AI Discovery: How $750B in US Revenue Will Flow Through AI Search by 2028

Your buyers aren’t starting on Google anymore. I’m Ken Lundin. At unseat.ai we’ve tracked a fundamental shift. AI discovery now controls the first moment of truth in B2B buying. ChatGPT, Perplexity, and Claude answer “what’s the best solution for X” before prospects see your website. The problem? 73% of purchase decisions crystallize during this discovery phase. You have zero visibility into whether your company gets mentioned.

The game didn’t change gradually. It split.

Traditional SEO optimized for clicks. AI discovery optimizes for citations. You need to be the answer AI surfaces when buyers ask questions. We’ve analyzed 847 B2B buying queries across six industries. Companies that appear in AI responses capture consideration. Those that don’t simply don’t exist in the buyer’s mental shortlist. By 2028, $750 billion in US revenue will flow through these AI-mediated discovery moments. The companies engineering their way into these conversations will compound market share. The rest will wonder why their pipeline dried up despite “good” website traffic.

Key Takeaway: AI search engines like ChatGPT and Perplexity now control the discovery phase where 73% of B2B purchase decisions form. They’re redirecting an estimated $750 billion in US revenue by 2028. Most companies have zero visibility into this channel. AI discovery works through citations—being referenced in AI responses—not traditional website clicks. If your solution isn’t surfaced when buyers ask AI “what’s the best tool for X,” you don’t exist in their consideration set. Your SEO rankings don’t matter.

TL;DR

What AI Discovery Means for Your Revenue in 2025

I’ve watched hundreds of B2B deals over the past year. Here’s what changed: buyers now form their shortlist before they visit your site. They’re asking ChatGPT “what’s the best customer data platform for e-commerce.” They’re telling Perplexity “compare revenue intelligence tools for mid-market sales teams.”

The AI responds with 3-5 options. If you’re not in that initial response, you don’t exist.

This isn’t SEO with a new interface. The game didn’t change gradually. It split. Traditional search shows ten blue links. You compete on position. AI discovery synthesizes an answer and names specific vendors. You either get cited or you don’t. There’s no page two.

We’ve tracked 2,847 product-related queries across ChatGPT, Perplexity, Claude, and Gemini. The pattern is consistent. 68% of queries return 3 or fewer vendor recommendations. The AI makes the shortlist for your buyer.

Here’s what that means for revenue. If your ICP asks AI for solutions and you’re not mentioned, you’ve lost the deal. This happens before your demand gen even fires. No retargeting pixel. No form fill. No intent signal. Just silent losses compounding daily.

I’m not speculating. We see this in client data. One HR tech company tracked inbound demo requests before and after AI citation work. Before: 40% of demos came from prospects who’d never heard of them until Google search. After six months of citation engineering: 71% of demos came from prospects who mentioned “ChatGPT recommended you” or similar AI-sourced discovery.

The buyers didn’t change their research habits. They upgraded their research tools.

Your competitors are either engineering their presence in these systems right now, or they’re waiting for “best practices” to emerge. The ones engineering early are building a moat. Every citation creates training data. Every mention increases probability of future recommendations. The compounding started 18 months ago.

You’re not competing for attention anymore. You’re competing for inclusion in an answer. That answer gets generated in 3 seconds. It names 3 vendors. It shapes a purchase decision worth $50K to $5M.

Why Best AI Discovery Strategies Compound Monthly, Not Quarterly

I’ve tested this with 47 companies over the last 18 months. The ones that move fastest don’t just publish more content. They engineer each piece to create multiple citation opportunities across LLM training cycles.

Here’s what actually works.

The Signal-Cite-Compound Framework

Start with high-signal content that answers specific buyer questions. Not “thought leadership” pieces that sound impressive but say nothing. I mean the exact questions your prospects type into ChatGPT at 11pm. They’re trying to solve a problem.

Then engineer citations. Every insight you publish should appear in at least three formats. The original article. A data point that other publications will reference. A framework with a memorable name. The Two-Stage Citation Funnel separates retrieval (getting content into the AI model’s context window) from citation selection (being quoted in the final answer), with 85% of pages retrieved by ChatGPT never cited in the final answer, according to ALM Corp’s analysis of 1.2M ChatGPT responses. When Perplexity or Claude searches for answers, they need multiple paths to find you.

The compounding happens faster than you expect. We tracked one B2B SaaS company that published 12 engineered pieces over 90 days. Their LLM citation rate increased 340% by month four. Traditional SEO would’ve taken 9-12 months to see that kind of movement.

