The fourth edition of the Gen AI 100 list represents a comprehensive, data-driven approach to tracking the consumer AI landscape. As Olivia Moore explains, the team tracks this ecosystem through daily engagement with startups, monitoring viral products on Twitter, and identifying AI-powered solutions that might not even market themselves as AI but are reaching mainstream consumers.

"This is one of my favorite reports that we put together a couple times a year. We track the consumer AI landscape through what we do every day, which is like meeting with consumer AI startups that come to pitch us, seeing what goes viral on Twitter, but actually there's a whole separate set of companies and products that might be reaching the true mainstream consumer that might not even be marketing themselves as AI products."

The methodology is entirely data-based, comprising two distinct lists: the top 50 AI products on web and the top 50 on mobile. For web rankings, they use Similar Web data to identify websites with the most monthly visits in January 2025, selecting the first 50 that are GenAI-first products. For mobile, they use Sensor Tower data to rank apps by monthly active users. Additionally, for the first time, they've analyzed the top 50 mobile apps by revenue—revealing surprisingly little overlap between popularity and monetization success.

Timestamp: [00:36-02:30]

The AI ecosystem has experienced several key moments that transformed consumer awareness and adoption patterns. Olivia Moore highlights that consumer activity typically lags behind research developments by 6-12 months, creating an interesting timeline of breakthrough moments.

Pre-ChatGPT breakthroughs included Midjourney and Character AI, which emerged in summer and fall 2022, capturing early adopter communities before the mainstream ChatGPT explosion. After ChatGPT, several products brought AI into widespread consumer consciousness.

"Snapchat's My AI with that little bot that appeared at the very top of your feed and like 150 million people used it. And for a lot of kind of younger consumers that was actually probably their first real chance having a conversation with an LLM."

In the image domain, the viral "Balenciaga Pope" moment in spring 2023 demonstrated AI's convincing creative capabilities to the general public. For AI music, the "BBL Drizzy" song in spring 2024 became a cultural phenomenon, while KO's AI-generated Christmas advertisement in late 2024 signaled AI's shift into enterprise creative consciousness.

Most recently, DeepSeek's launch challenged the assumption that no horizontal model could achieve rapid mass consumer scale after ChatGPT. Its free-to-use reasoning model and real-time display of thought processes captivated users and established a new standard for model transparency.

Timestamp: [02:38-05:20]

Each pivotal moment in AI's development has shattered previously held assumptions about the technology's limitations. From believing AI couldn't create convincing images to thinking ChatGPT had cornered the conversational AI market, these assumptions are consistently being overturned at an accelerating pace.

"I feel like you could actually match each pivotal moment with an assumption. Like an assumption being 'oh well like AI could never trick me into thinking a picture is real when it's not' right, or 'I would never actually listen to a top 100 song that's generated by AI,' or 'Chat GPT has cornered the market, no one else can penetrate it.'"

Despite the rapid progress, Anish Acharya believes we're still in the early adopter phase of AI, particularly when examining specific modalities. While LLMs seemed to be a solved problem, DeepSeek demonstrated otherwise. AI video currently excels at short clips but remains limited for longer-form content—suggesting we're still at the infrastructure-building stage of development.

Two particularly interesting assumptions that may prove incorrect:

1. The belief that AI will excel at transactional tasks while humans maintain the domain of relationship-building. In reality, AI assistants are demonstrating more patience, nuance, and consistency than humans in many conversational contexts.

2. The assumption that humans will delegate work to AI. Anish posits that the reverse might occur—AI could excel at organizing work while humans find joy and fulfillment in executing it.

"So many of these assumptions, and that's why they're assumptions, they seem intuitively correct, are going to turn out to be incorrect."

Timestamp: [05:20-08:20]

The consumer AI ecosystem continues to evolve rapidly, with significant market dynamics suggesting we're still in the early stages of development. The Gen AI 100 report revealed 17 new companies entering the web rankings, demonstrating the ongoing influx of innovation and disruption.

"One of the biggest trends among the newcomers is we are finally on the verge of AI video starting to work."

