Original · Unique Research / 非凡产研 · 2026-07-20 · Chinese source: https://view.inews.qq.com/a/20260720A0ADIY00
Editor's note: This is a complete English rendition of the source article. Product rankings, traffic figures, and company claims are retained as the source's own reporting. Source images are not processed per task scope.
---
AI Industry Observation
"The headline numbers already say it clearly: 63 are declining, but the total is still rising. Users' bookmarks and companies' procurement lists can't hold this many products."
Nearly $400 million in funding, valuation exceeding 10 billion RMB—on July 20, the same day WAIC 2026 closed in Shanghai, Meshy, the AI 3D modeling company, officially announced this Series B round: the largest in AI 3D to date, setting records simultaneously for single-round size and valuation.
The conference itself was lively enough. Yin Qi (印奇), Chairman of StepFun (阶跃星辰), declared at the opening: model capabilities are crossing a critical threshold in 2026, moving from tasks lasting only a few seconds to independently working for dozens of hours; agents are transitioning from chat tools to "the smallest unit of productivity."
Another detail worth singling out: WAIC established five new ecosystems this year, one called AI GRAVITY, dedicated to international partnership matching, with a focus on connecting Asia-Pacific and Europe. "Going overseas" is no longer just a single-point attempt by a few startups—it's an industry-level topic backed by dedicated ecosystems.
What does this have to do with June's data?
The connection is direct. Our analysis of the June overseas AI application rankings (measurement period: June 2026, 100 products) shows: the overall market is cooling, but traffic and capital have not retreated—they're just changing direction, moving closer to the real world.
Start with the most eye-catching stat: of 100 products, 63 saw visits decline month-over-month, 36 rose, and 1 was newly added with no MoM data.
By conventional logic, with 60% of products bleeding, the overall picture should be ugly. But when you sum the visits across all 100 products: May was approximately 488 million, June was 492 million—up 0.9% MoM.
Declines are the majority; the aggregate rises. Where did the traffic go? Look at concentration: the top 5 took 33% of all traffic; the top 10 took 47%. Moreover, the top 10's combined visits rose 4.3% MoM—more than four times the overall growth rate.
"Users haven't left AI; they're just no longer spreading it evenly. The market isn't upgrading as a whole—it's splitting. Winners eat the share vacated by losers; the middle ground is squeezed empty from both sides. Cooling doesn't mean the tide is going out. What's receding is the phase where 'everyone gets a piece.'"
The top spot is still the familiar PolyBuzz, at 65.39 million, +13.2% MoM, holding #1. But the top has little new story; the real June story is in the mid-tier.
Across the entire ranking, only one sector managed universal growth: 3D modeling. Three products combined reached 19.3 million, +15.5% MoM.
The old leader Meshy did 9.51 million, +6%—modest growth on the surface. But pull the monthly curve longer: April was ~6.09 million, May 8.97 million (+47%), June 9.51 million (+6%)—56% growth in two months, but the pace is clearly narrowing. Meshy is entering a plateau.
What's interesting is that this plateau collides with nearly $400 million in funding. This round, IDG Capital, Matrix Partners China, and Monolith (砺思资本) are new entrants; existing shareholders including Granite Asia, Sequoia China, BAI Capital, and Source Code Capital oversubscribed. Founder Hu Yuanming (胡渊鸣) has a Tsinghua Yao Class undergrad and MIT PhD; during his PhD he wrote the open-source graphics library "Taichi" (太极). Commercialization numbers are even more impressive: ARR grew 12x in the past year, registered users exceeded 12 million, cumulative generated models surpassed 100 million, and customers cover half of the world's top-10 tech companies by market cap—from NetEase Games and 37 Interactive to Bambu Lab and Hugo Boss. Capital is betting at this point clearly seeing beyond traffic growth.
The real explosion was Rodin (Hypershell Technology / 影眸科技, hyper3d.ai): visits +134.6% to 3.04 million, active users +112%. The surge has a concrete reason—in late June, Hypershell announced a new round of funding worth hundreds of millions of RMB (led by Cathay Capital and Shanghai Guotou Xiandao; previously backed by ByteDance, Meituan Longzhu, Sequoia, and BlueRun Ventures), and unveiled the new model Rodin Gen-2.5: the world's first 10-million-polygon-level 3D generation model, bringing the LLM's "think first, generate later" reasoning logic into 3D for the first time, generating a million-polygon model in as fast as 4 seconds. Official disclosure: in its first month, subscriber count and ARR both grew over 400% MoM, with overseas revenue at 70-80%. At this year's CES, Jensen Huang's keynote 3D generation segment used their model. One new model directly pulled a product line's curve back up.