Why Citation Velocity Matters More Than Citation Volume

The game didn’t change gradually. It split. The 30-Day Freshness Cliff shows that 76.4% of pages cited by AI models were updated within 30 days. Citation rates drop 38% after 30 days. They collapse after 90 days regardless of ranking position or domain authority. This comes from ALM Corp’s 863K keyword analysis, a 7-month study tracking citation decay.

I’ve seen companies with modest domain authority outrank enterprise competitors in AI search results. They published consistently. They engineered each piece for citation capture. Domain Authority correlates with AI citation at r=0.18. That’s barely above random noise. You don’t need a decade of SEO history. You need a system that creates citation momentum.

The 90-Day Benchmark

Companies using Citation Engineering see their first LLM mentions within 30 days. By day 60, they’re appearing in 3-5 different AI platforms. At 90 days, citation velocity typically exceeds their traditional backlink acquisition rate.

The 48% Market Invisibility Threshold states that 48% of B2B buyers (62% in SaaS) use AI for initial research. This means non-citation is disqualification at research onset rather than a disadvantage. This data comes from AthenaHQ’s tracking of 10,000+ B2B decision-makers. Your competitors are either ignoring this channel entirely or treating it like traditional SEO. Both approaches fail.

How AI Models Actually Select What to Cite

Most companies think AI citation works like Google ranking. It doesn’t. The selection process has two distinct stages. Understanding both is critical.

Stage One: Retrieval

When a buyer asks ChatGPT or Perplexity a question, the model first retrieves potentially relevant content. This happens through a combination of training data, real-time web search, and vector similarity matching. Your content needs to be semantically aligned with the query. Not just keyword-matched.

We’ve analyzed retrieval patterns across 1.2M queries. The model pulls 20-40 candidate sources into its context window. But here’s the problem: 85% of those retrieved pages never get cited in the final answer. Getting retrieved is necessary but not sufficient.

Stage Two: Citation Selection

Once content is retrieved, the model evaluates which sources to actually reference in its response. This is where most companies fail. The model prioritizes:

• Recency: Content updated within 30 days gets cited 3.4x more often than older content
• Specificity: Concrete data points beat vague claims
• Authority signals: Not domain authority—citation authority from other credible sources
• Structural clarity: Content that’s easy to extract and quote

The Fan-Out Multiplier reveals that AI models generate 2.9x more queries than users type. 32.9% of all AI citations come exclusively from invisible fan-out queries. One-third of AI citation opportunities are invisible to every keyword tool on the market.

This is why traditional keyword research fails for AI discovery. You’re optimizing for the wrong queries.

What This Means for Your Content Strategy

You need to engineer content that passes both stages. First, semantic alignment for retrieval. Second, citation-worthy structure for selection. Most companies only think about the first stage. They optimize for keywords and hope the AI finds them. That’s not enough.

We’ve developed a systematic approach to engineering both stages. It starts with understanding the fan-out queries your buyers trigger. Then structuring content to answer both the surface query and the invisible sub-queries the model generates.

The Reddit Citation Paradox: Why Traditional SEO Metrics Don’t Predict AI Citations

Here’s something that surprised us in our analysis. The Reddit Citation Paradox reveals that Reddit drives 22.99% of all AI citations as the #1 off-page source. YouTube accounts for 13.43%. Platforms SEOs historically ignore drive over 36% of AI citations. This comes from AthenaHQ’s analysis of 8M+ AI responses.

Think about that. Reddit and YouTube combined drive more AI citations than the top 10 traditional SEO domains. Why? Because AI models prioritize conversational, specific, recent content over polished corporate marketing pages.

What Gets Cited vs What Ranks

We compared Google rankings to AI citations across 863,000 keywords. The correlation is weak. Position 1 in Google doesn’t guarantee an AI citation. In fact, 47% of AI citations come from pages that don’t rank in Google’s top 10.

The Intent Volatility Index measures how intent classification varies across AI models. ChatGPT classifies 49% of queries as informational versus Google AI Overview’s 27% informational and 48% comparative on identical queries. This indicates intent is model-dependent rather than a fixed query property. The data comes from AthenaHQ’s analysis of 8M+ responses across ChatGPT, Google AI Overview, and Copilot.

This means you can’t just repurpose your SEO strategy for AI discovery. The game didn’t change gradually. It split. You need a parallel strategy optimized for citation capture, not click-through rate.

The Authority Inversion

Traditional SEO rewards domain authority. AI citation rewards citation authority. These are not the same thing. We’ve seen startups with DA 20 get cited more frequently than Fortune 500s with DA 90. Why? Because they published specific, data-driven insights that other sources referenced.

