Most popular AI 2021–2026
147.763M
52.986M
38.754M
20.124M
4.906M
2.744M
1.646M
1.554M
1.476M
1.476M
1.010M
815.973K
794.515K
247.622KWatch the AI boom unfold, from early writing assistants to ChatGPT, Gemini, Claude and DeepSeek. Scroll the timeline, choose a year or press play to explore how popular AI websites change places from January 2021 to September 2026.
This is the interactive companion to our “Most Popular AI 2021–2026” video. Open the event cards for memorable launches, interviews, demonstrations and viral moments—from the original DALL·E avocado chair and Toy Jensen to ChatGPT, Sora and Will Smith’s spaghetti. Which AI tool do you use most, and who do you think will lead next?
About the data: The bars show estimated worldwide monthly desktop and mobile website visits. Repeat visits count; these are not unique users, mobile-app usage, API calls or a measure of model capability. We compare 40 selected AI-focused websites and developer platforms, displaying up to 15 at a time. Fourteen are eligible at the opening; this is not a verified census of the world’s largest AI products.
The interactive uses the film’s underlying curves. The 2021 history is modelled, gaps are estimated, and transitions between observations are interpolated for animation. The timeline ends on 12 September 2026; the incomplete September period carries the latest monthly estimate rather than a measured September total. Standalone DALL·E ends with the Labs retirement in May 2025; image generation within ChatGPT is not added to its standalone series.
Source records include public traffic estimates from Semrush, Similarweb, Toolify and AIPURE. Providers can disagree, and the figures are not independently audited analytics. Event dates and source links are included with the cards; some interviews and archival clips were recorded after the event they illustrate.
Explore more interactive histories in Visuals.
Timeline summary and key events
In Jan 2021, the leading AI websites in this dataset were DeepL (147.763M), Grammarly (52.986M), remove.bg (38.754M). By Sep 2026, the leaders were ChatGPT (5.600B), Google Gemini (2.600B), Claude (950.200M).
- — 🎨 Imagine it. Describe it. DALL·E turns written prompts into images, including the now-famous avocado armchair. Source Source 2
- — 🎭 Can you trust your eyes? AI-altered Tom Cruise videos go viral, showing how convincing deepfakes have become. Source Source 2
- — 💻 Your AI pair programmer. GitHub Copilot enters technical preview, suggesting code inside the editor. Source Source 2
- — 🤖 You can run away. Introducing Tesla Bot, Musk jokes that people should be able to outrun or overpower the robot. Source
- — 🤖 Meet my digital double. NVIDIA unveils a talking toy version of Jensen Huang that answers questions in real time. Source Source 2
- — 🎨 From words to worlds. DALL·E 2 introduces more realistic text-to-image generation and image editing. Source Source 2
- — ⚠ AI needs rules. Musk calls for an AI regulator, arguing that public safety justifies oversight. Source
- — 🖼️ Image generation opens up. Stable Diffusion is publicly released, letting creators run the image model themselves. Source
- — 💬 A conversation changes everything. ChatGPT launches as a free research preview. Altman recalls the launch at DevDay 2023. Source Source 2
- — 🚀 The next generation arrives. OpenAI releases GPT-4. On the same day, Anthropic introduces Claude to a wider audience. Source Source 2
- — ⚠ A warning about AI. Musk warns that AI could threaten civilization, while acknowledging uncertainty about the probability. Source
- — ⚠ What could go wrong? At a Senate hearing, Altman warns that the AI industry could cause significant harm to the world. Source
- — 🦙 Open weights, new possibilities. Meta releases Llama 2 for research and commercial use under its community licence. Source Source 2
- — ✨ Meet Gemini. Google introduces Gemini, a family of models built to work across text, images, audio and video. Source Source 2
- — ⚖️ Who owns the training data? The New York Times sues OpenAI and Microsoft, alleging unauthorized use of its journalism to train AI. Source Source 2
- — 🎬 Words become moving pictures. OpenAI previews Sora, demonstrating text-generated videos up to one minute long. Source Source 2
- — 🎙️ AI finds its voice. GPT-4o is unveiled with demonstrations of real-time voice and vision interaction. Source Source 2 Source 3
- — 🧠 More time to think. o1-preview introduces a new model series trained to reason before responding. Source Source 2
- — 🐋 Reasoning gets a new rival. DeepSeek releases R1 and distilled reasoning models, with model weights under the MIT licence. Source Source 2
- — 🛠️ From answers to actions. Claude 4 arrives alongside the general availability of the Claude Code coding agent. Source Source 2
- — 🚀 GPT-5 arrives. OpenAI launches GPT-5, bringing fast answers and deeper reasoning into one system. Source Source 2
- — ✨ A new generation of Gemini. Gemini 3 launches in the Gemini app, adding new reasoning and interactive capabilities. Source Source 2
- — 🧠 Longer tasks. Bigger context. Claude Opus 4.6 launches with a one-million-token context window in beta. Source
- — 💼 AI gets to work. GPT-5.4 launches across ChatGPT, the API and Codex, with a focus on professional tasks. Source Source 2
- — 🍝 Same spaghetti. Different AI. The Will Smith spaghetti test: 2023 vs. a 2026 Kling recreation. Both clips are AI-generated. Source
Sign in or create an account to comment.
0 comments
Be the first to share your perspective.