Tripo (under VAST) did 6.75 million, +4.8%. VAST completed nearly $200 million in funding in June at a ~$1 billion valuation, then a sector record—only to have Meshy double that record a month later.
"Why 3D specifically? Our judgment: 3D generation is shifting from a 'content category' to 'infrastructure.' Games need 3D assets, e-commerce needs product models, 3D printing needs directly printable files, robotics needs massive spatial data for simulation training. Image generation solves 'looks right'; 3D generation solves 'can it go directly into the production process?' Traffic and capital voting for it simultaneously isn't wrong."
The second warming direction is emotional companionship: seven products combined reached 117 million visits, accounting for 23.7% of the top 100's total traffic, with the sector up 13.5% MoM. This is the largest and steadiest category in the ranking.
PolyBuzz at 65.39 million is roughly double the second-place Qwen (34.44 million)—note this is cross-sector comparison. Within the sector, the hottest is Emochi (under FlowGPT), +58% to 19.42 million, breaking into the overall #5; HiWaifu +13.1%; Joyland rebounded from May's -26% to +15.6%.
"This sector has no product launches, no grand narratives, but demand is stable and sticky. Users open these products not to complete tasks but to 'have someone there'—this kind of retention is naturally easier than tool products. Companionship isn't a fake demand; it's repeat purchase."
But Joyland's rebound is a reminder: top-tier dominance is real, but the list of who dominates is itself quietly changing.
The agent sector's June storyline, on the surface, was still divergence: eight products combined reached 46.76 million, -1.0% MoM—not collapsing, but broad-based growth is sparse.
On the up side: Z.ai (Zhipu) visits 16.52 million, +36.9%, active users +69.4%; Skywork (Kunlun Wanwei) visits +11.4%, active users +25%; Dola AI (ByteDance) visits 9.63 million, +31%, entering through the specific scenario of scheduling assistant. On the down side: Manus 23.22 million visits, -17.2%, active users -17.3%; MuleRun -26%.
But Manus has a contrast worth pondering: visits fell 17%, yet its 7.41 million active users are still #1 across the entire ranking—higher than PolyBuzz's 4.31 million and Qwen's 5.53 million. Traffic is falling; stickiness remains.
"Users never need 'an Agent.' They need concrete things: booking flights, organizing materials, doing research, managing schedules. 'Agent' is the implementation, not the demand itself. Products that write 'I'm a general-purpose agent' on their homepage force users to figure out what to do with it; Dola enters through scheduling, Skywork focuses on document and research workstations—users instantly know what they're there for."
At WAIC, one vendor shared a similar playbook: hand "understanding intent"—ambiguous work—to the Agent, and hand "deterministic execution" to traditional workflows, thereby lowering token cost per call. For price-sensitive overseas markets, "too expensive per run" directly determines whether an Agent product can land.
Coding tools are also generational: open-source opencode took 10.87 million visits, ranking #11; ByteDance's TRAE fell -25.5%.
A few scattered but noteworthy items:
ByteDance's Dreamina, image generation, +23.5% to 12.3 million, active users +41%. Note: its growth will come back later as important evidence.
Fish Audio (text-to-speech) +15.5%, Pollo.ai +19%, PLAUD (AI recording cards) +15.8%. Voice, video, hardware—someone is growing in all of them.
"See the pattern? The warming list has one common thread: it's all 'generation'—generating 3D, generating images, generating sound, generating video. The declining list coming up is mostly doing 'organizing' and 'editing.' This contrast is the most valuable clue in the June rankings."
Start with a set of easily misread data: the video generation sector has 19 products, combined June visits of 61.23 million, down only 0.8%—basically flat. Flat doesn't mean stagnant—internal rankings are being rearranged: Dreamina active users 5.88 million (+41%), overtaking KLING's 5.22 million to become the sector's MAU #1; KLING's visit growth has narrowed to 1.19%, entering a plateau; HeyGen -8.9%, Pika turned negative at -9.6%.
What's actually contracting is the neighboring video editing: six products combined reached 16.19 million, down 11.8%. Media.io -22%, UniConverter -26%, Filmora -15%—Wondershare's video trio is all declining.
"Put the two sets together and the picture is clear: users aren't making fewer videos; they just don't need standalone post-production tools anymore. When the generation model itself can move the camera, extend clips, dub, and add subtitles, the workflow of moving assets between multiple tools is being directly eaten by the generation step."