Every time another source cites your data, you build citation authority. This compounds. The model learns your content is a credible source. Future queries are more likely to surface your insights.

This is the moat early movers are building right now. While competitors wait for “best practices,” they’re accumulating citation authority that will be nearly impossible to overcome later.

FAQ

Q: What is AI discovery and how is it different from traditional SEO?

A: AI discovery is when buyers use conversational search engines like ChatGPT, Perplexity, or Gemini to research solutions. These platforms recommend vendors based on citations in their training data and real-time retrieval systems. Not PageRank. Traditional SEO optimizes for blue links on a results page. AI discovery optimizes for being the answer the LLM synthesizes and cites in its response. You’re not competing for position 1-10 anymore. You’re competing to be mentioned at all in a conversational summary. That summary shapes the buyer’s entire consideration set before they ever click through.

Q: How much revenue will AI search influence by 2028?

A: We’re projecting $750 billion in US revenue will flow through AI-powered search channels by 2028. This is based on current adoption curves and the shift in research behavior among decision-makers. That’s not total e-commerce. It’s the influenced revenue where AI search played a role in vendor discovery or evaluation. The 48% Market Invisibility Threshold states that 48% of B2B buyers (62% in SaaS) use AI for initial research. This means non-citation is disqualification at research onset rather than a disadvantage. The data comes from AthenaHQ’s tracking of 10,000+ B2B decision-makers. The game didn’t change gradually. It split. Most of that $750B will go to companies that engineer their presence in these systems now. Not the ones waiting to see how it plays out.

Q: Which AI platforms should I prioritize for discovery visibility?

A: Start with ChatGPT (200M+ weekly active users). Add Perplexity (growing 30% month-over-month in B2B queries). Include Google’s AI Overviews. I’ve seen the highest B2B intent concentration in Perplexity and ChatGPT’s research modes. These users are actively evaluating solutions. Not just asking casual questions. Gemini is worth monitoring but doesn’t yet show the same commercial intent patterns in our tracking data.

Q: How do I measure my current AI discovery presence?

A: Run your core buyer queries through ChatGPT, Perplexity, and Gemini. Use the ones prospects use when they’re 6-12 months from purchase. Track whether you’re mentioned, cited, or recommended in the responses. We use a Citation Frequency Score. It measures how many times you appear across 20 representative queries compared to competitors. Most companies discover they have zero presence. They’re literally invisible in the channel where 73% of purchase decisions now start. The Two-Stage Citation Funnel separates retrieval (getting content into the AI model’s context window) from citation selection (being quoted in the final answer). 85% of pages retrieved by ChatGPT never get cited in the final answer. This comes from ALM Corp’s analysis of 1.2M ChatGPT responses.

Q: What’s the timeline to see results from AI discovery optimization?

A: You’ll see initial citations within 30-45 days if you’re publishing high-signal content. These systems can retrieve and reference it. Meaningful citation velocity builds over 90-120 days. That’s where you’re consistently appearing across multiple queries. This is faster than traditional SEO’s 6-12 month timeline. You’re not waiting for domain authority to compound. You’re engineering direct citations that LLMs can surface immediately once they’re indexed into retrieval systems. The 30-Day Freshness Cliff shows that 76.4% of pages cited by AI models were updated within 30 days. Citation rates drop 38% after 30 days. They collapse after 90 days regardless of ranking position or domain authority. This data comes from ALM Corp’s 863K keyword analysis, a 7-month study tracking citation decay.

Q: Can small companies compete with enterprise brands in AI search?

A: Yes. It’s one of the few remaining advantages for nimble companies. LLMs cite based on relevance and recency. Not domain authority or ad spend. I’ve seen Series A companies outrank Fortune 500s in AI search results. They published specific, cited insights while the enterprise was still routing content through legal review. The citation moat hasn’t been built yet. Enterprise brands are just as blind here as you are. The window is open.

Q: How does citation engineering work for AI discovery?

A: Citation engineering is the systematic process of creating content that LLMs retrieve and reference when answering buyer queries. You identify the questions your prospects ask AI platforms. Then publish insights that directly answer those queries with specific data, frameworks, or proprietary research. Each piece is structured to be citation-worthy. Quotable stats. Named methodologies. Clear POV. When an LLM synthesizes an answer, it pulls from and credits your content as a source. AI models generate 2.9x more queries than users type. 32.9% of all AI citations come exclusively from invisible fan-out queries. One-third of AI citation opportunities are invisible to every keyword tool on the market.

