The State Of AI: July 2026. The Top 20 Chatbots, Ranked And Explained
Eighteen months ago, one chatbot held 87% of the market. Today five platforms top 100 million users each, two AI labs have filed trillion-dollar IPOs, and China runs a parallel market most Western charts never count. This is the full picture, all 20 players, every number sourced.
Eight Numbers That Define AI In July 2026
- People are on pace to download nearly 2.3 billion AI apps and spend over $4.2 billion on them in the first half of 2026, up from $1.83 billion a year earlier, per Sensor Tower.
- Hours spent in AI apps roughly doubled year over year, from 17.2 billion in H1 2025 to about 36 billion in H1 2026. The top three assistants take 89% of that time.
- ChatGPT’s share of the seven biggest assistants’ web visits fell from 76.5% in February 2025 to 53.9% in May 2026, while Gemini rose from 5.6% to 27.9% and Claude from 1.4% to 9.2% (Similarweb via Momentic).
- Five platforms now clear 100 million+ monthly users: Meta AI, ChatGPT, Gemini, Copilot, and Doubao. Claude, Yuanbao, and Grok sit just behind.
- Anthropic became the world’s most valuable AI company at $965 billion, and both it and OpenAI filed confidential IPOs a week apart in June 2026.
- China’s market runs on its own leaderboard: ByteDance’s Doubao hit 330 million users by May 2026 and added paid tiers, while DeepSeek, Yuanbao, and Qwen round out the domestic top tier.
- Users stack assistants instead of switching: 79% of OpenAI’s paying customers also pay Anthropic, and about 20% of ChatGPT’s weekly web users also use Gemini.
- AI referral traffic converts at 16% versus 1.8% for Google organic (Seer Interactive), and the same brand can get cited 615x more on one chatbot than another (Superlines).
AI Stopped Growing Users And Started Making Money
The defining shift of 2026 isn’t a model release. It’s monetization. Consumer spending on AI apps more than doubled year over year while download growth cooled, average revenue per user climbed industry-wide, and the two biggest private labs raced each other to the public markets. The land-grab phase is ending. The extraction phase has begun.
DataReportal’s Digital 2026 report counts over 1 billion monthly users of standalone AI tools, and that excludes embedded assistants like Meta AI inside WhatsApp, Copilot inside Office, and My AI inside Snapchat. Count those and generative AI reaches well past 2 billion people.
On pace for H1 2026, per Sensor Tower’s State of AI report.
Up from $1.83 billion in H1 2025. Monetization is the 2026 story.
Estimated for H1 2026, roughly double the 17.2 billion of H1 2025.
Anthropic ($965B) plus OpenAI ($852B) on the private market, both IPO-bound.
The Western Web Leaderboard, May 2026
Start with the cleanest apples-to-apples measure available: Similarweb web-visit share across the seven largest standalone assistants. One warning before you quote it anywhere: this panel excludes embedded assistants (Meta AI), character apps (Character.AI), aggregators (Poe), and every major Chinese platform. It’s the Western web race, not the whole war.
Source: Similarweb via Momentic, May 2026. Share of combined web visits across the seven largest general-purpose standalone assistants.
The quarter-by-quarter movement matters as much as the snapshot. From February to May 2026, Claude grew fastest at +228% in web visits, followed by DeepSeek (+58%, rebounding from a February trough to 430 million visits), then Gemini and Copilot. Grok and Perplexity declined, and ChatGPT was roughly flat. And by Sensor Tower’s broader app-based metric, ChatGPT’s global user share fell below 50% for the first time ever in March, landing at 46.4% by May.
The Top 20 AI Chatbots, July 2026
Ranked by best-available monthly user estimates, with one honest caveat: no two platforms count users the same way. Embedded assistants inherit their host app’s audience, Chinese platforms report cumulative or domestic figures, and several players simply don’t disclose. Where a number is contested, the range or metric is noted. Treat the order as directional, and the stories as the point.
Four Games Being Played On One Board
The top 20 isn’t one competition. It’s four different games, and each tier wins on a different resource.
The Embedded Giants
Meta AI, Copilot, My AI, and soon Siri-via-Gemini. They win on distribution: the assistant is simply there, inside apps billions already open daily. Their weakness is intent. Nobody opens WhatsApp to think.
The Destination Apps
ChatGPT, Gemini, Claude, Grok, Perplexity. People choose them on purpose, which makes their usage high-intent and monetizable. This is where the ads, subscriptions, and $1.8 trillion of IPO paper live.
China’s Parallel Market
Doubao, Yuanbao, DeepSeek, Qwen, Kimi, Ernie. Half a billion users Western charts never count, models within a few benchmark points of the frontier, and pricing up to 95% cheaper. The catch: a brutal monetization gap, which Doubao’s new paid tiers are the first real test of.
The Specialists
Character.AI (companionship), GitHub Copilot (code), Le Chat (EU sovereignty), Poe (aggregation), Pi (the cautionary tale). Sensor Tower notes these adjacent categories remain fragmented and wide open, unlike the brutally concentrated assistant race above them.
Who’s Actually Getting Paid
User charts and revenue charts tell opposite stories. Anthropic makes roughly $192 per monthly user; OpenAI makes about $23; China’s entire iOS AI market once generated $500K against ChatGPT’s $1.7 billion. Here’s the scoreboard where it counts.
The multiple math is where it gets spicy. Anthropic trades near 20x revenue; OpenAI’s implied multiple runs 34 to 42x; Perplexity sits around 45 to 50x. If the twin October-window IPOs price anywhere near their filings, public markets will finally referee which model of AI business (enterprise depth, consumer scale, or trust premium) deserves the bigger number.