This emergence of functional AI video capabilities represents an important milestone in the evolution of generative AI technologies. While currently limited to shorter clips of three to six seconds, the technology is steadily improving and following the established pattern of AI drastically reducing creation costs across various media formats.

The ongoing innovation in the AI video space suggests that we're witnessing just the beginning of what might eventually enable the generation of minute-long or even hour-long AI-created video content. This development is particularly significant as it continues the trend of democratizing creative production across different modalities.

Timestamp: [08:20-08:53]

• Consumer AI adoption typically lags 6-12 months behind research and model development
• The Gen AI 100 list uses rigorous data methodology tracking both web visits and mobile active users
• For the first time, the report includes revenue rankings for mobile apps, showing little overlap with popularity metrics
• Key cultural moments accelerated mainstream AI awareness: Midjourney (pre-ChatGPT), Snapchat's My AI (150M users), the Balenciaga Pope image, BBL Drizzy AI music
• DeepSeek's launch proved horizontal models can still achieve mass scale quickly, challenging industry assumptions
• We remain in the early adopter phase of AI, primarily in the infrastructure-building to application-building transition
• Many intuitive assumptions about AI are being proven incorrect, such as AI excelling at transactions but not relationships
• The emergence of functioning AI video represents one of the most significant recent developments
• Market dynamism is evidenced by 17 new companies entering the web rankings in this report
• AI is consistently following a pattern of dramatically reducing creation costs across multiple media formats

Timestamp: [00:00-08:53]

The Gen AI 100 list has experienced significant movement with several newcomers "rewriting the leaderboard overnight." Among these emerging trends, AI video stands out as a category finally coming into its own.

"One of the biggest trends among the newcomers is we are finally on the verge of AI video starting to work, to really work."

Three new video models made the list: Hyo and Clling (both Chinese models) and Sora (OpenAI's long-anticipated model that was announced over a year ago). The upcoming Google model V2 is expected to push capabilities even further in the next 3-6 months, potentially causing another major shakeup in the rankings.

Another significant category of newcomers were "vibe coding" products—tools that dramatically lower barriers to software creation. Cursor, an agentic IDE for technical users, and Bolt, which allows non-technical users to go from text prompts to functioning web apps, both made the list. However, a substantial portion of users for these products are still technically proficient individuals using them for rapid prototyping rather than true novices creating from scratch.

"It sort of follows the trend of AI decreasing the cost of creation in every way and people just trying more ideas. Just think about what that says about the untapped market of people who wanted to build things with code that this is on the top 50 list."

This trend enables two emerging software concepts: "personal software" (designed for an audience of one, which was never economically viable before) and "disposable software" (applications with extremely short lifespans created for specific moments or jokes, similar to how Sunno and Yo democratized quick music creation).

Timestamp: [08:48-11:53]

The fourth edition of the Gen AI 100 introduced a new "Brink List" — highlighting the five companies in both web and mobile categories that just missed the main rankings. This data-driven addition provides valuable insights into market dynamics and emerging players.

"The Brink list is essentially the five companies that almost made the list and were right below the cutoff again purely based on the data. So we pulled the five websites and the five mobile apps."

The Brink List revealed two interesting patterns. First, it captured established players like Runway, Otter, and UMAX that had previously made the top 50 but were edged out by newer entrants like DeepSeek. These companies still maintain massive usage despite temporarily falling off the main rankings, demonstrating the intensely competitive nature of the space.

Second, it identified rising stars like Korea and Lovable that haven't previously appeared in the main rankings but show consistent upward momentum. These companies represent the next potential wave of mainstream consumer AI products if their growth trajectories continue.

The introduction of this list reflects how rapidly the consumer AI landscape is evolving, with companies moving in and out of the spotlight as new innovations emerge and user preferences shift. It provides a window into both established players maintaining relevance and newcomers poised for breakthrough success.

Timestamp: [11:53-12:57]

While the Gen AI 100 list captures current consumer preferences, several anticipated technologies and approaches haven't yet reached mainstream adoption. These gaps highlight the lag between model development and consumer-facing applications.

