AI-native search and the post-Google information access market
September 13 at 20:38 · $0.120 total
AI-Native Search & Post-Google Information Access: Market Map
1. Thesis
The search market is being unbundled from a single ad-funded index into intent-specific answer engines, where value migrates from ranking links to owning the last mile of reasoning — and the durable moats shift from crawl scale to proprietary data access, distribution defaults, and workflow integration.
---
2. Segments
A. Consumer Answer Engines
Companies: Perplexity, OpenAI (ChatGPT Search), Google (AI Overviews / Gemini), Microsoft (Copilot/Bing), You.com (pivoted toward enterprise/agents)
Dynamics: Brutal distribution war. Incumbents defend defaults (Chrome, iOS deal) while challengers burn capital on retention; monetization is unproven since AI answers collapse the ad-impression surface that funded search for 25 years.
B. Enterprise / Internal Search (RAG-as-product)
Companies: Glean, Hebbia, Elastic, Coveo, Sinequa (now part of ChapsVision), Microsoft (Copilot for M365)
Dynamics: Highest willingness-to-pay in the map. Winners are defined by connector breadth and permissions-aware retrieval, not model quality. Microsoft's bundling is the existential threat to every independent here.
C. Vertical & Professional Answer Engines
Companies: Harvey (legal), OpenEvidence (medical), Consensus (scientific literature), AlphaSense (financial/market intel), Casetext/CoCounsel (acquired by Thomson Reuters)
Dynamics: Trust and liability create defensibility — users pay for citable, domain-validated answers. Proprietary or licensed corpora (case law, filings, clinical literature) are the moat generalist LLMs can't crawl.
D. Search Infrastructure & Retrieval Plumbing
Companies: Pinecone, Weaviate, Exa, Tavily, Algolia, Vespa (spun out of Yahoo/Verizon Media)
Dynamics: Picks-and-shovels for agentic retrieval (search APIs built for LLM consumption, not humans). Growing fast but commoditizing fast — vector search is now a feature inside Postgres, Elastic, and every cloud.
E. Agentic / Action-Oriented Search
Companies: OpenAI (Operator/Deep Research), Google (Project Mariner), Anthropic (Claude with computer use), Perplexity (shopping/agents); less sure: MultiOn's current status post-pivot (rebranded, reportedly toward "Please" — verify)
Dynamics: Earliest-stage segment; search collapses into task completion ("book it" vs. "find it"). Currently a feature of foundation labs rather than a standalone market — independents get absorbed or squeezed.
---
3. Where Value Accrues: Segment C (Vertical/Professional)
Consumer search (A) has the largest TAM but the worst structure: monetization is unsolved and foundation labs subsidize it as a loss leader. Enterprise (B) is attractive but Microsoft taxes it.
*Vertical answer engines win the most capturable value* because:
- They monetize per-seat at $500–$1,500+/user/year against professional labor costs, not ad CPMs
- Data moats (licensed corpora, feedback loops from experts) resist frontier-model commoditization
- Liability and accuracy requirements block generalist substitutes — a hallucinated citation is a fired lawyer
- Evidence: Harvey's rapid ARR growth and multi-billion valuation, AlphaSense's ~$4B valuation, Thomson Reuters paying ~$650M for Casetext
4. White-Space Opportunities
1. Content-licensing exchange / attribution clearinghouse. There's no neutral marketplace where publishers price, meter, and get paid for AI retrieval of their content per-query (current state: bilateral OpenAI/News Corp-style deals plus lawsuits; TollBit and ProRata are adjacent but nascent and narrow — closest to this space, worth watching, but neither is yet a true programmatic exchange). Whoever builds the "ad exchange for AI answers" owns the economic rail of the entire market.
2. Deep-web / authenticated-data search for agents. No player offers permissioned, credentialed retrieval across the ~90% of valuable data behind logins — government portals, B2B databases, ERP systems, paywalled records — with consent, audit, and compliance built in. Enterprise search (Glean) stops at corporate connectors; consumer agents stop at the paywall. An "OAuth-native search layer" for agentic access to gated data is unclaimed.
---
Confidence notes: Company categorizations and cited deal figures (Casetext ~$650M, AlphaSense valuation) reflect widely reported data through early 2025; private-company valuations and MultiOn's status should be re-verified as this market moves quarterly.
Thesis: The post-Google information access market is being restructured from a single ad-funded index-and-rank monopoly into a layered stack of answer engines, permissioned retrieval agents, vertical research tools, and agentic action layers, where value migrates from the raw index to whoever owns user intent, permissions, and the transaction.
Segments:
- Consumer AI answer engines Companies: Perplexity, OpenAI ChatGPT search, Google Gemini/AI Overviews, Microsoft Copilot/Bing, You.com, Kagi Dynamics: High engagement but low switching costs; monetization is split between subscriptions and ads, while Google defends its core search business and risks ad cannibalization.
- Enterprise knowledge retrieval & agents Companies: Glean, Microsoft 365 Copilot, Google Vertex AI Search, OpenAI ChatGPT Enterprise, Elastic, Vectara Dynamics: Value comes from secure, permission-aware retrieval over internal data and turning search into workflow actions; differentiation is driven by connectors, auditability, and answer accuracy.
- Vertical AI research/answer engines Companies: Harvey (legal), Elicit (academic), AlphaSense (market intelligence), Hebbia (finance/legal), Sourcegraph (code), Consensus (scientific) Dynamics: Domain-specific citations, proprietary data access, and workflow integration create high willingness to pay and retention; each vertical likely consolidates to one or two leaders.
- Agentic browsing & task execution Companies: OpenAI Operator, Anthropic Computer Use, Google Project Mariner (less sure), Perplexity Comet (less sure), Browser Use Dynamics: This layer moves from retrieving information to completing tasks in the browser; security, permissioning, and platform risk are the main bottlenecks.
- Retrieval/search infrastructure & APIs Companies: Exa, Tavily, Brave Search API, SerpAPI, Algolia, Vectara Dynamics: Provides clean, real-time, semantic search and grounding for AI applications; usage-based pricing and low switching costs make it high-volume but prone to commoditization by foundation model providers.
Which segment wins the most value: Enterprise knowledge retrieval/agents and vertical AI answer engines capture the most value because they monetize per seat and per workflow, not per query, and they own permissioned data, auditability, and the action layer. Consumer answer engines may win distribution but face ad margin pressure and low switching costs; infrastructure is essential but low-margin.
White-space opportunities:
- AI answer verification and provenance/trust layer An independent, third-party service that audits AI-generated answers for timeliness, source reliability, conflicts, and licensing across web, enterprise, and vertical sources. No current player covers this end-to-end for AI-native search.
- Personal memory and permissioned personal retrieval graph A user-owned index of everything a person has read, written, searched, and bought across apps, combined with agentic retrieval and fine-grained permissions. Current tools like Limitless or Rewind are partial, but no one owns the full personal-to-web memory layer.