedia, long-form content on high-DA sites, Reddit threads, review platforms like G2, YouTube transcripts, and PR coverage. If your brand isn’t in these sources, you effectively don’t exist to the model – even if your website is excellent.

Retrieval-Time Visibility: Being Found During the Query

Perplexity, ChatGPT Search, Bing Copilot, and Gemini all use Retrieval-Augmented Generation (RAG). When a user asks a question, the LLM searches the web in real time, fetches results, and synthesises an answer from them. Your URLs must be crawlable, indexed, and linked from authoritative sources for this to work.

What Breaks Visibility

• Over-indexing on branded SEO while being absent in non-branded category queries
• Weak entity footprint – not listed on Wikidata, Crunchbase, or analyst reports
• No presence in third-party authoritative lists (G2, Capterra, industry roundups)
• AI crawlers blocked in robots.txt or content hidden behind JavaScript rendering

How to Fix It

1. Create a Wikipedia page and Wikidata entity for your brand – this single action feeds Google Knowledge Graph, Siri, and most LLM training pipelines.
2. Claim and fully populate G2, Crunchbase, and industry directory listings.
3. Run a PR programme targeting Search Engine Journal, MarTech, and vertical trade press – these are high-DA domains LLMs actively cite.
4. Ensure GPTBot and Perplexitybot are not blocked in your robots.txt.
5. Publish content that wins non-branded, category-level queries — not just your brand name.

Stage 2: Citability – Do LLMs Trust You Enough to Quote You?

You can be visible and still never get cited. Citability is about trust – the signals that tell an LLM your content is accurate, authoritative, and worth including in an answer.

The hard truth: 77% of enterprise companies in Pepper’s benchmark study hide their best insights in PDFs or gated content. That content might as well not exist for any LLM.

What Makes Content Citable

LLMs are pattern-matching for authority. There are five signals they weight most heavily:

• Expert attribution – named authors with credentials, job titles, and institutional affiliations
• Data and statistics – original, sourced numbers that support a clear claim
• Definitional authority – clearly structured ‘What is X?’ content that owns a concept
• Structured formatting – TL;DRs, bullets, numbered lists, Q&A blocks that can be extracted atomically
• Cross-source agreement – when multiple trusted sources make the same claim, LLMs weight it higher

Myth vs. Reality

Real-World Example

One company in Pepper’s client portfolio – a B2B SaaS brand – was seeing a dramatic drop in website traffic from AI search. Within their first content overhaul with Pepper, their team restructured 40 blogs to lead with direct answers, added expert attribution, and embedded FAQ schema. High-intent traffic from AI sources increased significantly because the content was now actually answering the questions LLMs were receiving.

Stage 3: Retrievability – Can LLMs Actually Use Your Content?

Retrievability is the most technical stage – and the one most brands have done nothing about. It’s the difference between content an LLM can see and content an LLM can use.

The critical distinction: traditional SEO creates content. GEO engineers the retrievability of that content. You can publish the world’s best article on enterprise security, and an LLM will still skip it if the page has no schema, the content isn’t chunked, and your site architecture is fragmented across three subdomains.

How RAG Actually Works

When a user asks Perplexity or ChatGPT Search a question, the system runs a live web search, fetches the top results, breaks each page into 300-500 word semantic chunks, scores each chunk for relevance, and synthesises an answer using the highest-scoring chunks. The sources of those chunks become the citations.

Your content needs to be structured so those chunks each answer a complete question on their own.

The LLM Retrieval Score Formula

Pepper’s framework identifies five multiplicative factors that determine how well content is retrieved and cited:

LLM Retrieval Score ∝ (Chunking × Structure × Schema × Source Weight × Trust Signals)

What Breaks Retrievability

• No llms.txt file – AI crawlers don’t know which content to prioritise
• Content fragmented across microsites and subdomains (no coherent architecture)
• Publishing too late – if your competitors published the definitive piece first, they own the citation
• JavaScript-rendered content that AI crawlers can’t parse
• No FAQ schema – the single most impactful structural addition for LLM extractability

The Benchmark: What 110 Enterprise Companies Got Wrong

Pepper’s AI Search Mistakes Benchmark – drawn from analysis of 110 enterprise companies with 500+ employees – maps exactly where brands fail across all three V→C→R stages. The patterns are consistent. They are also fixable.

The data tells a clear story: enterprise brands are losing AI search not because AI search is hard – but because they haven’t applied a systematic framework to it. The V→C→R model turns a scattered problem into three actionable workstreams.

Industry Updates: What’s Changing in AI Search Right Now

The V→C→R landscape is shifting fast. Here are the five developments every marketing leader needs to track heading into H2 2026.

1. ChatGPT is Now the #1 Starting Point for B2B Research

G2’s survey data from Index’26 showed that ChatGPT has overtaken Google as the preferred starting point for software discovery searches. Claude has seen the fastest growth, up 21% between August 2025 and March 2026. The implication: if you’re only optimising for Google, you’re blind to where the buying journey now begins.

