The AI buyer funnel is the classic marketing funnel (TOFU, MOFU, BOFU) reframed for AI search behavior. Each stage has distinct query patterns, citation behaviors, and AEO levers in AI engines, even though buyers can traverse all three stages inside a single ChatGPT session. The framework is the same as it has been since 1898; the mechanics inside each stage are different in AI search than in Google search. This entry covers the AI-context translation: what TOFU, MOFU, and BOFU look like in answer engines, and how to map content and optimization to each stage.
What is the AI buyer funnel?
The marketing funnel is one of the oldest frameworks in commercial marketing. Elias St. Elmo Lewis introduced the AIDA model (Awareness, Interest, Desire, Action) in 1898, and the modern three-stage abbreviation (TOFU, MOFU, BOFU) is the same arc: buyers move from broad awareness, through narrower consideration, to specific vendor evaluation. The classic funnel still organizes strategy in 2026, but the buyer behavior inside it has shifted as AI search took over an increasing share of B2B research.
The AI buyer funnel is the AI-context application of that framework. The classic stages still apply (TOFU = research, MOFU = comparison, BOFU = evaluation), but each stage has distinct query patterns, citation behaviors, and AEO levers when the buyer is using ChatGPT, Perplexity, Gemini, Copilot, or Claude instead of Google. McKinsey and Demandbase data indicates 40 to 70 percent of B2B buyer research now involves at least one AI touchpoint, concentrated in TOFU and MOFU (discovery and consideration); BOFU still skews toward direct vendor engagement, but AI-assisted shortlisting now precedes most BOFU contacts. The funnel is still the right map. The terrain inside each stage is new.
See AI-assisted research for the underlying B2B buyer behavior that produced this remapping, and AI shortlist for the BOFU-stage artifact this funnel ultimately produces.
The three stages in AI search
Each stage has distinct query types, citation patterns, and AEO levers. The leverage you have on each stage is different.
TOFU - open-ended research queries
Buyers at the top of the funnel ask AI broad research questions. "What is generative engine optimization." "How do I measure AI citations." "Why is my organic traffic dropping." These queries fan out into many sub-questions via query fan-out in engines like Google AI Mode and Perplexity. The synthesized answer pulls from category-level sources: trade publications, broad reviews, foundational explainer content, Wikipedia, and authoritative how-to pages. Brand-specific content is rarely cited at TOFU because the buyer is not asking about a brand.
TOFU citation patterns favor topical authority and content cluster depth over brand specificity. The AEO levers that move TOFU citation rate are: deep topic-cluster coverage of category questions, strong schema markup that lets retrieval extract clean answers, direct-answer paragraphs on category pages, and content freshness on the explainer content that AI engines re-cite as the category evolves.
MOFU - category comparison queries
Buyers in the middle ask AI to compare options. "Best CRM for B2B SaaS." "Alternatives to Salesforce." "Top AI citation tracking tools." These queries pull from third-party comparison content: G2 and Capterra profiles, review aggregators, Best Of listicles, comparison articles in trade publications. First-party brand pages get cited at MOFU when they are the authoritative source on a sub-feature (a pricing page, a security overview, a specific integration), but the majority of MOFU citations are third-party.
MOFU citation patterns favor brands with presence across multiple comparison sources and consistent positioning across them. The AEO levers that move MOFU citation rate are: claiming and completing review-platform profiles, earning placement in third-party Best Of listicles, structured comparison tables on the brand's own site (cited at MOFU when the brand's product is a candidate), and consistent positioning across all third-party sources so AI engines see editorial consensus on what the brand actually does.
BOFU - brand-specific evaluation queries
Buyers at the bottom ask AI specific brand-evaluation questions. "How much does X cost." "What features does X have." "X vs Y." "Alternatives to X." These queries pull from a mix: first-party brand pages (the pricing page, the feature pages, the comparison pages), high-trust review platforms, and trade-press coverage of the specific brand. The AI shortlist forms at BOFU and determines which 3 to 5 brands the buyer evaluates next.
