AI-Review.net
A market-intelligence dashboard for tracking which AI products are winning — by category, adoption, and enterprise spend. Release markers show when important products entered the market.
Market trackerUpdated Sep 14, 2026
Prototype / source-aware data model
Biggest structural shift
Claude ↑ / OpenAI ↓
In enterprise LLM spend, Anthropic rose from ~12% in 2023 to ~40% in 2025, while OpenAI fell from ~50% to ~27%.
Coding adoption
39%
Claude Code usage at work among professional developers, May–Jul 2026 (JetBrains).
Agent penetration
90%
Professional developers using coding agents at least weekly, May–Jul 2026 (JetBrains).

General AI assistants proxy share

Relative consumer/web usage index. Useful for direction, not audited MAU market share.
Release markers: ChatGPT (Nov 2022), Claude (Mar 2023), Gemini model family (Dec 2023). Series are normalized illustrative trend estimates anchored to public traffic/adoption reports.

AI coding tools survey measured

% of professional developers using each tool at work. Multi-select; lines do not sum to 100%.
JetBrains Developer Ecosystem / AI Pulse surveys. Known points include Jan 2026 and May–Jul 2026. Intermediate points are visually interpolated.

Enterprise foundation-model share spend / usage

Estimated share of enterprise LLM API usage / spend.
Menlo Ventures: OpenAI ~50%→27% (2023→2025), Anthropic ~12%→40%, Google ~7%→21%. Remainder grouped as Other.

Open-weight model ecosystem proxy share

Illustrative share of open-model usage across developer/startup channels.
Enterprise open-weight share is about 11% of total LLM usage in Menlo's 2025 estimate. Family-level trend below is a proxy inspired by vLLM/OpenRouter ecosystem direction.

AI image generation traffic / usage proxy

Standalone + embedded product share is difficult to measure; this chart is intentionally labeled as a proxy.
Embedded image generation inside ChatGPT and Gemini makes standalone-web market-share comparisons incomplete.

AI video generation traffic / usage proxy

Relative product adoption proxy across major video-generation platforms.
Veo and Sora distribution is partly embedded inside broader ecosystems; Runway/Kling/HeyGen have clearer standalone traffic signals.

Release timeline

Key launches used as context markers. “Release” refers to public product/model availability, not company founding date.
CategoryProduct / familyPublic releaseWhy it matters
AssistantChatGPTNov 30, 2022Consumer generative-AI breakout
AssistantClaudeMar 14, 2023Anthropic general assistant/API
AssistantGeminiDec 6, 2023Google multimodal foundation-model family
CodingGitHub Copilot (GA)Jun 2022Early mass-market AI coding assistant
CodingClaude Code2025Rapidly became leading coding agent by 2026 survey adoption
CodingOpenAI Codex agent2025Agentic coding product from OpenAI
Open modelsLlama familyFeb 2023Major open-weight ecosystem anchor
Open modelsDeepSeek R1Jan 2025High-profile reasoning-model release
ImageMidjourneyJul 2022Early high-quality consumer image generation
ImageStable DiffusionAug 2022Open image-generation ecosystem
VideoSora2024 preview / 2024 productHigh-profile text-to-video system
VideoGoogle VeoMay 2024Google's flagship video-generation family

Data methodology

Enterprise APIs: Menlo Ventures 2025 State of Generative AI in the Enterprise. 2023–2025 values are reported/estimated by Menlo.
Coding: JetBrains Developer Ecosystem Survey 2026 and AI Pulse research. Multi-tool adoption means percentages can exceed 100% when summed.
Assistants: Public consumer-adoption and web-traffic reports. Because platforms report MAU differently, this demo uses a normalized share proxy.
Image / Video: Public traffic and ecosystem signals. Embedded generation makes a universal denominator unavailable; values are illustrative until a consistent dataset is licensed or collected.
Release dates: Official vendor announcements where available. Dates should be stored independently from chart observations in production.
Production recommendation: Store each observation as {date, category, product, metric, value, denominator, geography, source_url, confidence} and render from an API rather than hard-coded arrays.