2. AI Overviews Are Compressing the Organic Funnel

Fortune 500 companies saw organic traffic from Google drop 30-40% on average in 2024-25. For some, the fall was 70-80%. This is the structural impact of AI Overviews – the answer appears above the links, and click-through rates for ranked pages collapse. Visibility without citability is now worthless.

3. Conductor’s 2026 CMO Survey: Enterprise Investment Is Accelerating

Conductor’s 2026 AEO/GEO CMO Investment Report confirmed that nearly all enterprise brands plan to increase investment in AI search visibility in 2026. The warning in the data: organisations already ahead are accelerating at a pace that will leave late movers unable to catch up. The compounding effect of citation history favours first movers.

4. RAG Architecture Is Rewriting How Source Authority Works

Perplexity, ChatGPT Search, and Gemini all use Retrieval-Augmented Generation. This means content published on high-authority domains – G2, Reddit, Wikipedia, YouTube, press coverage – carries disproportionate citation weight compared to brand websites. Your off-site content programme is now a core GEO asset, not a PR nice-to-have.

5. Fan-Out Queries Are the New Keyword

LLMs decompose complex queries into sub-queries before searching. When someone asks ‘What’s the best CDP for an enterprise retail brand?’, the model might search for ‘best CDP enterprise,’ ‘CDP retail use cases,’ and ‘CDP pricing comparison’ separately. You need content that answers each fragment independently – not just a single long-form page.

The V→C→R Diagnostic Checklist

Run this against your current content programme. Every ‘No’ is a gap an LLM is filling with your competitor’s content.

Visibility Checklist

• Do you have a Wikipedia page or Wikidata entity for your brand?
• Is your brand listed on G2, Crunchbase, and relevant analyst lists?
• Are you mentioned in press coverage on high-DA domains (DA 50+)?
• Are GPTBot, PerplexityBot, and ClaudeBot allowed in your robots.txt?
• Do you win non-branded, category-level queries in your market?

Citability Checklist

• Does every blog post have a named author with credentials?
• Do you lead every section with a direct answer (not a setup paragraph)?
• Do you have FAQ schema deployed across your key content pages?
• Do you publish original data, stats, or proprietary research?
• Do you own the definitional ‘What is X?’ queries in your category?

Retrievability Checklist

• Do you have an llms.txt file at your root domain?
• Is Organization, Article, and FAQPage schema implemented sitewide?
• Is your content structured in 300-500 word semantic chunks with clear H2/H3 headers?
• Is your site architecture unified (not fragmented across microsites)?
• Are you monitoring which prompts lead to brand citations – and closing the loop?

FAQ

What is the V→C→R Framework in AI search?

The V→C→R Framework is Pepper’s proprietary AI search operating model. It stands for Visibility (can LLMs detect your brand’s existence?), Citability (will LLMs select your content as trustworthy and quote-worthy?), and Retrievability (can LLMs technically parse and use your content to construct an answer?). The three stages are sequential – failing any one makes the others irrelevant.

How is AI search different from traditional SEO?

Traditional SEO optimises for ranking – the goal is to appear in a list of links. AI search optimises for answering – the goal is to be the source an LLM synthesises its response from. The signals are different: AI search weights entity recognition, content structure, schema markup, and cross-source authority rather than keyword density and backlink count alone.

What is retrievability in the context of GEO?

Retrievability refers to the technical and structural properties that allow AI crawlers and RAG systems to find, parse, and use your content when generating answers. It includes having an llms.txt file, implementing schema markup, structuring content in 300-500 word semantic chunks, and ensuring your site architecture is coherent enough for AI crawlers to map.

Why are most enterprise brands failing AI search?

Pepper’s benchmark data from 110 enterprise companies shows that 72% over-index on branded SEO and are absent from non-branded category queries. 77% hide insights in gated content. 69% have no llms.txt. The pattern is consistent: enterprise brands have invested heavily in traditional SEO but applied almost none of those systems to AI search.

How do I know if my brand is visible in AI search?

Run your core category queries – ‘best [product category] for enterprise,’ ‘what is [concept you should own],’ ‘[your brand] vs [competitor]’ – in ChatGPT, Perplexity, and Gemini. If your brand doesn’t appear in the generated answers, you have a visibility or citability gap. Pepper’s Atlas platform tracks this systematically across prompts, engines, and competitors.

Your Next Move

The brands winning AI search in 2026 aren’t waiting for the algorithm to figure them out. They’re engineering their visibility, citability, and retrievability – systematically.

Run Pepper’s V→C→R diagnostic against your current content programme. If you’re scoring below a 50% on any stage, you have a structural gap – and your competitors are filling it.

pepper.inc  |  AI Search Strategy  |  GEO & AEO  |  Search Everywhere Optimization