Now the decliners. The worst hit is education: three products combined 6.72 million, -29.1% MoM. Gauth (ByteDance) -39.9% to 3.99 million, active users -39.7%; Solvely -47%; Mathos AI -23%.
It looks scary, but fairness requires saying: North American schools entered summer break in late June; photo-search-homework demand naturally cliffs—that's seasonality, not demand disappearing.
The real test is the September back-to-school season. Whether Gauth and its peers can bring users back determines whether education AI is cyclical fluctuation or structural decline.
Smart search has no seasonal excuse: the sector combined 15.44 million, -21.1% MoM. Genspark -25.1%, active users -36.5%; Alibaba's Accio -12.5%.
AI search's original logic was to redo the search engine with large models—give answers directly instead of links. But the problem is, when ChatGPT, Gemini, Qwen, and various Agents all come with built-in web access and information organization, "AI search" itself becomes a feature easily taken over by platforms. The disrupters are being disrupted a second time.
Look at Gensark's two-month curve connected: per our May ranking, it already fell 24% in May visits and 33% in active users; cumulative over two months, visits are down about 43% from April, active users down nearly 58%. This is no longer single-month fluctuation.
"In one sentence: a real demand doesn't mean it can support an independent product. Search demand has never been stronger, but the 'independent AI search' product form is being nestedly absorbed."
The most worth re-examining thing in the June ranking is a set of "collective casualties":
PPT generation trio all fell—PPT.AI -10%, AiPPT -19.9%, Dokie -27%, sector combined -18.9%. Mind maps all fell—Xmind -14.3%, Mapify -20.2%, GitMind -16%, sector combined -16.3%. Image editing mostly declined—Fotor -13.5%, Airbrush -14.3%, insMind -19.2%, sector combined -6.0%. Music generation didn't escape either: Mureka -6%, Udio -8.6%. But music is another story—Suno's relentless pressure, combined with copyright litigation shadows (Udio was previously mired in disputes with the three major labels), is an industry-wide disturbance; set it aside for now.
Single-category declines always have various explanations; summer break might contribute a few points. But summer can't explain the mirror divergence within the same sector: generation up, editing down. PPT, mind maps, editing, and photo retouching collectively bowed in the same month; the explanation likely has only one: Agents are consuming their share.
Remember Dreamina? In the same image sector, editing tools fell double digits while generation tools rose 23.5%. Users aren't making images less—they're not retouching them; they're generating directly. Video is the same: editing fell 11.8%, generation flat.
"Users don't want to open five separate tools to make PPTs, mind maps, and images. They want to say one sentence to an Agent and get the result directly. This isn't one company's loss—it's the 'single-point AI tool' product form becoming obsolete."
Of course, self-calibration is needed: robots working on the WAIC show floor and Agents operating dozens of steps in a row shouldn't be mechanically applied to web traffic data—embodied intelligence and consumer AI apps are two different sectors with completely different metrics. What June data can say only goes as far as "web-side single-point tools are being integrated."
There are two outliers in the ranking worth noting separately.
MaxAI.me: visits rose nearly 133x MoM to 2.41 million; active users rose over 224x. This magnitude is basically impossible to be organic growth—it's likely a measurement methodology change, domain switch, or traffic diversion anomaly. The first reaction to such a number should be to check the methodology, not write a funding announcement.
The other is Agnes AI: visits +570% to 2 million.
Agnes is a product under Singapore's Sapiens AI, founded by Bruce Yang. On June 1, it announced indefinite free API access to its three core models for text, image, and video. The effect was immediate: first-week text model calls exceeded 1 trillion tokens, image generation exceeded 2 million images, video exceeded 2 million seconds; second-week multimodal weekly calls hit 3.12 trillion tokens, even surpassing Claude Opus 4.7's weekly call volume on OpenRouter.
The audacity to burn this much has a basis: its text, image, and video models have all ranked on Claw-Eval and Artificial Analysis blind leaderboards, globally ranked #9 among AI labs, with 10 million registered users.
"Put simply, this is a 'price butcher'-style breakthrough—trading free for developer ecosystem and user mindshare. The +570% visit surge is the most direct surface reaction to this free storm."
One line from Meshy's funding announcement stuck with me: "What used to require a professional art team, expensive software, and two weeks of work can now be done in one minute and one dollar."