Nobody Is Switching. Everybody Is Stacking.
The most misread trend in AI is churn. ChatGPT losing share doesn’t mean people quit it. Similarweb cross-panel data shows about 20% of ChatGPT’s weekly web users also use Gemini, and a remarkable 79% of OpenAI’s paying customers also pay Anthropic. The share of ChatGPT users who also use Claude nearly tripled between January and May, while the reverse overlap climbed from 58% to 66%.
People are assembling toolkits: one assistant for search, one for writing or code, one that lives inside their work suite. That’s why total AI app hours doubled to 36 billion even as the leader’s share fell.
Loyalty is real but conditional. Sensor Tower found specific events accelerate switching: OpenAI’s Department of Defense deal in February triggered a measurable spike in uninstalls, and Grok’s content controversies preceded its plateau. In 2026, brand trust moves usage numbers, not just benchmarks.
Search Didn’t Die. It Split Into Twenty Pieces.
Every platform on this list answers questions that used to be Google searches. Google’s own AI Overviews put Gemini answers in front of 2 billion people a month. ChatGPT handles 2.5 billion prompts a day, roughly 30% of Google’s search volume. And the traffic these platforms send out is gold: AI referrals convert at 16% against 1.8% for Google organic, per Seer Interactive, even though they’re still under 1% of most sites’ volume.
The trap is optimizing for one chatbot. Superlines measured a 615x gap in citation volume for the same brand across different platforms in March 2026. A client can dominate ChatGPT answers and not exist in Claude’s. Meanwhile the crawler war decides tomorrow’s answers today: ClaudeBot led AI crawlers at 19.8% of bot traffic in June with Anthropic’s combined footprint near 23%, GPTBot held about 9.4%, and ByteDance’s Bytespider keeps surging in and out of the top five. What these bots read now is what the models recommend in six months.
The strategy that covers all twenty platforms at once is the oldest one in the book: authoritative backlinks and brand mentions on the trusted sites every crawler reads. AI didn’t replace that playbook. It raised the payout.
Five Things To Watch In H2 2026
The Twin IPOs
Anthropic reportedly targets an October Nasdaq listing seeking $60B+; OpenAI’s follows close behind near a $1 trillion goal. Public markets will price the AI business model debate for the first time.
Siri On Gemini, iOS 27
The rebuilt Siri ships late 2026 on a custom 1.2-trillion-parameter Gemini model, putting Google’s AI in front of 2 billion+ Apple devices. The biggest distribution event since ChatGPT launched.
The Ads Race
OpenAI’s ad pilot hit $100M annualized in six weeks and self-serve is open. Google shows ads in AI Overviews. Perplexity refuses on principle. Watch whether users punish the monetizers or shrug.
Grok 5 And The Missing Flagships
Grok 5 (reportedly up to 10 trillion parameters) has missed two launch windows. DeepSeek’s R2 cycle slipped too. Whether the laggards land in H2 decides if the frontier stays a three-horse race.
China’s Monetization Test
Doubao added subscription tiers in May with 330 million users watching. If Chinese consumers start paying for AI at scale, the global revenue map redraws overnight.
Quick Answers On The AI Market In July 2026
What is the most used AI chatbot in 2026?
Depends on the ruler. Meta AI leads raw monthly actives at 1.2 billion because it’s embedded in WhatsApp, Instagram, and Facebook. ChatGPT leads every measure of chosen usage: 1.1 billion monthly actives, 900 million weekly, and 53.9% of standalone-assistant web visits.
Why do market share numbers contradict each other?
Because every panel measures something different. The same quarter can read 76.85% for ChatGPT on Statcounter’s global web panel and 38.7% on Apptopia’s US mobile panel. Always check the source, geography, and metric before quoting any share figure, including ours.
Which AI company makes the most money?
Among the labs, Anthropic: a $47 billion annualized run rate as of May 2026 against OpenAI’s roughly $25 to 30 billion, driven by enterprise API contracts. Count the hyperscalers and Google’s Cloud division alone did $20 billion in a single quarter, up 63%.
Are Chinese AI chatbots really competitive?
On capability, yes: DeepSeek V4 Pro, Qwen3-Max, and Kimi K2 score within a few points of Western frontier models at a fraction of the cost. On revenue, not yet. The user bases are enormous, the willingness to pay is the open question, and Doubao’s new paid tiers are the live experiment.
How should businesses respond to twenty AI platforms?
Don’t chase twenty strategies. All of these systems learn who to recommend from the same open web, so authority built on trusted, crawlable sites (links, brand mentions, structured content) compounds across every platform at once. Then track citations on the two or three platforms your buyers actually use.
Deepan Paul
AuthorDeepan Paul is a SEO Lead with four years of experience helping brands recover, scale, and sustain organic growth across global B2B, B2C, and D2C markets. He is recognized as a ranking revival expert, specializing in diagnosing traffic drops, fixing indexing and technical issues, and restoring lost search visibility. He has managed international clients and led cross-functional teams, aligning SEO strategies with core business goals. His expertise spans technical SEO, content strategy, indexing optimization, and building scalable growth systems that adapt to constant algorithm changes. Beyond execution, Deepan is also an SEO trainer and guest speaker, mentoring professionals and contributing insights to leading digital marketing publications. His approach is focused on sustainable, system-driven SEO that delivers long-term results rather than short-term gains.