BOFU citation patterns favor brands with strong entity recognition, canonical brand pages, and first-party transparency on pricing, features, and positioning. The AEO levers that move BOFU citation rate are: clear and current pricing pages, structured feature pages, Organization and Product schema with sameAs links into authoritative listings, and presence in knowledge graphs (Wikidata, Wikipedia where notable) so AI engines treat the brand as a first-class entity rather than a string of characters that might be confused with a similarly named competitor.
How AI search remaps the funnel
The classic funnel assumes buyers traverse stages over weeks or months, visiting different surfaces at each step (a blog post for TOFU, a comparison article for MOFU, a vendor's product page for BOFU). AI search changes the mechanics three structural ways.
Multi-turn conversation depth replaces single-query funnel position
A buyer can traverse TOFU through BOFU within a single ChatGPT session. They ask a broad question, follow up with a comparison, then ask for specific vendor pricing - all in one conversation. Funnel position is no longer tied to a single visit to a single surface. The implication: brands need to be retrievable at all three stages, because the buyer might encounter the brand at any of them within the same session.
Query fan-out hides sub-queries from publishers
When a buyer asks a TOFU question in Google AI Mode or Perplexity, query fan-out decomposes it into 5 to 50 sub-queries the publisher cannot directly see. Some of those sub-queries are TOFU-relevance; some are MOFU-relevance. Content optimized for the literal TOFU question may miss the MOFU sub-queries that determine which competitor brands get surfaced in the same synthesized answer. The implication: TOFU content needs to anticipate the MOFU sub-questions that fan out from it.
AI engines collapse buyer-journey stages into single synthesis
Ask a TOFU question, and the AI may already include a BOFU shortlist in the answer. "What is the best CRM for small teams" is structurally TOFU phrasing, but Perplexity and ChatGPT routinely respond with a 3-5 brand shortlist plus a brief category-level definition. The implication: brands cannot afford to be absent at any single stage, because the AI may render a single answer that pulls from all three stages of source content simultaneously.
AI buyer funnel vs related concepts
The framework overlaps with several adjacent concepts in the catalog. The cleanest way to keep them straight:
| Concept | What it is | Relation to funnel |
|---|---|---|
| AI buyer funnel | The framework organizing stages | The structural map |
| AI-assisted research | The buyer behavior happening at TOFU + MOFU | The behavior inside the upper stages |
| AI shortlist | The 3-5 brands AI surfaces in a category answer | The BOFU-stage output artifact |
| Context hint | ChatGPT Ads targeting primitive | Maps to specific funnel stages on the paid side |
| Brand query | A search query containing the brand name | Almost always BOFU-stage |
How to map your content to AI funnel stages
Three layers, each with concrete tactics.
Per-stage content coverage
- TOFU: foundational category explainers, glossary entries, methodology posts, broad how-to guides. Optimize for topical authority + clean structure.
- MOFU: Best Of listicles, comparison articles, alternative roundups, sub-feature explainers. Optimize for third-party citation presence + consistent positioning.
- BOFU: pricing page, feature pages, comparison pages, About page, knowledge graph signals (Wikidata, schema with sameAs). Optimize for entity recognition + first-party transparency.
Per-stage measurement
Build three sub-sets inside your AI prompt monitoring set: TOFU queries (open-ended research, 30 to 50% of the set), MOFU queries (category comparison, 30 to 40%), BOFU queries (brand evaluation, 20 to 30%). Track citation rate and citation share separately for each. Most brands win one stage and lose the others, and the per-stage split is what reveals the actual gap. The Answer Engine Insights platform module runs the per-stage split across all five major engines on a weekly cadence.
Per-stage AEO lever prioritization
Different stages reward different optimization work. If TOFU citation share is weak, invest in content cluster depth and freshness. If MOFU is weak, invest in third-party listicle placements and review-platform completeness. If BOFU is weak, invest in knowledge graph presence and first-party page structure. Auditing the per-stage gap with the Quick AEO Audit or the AI Visibility Checker shows which stage to prioritize first.