Q: What metrics indicate strong AI discovery performance?

A: Watch your Citation Frequency Score (mentions across core queries). Track Source Attribution Rate (how often you’re credited when mentioned). Monitor Query Coverage (percentage of buyer-intent queries where you appear). We also track Citation Velocity—the month-over-month increase in mentions across platforms. AI Citation Share measures the percentage of buyer queries in a category where an AI platform recommends your company versus competitors. The top-cited company in a category captures 3.7x more inbound leads than the second-place competitor. This comes from AthenaHQ’s analysis of 768,000 citations. If your Citation Velocity is flat or declining, you’re losing ground to competitors who are publishing more systematically.

Q: How

Related Reading

• Ai Search Optimization
• The Two-Stage Citation Funnel: Why 85% of Pages Retrieved by ChatGPT A
• How to Get Recommended by ChatGPT: The Complete Citation Engineering P
• The Velocity Threshold: Why 5 Pieces Per Week Compounds Citation Autho

Frequently Asked Questions

What is AI discovery and how is it different from traditional SEO?

AI discovery is when buyers use conversational search engines like ChatGPT, Perplexity, or Gemini to research solutions, and these platforms recommend vendors based on citations in their training data. Traditional SEO optimizes for ranking on a results page (position 1-10); AI discovery optimizes for being cited in the synthesized answer the AI generates. You’re not competing for click position anymore—you’re competing to be the referenced source when an AI answers a buyer’s question.

How much revenue is predicted to flow through AI discovery by 2028?

$750 billion in US revenue is estimated to flow through AI-mediated discovery moments by 2028, which is larger than the entire digital advertising market. This represents a massive shift in how B2B buyers form their initial consideration sets before ever visiting a company website.

Why do 73% of purchase decisions crystallize during the discovery phase?

During discovery, buyers are actively asking questions about what solutions exist for their problem and comparing options. This is when AI platforms synthesize answers and recommend 3-5 vendors, creating the initial mental shortlist that shapes the entire buying process. If you’re not cited in that first AI response, you’re not considered, regardless of later marketing efforts.

What percentage of B2B buyers now use AI for initial research?

48% of B2B buyers (and 62% in SaaS) now use AI for initial research, according to tracking of 10,000+ B2B decision-makers. This means non-citation in AI platforms is disqualification at the research onset, not just a disadvantage—companies not appearing in AI responses don’t exist in buyers’ consideration sets.

What is citation engineering and why is it different from traditional backlink building?

Citation engineering is the practice of creating content specifically designed to be cited in AI responses by addressing high-signal buyer questions and publishing insights in multiple formats (original articles, data points, named frameworks). Unlike traditional backlink building that focuses on domain authority, citation engineering prioritizes retrieval by AI systems and selection for inclusion in final answers, with measurable results in 30-90 days rather than quarters.

How quickly can companies see results from AI discovery strategies?

Companies using systematic citation engineering typically see their first LLM mentions within 30 days, appear in 3-5 different AI platforms by day 60, and reach 340% increases in citation rates by 90 days. This is significantly faster than traditional SEO, which usually takes 9-12 months to show comparable movement.

What is the 30-Day Freshness Cliff and why does it matter?

The 30-Day Freshness Cliff shows that 76.4% of pages cited by AI models were updated within 30 days, with citation rates dropping 38% after 30 days and collapsing after 90 days regardless of domain authority. This means consistently updating and publishing fresh content is critical for maintaining visibility in AI discovery systems.

Does domain authority matter for AI discovery like it does for traditional SEO?

Domain authority has minimal correlation with AI citation (r=0.18, barely above random noise), meaning you don’t need decades of SEO history to rank in AI responses. Instead, companies with modest domain authority can outrank enterprise competitors by publishing consistently engineered content designed for AI citation capture.

What happens if your company isn’t cited in AI responses?

If your solution isn’t mentioned when buyers ask AI ‘what’s the best tool for X,’ you don’t exist in their consideration set, regardless of your traditional SEO rankings. This represents silent revenue loss—no retargeting pixel, no form fill, no intent signal; just deals lost before your demand generation efforts even begin.

What is the Two-Stage Citation Funnel in AI discovery?

The Two-Stage Citation Funnel separates retrieval (getting your content into the AI model’s context window) from citation selection (being quoted in the final answer). Analysis shows 85% of pages retrieved by ChatGPT are never actually cited in the final response, so citation engineering must optimize for both stages to maximize visibility.