Two weeks to one minute. That's why the 3D sector rose across the board and capital broke records twice in a month. And the collective decline of PPT, mind maps, and editing says the flip side of the same thing: the second half of overseas AI isn't about how long your feature list is, but how close you are to the real world.
The headline numbers already say it clearly: 63 are declining, but the total is still rising. Users' bookmarks and companies' procurement lists can't hold this many products; the market's window for storytelling is shrinking. Those who remain at the table next will either be close enough to the real world, or have made one specific thing indispensable.
"June is just a snapshot. Whether education AI returns in September's back-to-school season, whether opencode holds its position, how long Agnes's free storm lasts—next month's ranking, we'll talk again."
---
Unique Research (非凡产研) — Global AI Market Insight Authority
The detailed May 2026 AI Overseas Web rankings published by Unique Research are as follows:
AI Overseas Website Traffic Ranking: TOP 100 list of overseas AI Web products by visits and monthly active users.
All rankings also provide growth-rate leaderboards to help investment institutions discover promising projects and entrepreneurs/developers find potential opportunities.
Unique Research's data and insights have earned widespread trust and citation from top domestic and international investment institutions, academic bodies, and media, becoming an important benchmark for market analysis, investment decisions, and trend judgment. Our partners and citers include (but are not limited to):
Securities & investment banks: Sinolink Securities, China Securities, Huayuan Securities, Haitong Securities, etc.
Authoritative media: China Securities Journal, TMTPost, Jiemian News, etc.
We firmly believe that only data that withstands validation by core industry participants has long-term value.
We believe in the power of open source and openness. Unique Research hereby makes three commitments:
1. This AI product ranking is permanently open-source and free; the complete methodology and raw data support full reproducibility.
2. Data sources for this AI product ranking are strictly limited to independent third-party monitoring platforms; self-submitted corporate data is not accepted as an evaluation basis.
3. This AI product ranking system remains absolutely neutral; no commercial cooperation or paid activity that interferes with ranking results is accepted.
If you are an investor/analyst: This ranking provides a clean, reverse-engineering-verifiable data baseline—a reliable foundation for building investment models, discovering underwater projects, and cross-validating.
If you are an AI entrepreneur/product manager: You can precisely identify competitors and deeply reproduce our methodology to analyze your own product's market performance and optimize iteration.
If you are a researcher/student: The fully open-source nature makes this report an excellent free resource for academic research and market analysis. We encourage innovative research based on this data.
For the complete data behind the charts and rankings shown, please visit 100aiapps.cn. All visible data can be copied and is freely available for non-commercial use such as personal study and research. For commercial use, please consult Unique Research for authorization in advance. We welcome listed teams to apply rankings to business plans, recruitment materials, and other scenarios under compliant usage, providing authoritative support for your capabilities.
Unique Research not only focuses on AI industry research but also aims to build an open-source community promoting transparency in AI market data. We invite developers, data scientists, and industry analysts:
Reproduce & verify: Visit our Feishu knowledge base for complete data and methodology; verify our results and suggest improvements.
Co-build: Contribute code to help us expand data dimensions (e.g., API call volume estimation, agent monitoring models).
Exchange insights: Join our Feishu group to share unique insights derived from Unique Research data.
One person's insight is limited; a group's wisdom is boundless. Let's define the metrics for the AI market together.
---
Based on its independent research of the AI industry and continuous tracking of thousands of AI companies and nearly 10,000 AI products worldwide, Unique Research ranks the top 100 AI WEB products by Chinese or ethnic-Chinese entrepreneurs targeting overseas markets, sorted by cumulative monthly visits. The specific ranking is as follows:
Based on its independent research of the AI industry and continuous tracking of thousands of AI companies and nearly 10,000 AI products worldwide, Unique Research ranks the top 100 AI WEB products by Chinese or ethnic-Chinese entrepreneurs targeting overseas markets, sorted by monthly active users. The specific ranking is as follows:
AI product revenue data in this report, article, ranking, or chart is based on real-time tracking of redirect traffic from AI product websites to payment gateways (e.g., Stripe), combined with time-series analysis, deeply examining historical visits, monthly active user fluctuations, and pricing strategy changes to build a dynamic revenue estimation model. Core model parameters are determined through a triple verification system: paid conversion funnel monitoring, reverse calibration of company-disclosed data, and industry benchmark conversion rate comparison, with timely corrections against company financials, funding announcements, and authoritative media disclosures. Sampling verification shows that the mean absolute error between predicted and actual values is stable within ±10%, meeting industry research-level accuracy standards.