The full per-stage citation-share methodology lives in our how to measure AI citation share guide, and the 7-step playbook for shifting per-stage share is in how to increase your AI citation share.
Common misconceptions
AI search bypasses the funnel
It does not. The stages still exist; the buyer still moves through them, just often within a single conversation. What changed is the mechanics inside each stage and the ability of a single AI session to traverse stages quickly. Treating AI as a stage-less surface produces undifferentiated content that wins no stage cleanly.
TOFU traffic doesn't matter in AI search because nobody clicks
Clicks are not the only outcome. TOFU content is the retrieval pool AI engines pull from when synthesizing MOFU and BOFU answers. A brand with weak TOFU topical authority loses citation share even on the queries where it most wants to win, because the AI does not have grounding material on the underlying category to anchor the synthesized answer in. The right frame is contribution-to-synthesis, not contribution-to-clicks.
BOFU is the only stage AI engines surface brands
Untrue. Brands surface at all three stages, but in different formats. TOFU surfaces brands as side-mentions or category exemplars ("for example, brands like X have written about this"). MOFU surfaces brands as comparison entries in shortlists. BOFU surfaces brands as the direct answer to evaluation questions. Each stage has its own citation type, and a complete AEO program competes for all three.
Frequently asked questions
#What is the AI buyer funnel in simple terms?
The AI buyer funnel is the classic marketing funnel (TOFU, MOFU, BOFU) reframed for how buyers actually behave when they use AI search engines. TOFU is open-ended research ("what is X," "how do I solve Y"); MOFU is category comparison ("best X for Y," "X alternatives"); BOFU is brand-specific evaluation ("X pricing," "X vs competitor"). Each stage has distinct AI query patterns, citation behaviors, and AEO levers, which means a B2B marketer needs different content and optimization tactics for each stage, even though the buyer might traverse all three within a single ChatGPT session.
#How is the AI buyer funnel different from the classic marketing funnel?
The framework is the same; the buyer behavior inside it changes in three structural ways. First, multi-turn AI conversations let buyers move through TOFU to BOFU in a single session, so funnel position is no longer tied to a single visit. Second, query fan-out means a TOFU question generates MOFU-relevance sub-queries the publisher cannot directly see, but which determine which competitor content gets surfaced. Third, AI engines often collapse buyer-journey stages into a single synthesis: ask a TOFU question and the answer may already include a BOFU shortlist of 3-5 brands. The classic funnel still organizes the strategy; the execution mechanics differ.
#Which AI engines surface brands at which funnel stage?
Stage behavior varies by engine. Perplexity and Google AI Mode are TOFU-heavy because users land there for research questions. Google AI Overviews surface across all three stages but skew MOFU (comparison and category queries are the highest-trigger queries). ChatGPT is the broadest performer across all stages because of its conversational depth and 900M+ weekly users. Microsoft Copilot is BOFU-heavy in enterprise contexts where Microsoft 365 users surface vendor evaluation queries. Track the same prompt set across all five engines to see which surface each brand at each stage.
#Should I still create TOFU content if AI engines collapse stages?
Yes. Stage collapse changes the rendering of the answer, not the underlying source-selection mechanics. AI engines still retrieve TOFU-style category content when they need to ground the broader claims inside a synthesized BOFU response. A brand with deep TOFU topical authority shows up in the citation list even when the user-visible answer surfaces a BOFU shortlist. Skipping TOFU content removes you from the retrieval pool that feeds MOFU and BOFU answers; you would lose share even on the queries you most want to win.
#How do I measure AI funnel performance?
Per-stage measurement is the cleanest approach. Build three sub-sets inside your AI prompt monitoring set: TOFU queries (open-ended research), MOFU queries (category comparison), BOFU queries (brand evaluation). Track citation rate and citation share separately for each. Most brands win one stage and lose the others; the per-stage split is what reveals the actual gap. Weekly cadence across five AI engines through Answer Engine Insights is the operational pattern most B2B teams use.