AI product traffic data in this report, article, ranking, or chart is based on multi-source data integration and Unique Research's intelligent algorithm processing. Monitoring scope is strictly limited to users' direct access through official apps and websites. Specifically, this data only covers direct access to AI products through websites and native apps, and does not include traffic from: browser plugins/extensions, desktop client software, WeChat/Alipay mini-program ecosystems, embedded services on third-party platforms such as Discord, local deployments of open-source models, API interface calls, and other non-direct access scenarios. This data focuses on core endpoint access behavior of AI products, aiming to objectively reflect direct usage on mainstream user terminals.
Given the complexity and dynamism of data collection and processing, and the evolving market environment, the displayed data may contain some degree of error and omission. Therefore, this data should be viewed as a reference for research and analysis, providing a window into the AI product market, rather than a definitive basis for formulating specific investment strategies or providing advisory recommendations.
Data coverage expansion notes:
We are actively developing and testing monitoring models for the following emerging areas, expected to be gradually incorporated in future quarterly reports:
AI Agent ecosystem
AI hardware device sales and activity
Large model API call scale estimation
Stay tuned.
Geographic dimensions
Overseas AI products: Products founded by non-Chinese entrepreneurs or teams, primarily launched for global markets (including but not limited to their home market).
Domestic AI products: Products founded by Chinese-national entrepreneurs or teams, primarily launched for the Chinese domestic market.
Going-global AI products: Products founded by Chinese or ethnic-Chinese entrepreneurs or teams, but with primary market targets positioned in overseas markets (non-Chinese domestic market).
Functional dimensions
AI-native applications: Applications that deeply integrate AI technology and algorithms from product design inception, whose core value, business processes, or user experience are entirely built on AI. These applications don't merely use AI to optimize or enhance existing features; AI is a core component—without AI, these applications would not exist or would lose their core value.
AI-feature applications: Applications built on existing business logic and application frameworks, enhancing existing features or providing new ones through integrating or embedding AI technology. These applications may have already had mature business models and market positioning, but introduce AI to significantly improve efficiency, user experience, or create new value points.
AI ecosystem applications: Platform-type applications that support, connect, or facilitate exchange within the AI ecosystem of AI technology developers, application developers, service providers, and end users. These applications extend beyond the technical layer into market, community, resource sharing, and other dimensions, aiming to promote AI technology adoption, innovation, and collaboration.
WEB data metrics
Visits: The total number of all visits to a website during a specific period (e.g., a month), measuring website traffic scale. Page views within a continuous active period are counted in the same session, while a user reactivating after a 30+ minute interval or starting a new day is counted as a new visit, comprehensively and accurately measuring website traffic and visitor activity.
Unique Visitors: The number of independent IP addresses visiting the target website during a specific period (e.g., a month). If a person visits a website on multiple days in a month, they are counted as only one unique visitor.
ARR (Annual Recurring Revenue): The total revenue a website generates from subscription services during a period (e.g., a year), excluding advertising revenue, transaction commissions, and one-time revenue such as professional services.
APP data metrics
Downloads: The number of times an App is downloaded by users during a specific period (e.g., a month), typically measured as downloads per App on app stores (e.g., App Store, Google Play). Downloads are an important metric for measuring App popularity and user appeal, and can reflect the effectiveness of App promotion and marketing campaigns.
Active Users: The number of independent users who performed at least one activity during a specific period (e.g., a month). Active users measure user engagement and activity, typically used to assess App user scale and user activeness.
IAP (In-App Purchase): Revenue generated within an App from users purchasing virtual goods, additional services, or premium features. This excludes advertising revenue, direct user payments (e.g., tips), and revenue from third-party Android app stores.
This report is published by Unique Research. Copyright belongs to Unique Research. Any Chinese reprint or citation must credit the report source; international institutions wishing to reprint or cite must contact for authorization in advance. When citing data, please label: [Data source: Unique Research]. If directly citing charts, please label: [Data source: Unique Research, chart design: Unique Research].
This report is independent original analysis by Unique Research as a third-party institution. Report content does not represent any enterprise's position and does not constitute investment advice to anyone. Investors should note that any investment decisions made based on this report are unrelated to Unique Research, its employees, or its affiliated institutions.
Where legally permissible, Unique Research and its affiliated institutions may hold equity in companies mentioned in the report, or provide or seek to provide fundraising or financial advisory services for them; its employees may serve as directors of companies mentioned in the report.