One surprising absence was the limited presence of style transfer for video generation. As Anish explains:

"One thing I thought we'd see more of is style transfer as an approach to scalable video because style transfer is just a much more tractable problem and has a lot lower cost of inference versus raw text to video."

Despite this expectation, developers have focused primarily on text-to-video approaches, suggesting a preference for maximizing creative flexibility over computational efficiency.

Other notable absences included consumer voice products, screen-aware AI assistants like Gemini Flash that can interact with what's happening on your device, and new agentic interfaces like OpenAI's Operator model that can complete browser-based tasks independently.

"I built something to yell at me like if I go on Netflix or something. 'It's time to get back to work now! You've got this and can accomplish all of your goals.'"

This pattern reflects the natural development cycle where cutting-edge models require time to be integrated into consumer applications. Developer customization and product development typically lag 6-12 months behind model releases, suggesting these capabilities will likely appear in future rankings.

Deep Research exemplifies this gap—a "magical" but primitive tool that hasn't yet been transformed into specialized consumer applications. Its current prescribed use focuses on market research reports, but the team sees vast untapped potential for more creative implementations, such as tracing meme origins or other specialized information retrieval tasks.

"The known or the prescribed use of Deep Research right now is basically market research reports... But if you try other things like tracing the origin of a meme, Deep Research is like a 100x better version of that Know Your Meme website."

This reveals an opportunity for developers to create more constrained, purpose-built applications around powerful foundation models rather than leaving users with the "blank page problem" of unlimited but undefined possibilities.

Timestamp: [12:57-16:08]

The Gen AI 100 list revealed several unexpected patterns in consumer AI adoption, particularly around which categories have gained mainstream traction and which products have maintained staying power.

One of the biggest surprises was the rapid adoption of "vibe coding" products like Bolt, Cursor, and Lovable among technical audiences. These tools have achieved remarkable saturation within developer communities, demonstrating their immediate practical value.

"Gary Tan had some tweet that like 95% of YC companies or something are now kind of building using those tools. So it's something that every, nearly every developer now is probably using, which was maybe a surprise to me how quickly we reached saturation."

Another consistent surprise has been the enduring popularity of AI companion products, with three such applications appearing in the top 10. Notably, two of these were NSFW-oriented, which aligns with broader internet traffic patterns. Many users are employing these companions as interactive fanfiction platforms, similar to how traditional fanfiction sites rank among the top 100-200 global websites.

Despite the rapid pace of innovation in AI, there's been remarkable consistency across editions of the report. Sixteen web companies have appeared in all four rankings, maintaining their positions despite intense competition from newcomers.

"Across the four lists, there's now 16 companies on the web ranks who have made it every single time and have kept the streak going, which is pretty remarkable when you think of how early we are in AI."

This consistency demonstrates that certain AI companies have successfully cemented their brands and positions in consumer consciousness, building sustainable businesses even in this early, volatile phase of AI development.

The team also noted the absence of more multimodal companion experiences despite the success of text-based ones. While products like Grock have added voice capabilities with distinctive aesthetics and Character.AI has incorporated voice modes, the full potential of multimodal companionship remains largely untapped.

Timestamp: [16:08-18:31]

The conversation reveals fascinating insights about where consumer AI is heading, particularly around app creation and distribution. A key revelation is that traffic patterns for "vibe coding" platforms show more people creating than consuming—suggesting an enormous untapped desire for personal software creation.

"You can track the traffic of like apps that people have launched on lovable.app versus visits to lovable.dev which is where people go to make a lovable product. And lovable.dev has more visits significantly than traffic to lovable.app."

This indicates we're just at the beginning of a potential explosion in AI-generated applications. The team notes that we haven't yet seen the first wave of viral products built on platforms like Lovable and Bolt. When these breakthrough apps emerge, awareness of no-code AI platforms will likely increase dramatically.

The implications for app distribution platforms are profound. As AI dramatically lowers barriers to software creation, traditional gatekeeping mechanisms will be overwhelmed by volume.

"The app store is going to be chaos."

This proliferation of AI-generated apps will create its own meta-problem: how to discover, organize, and manage the flood of new applications. The team suggests we'll eventually need AI solutions just to navigate the AI app ecosystem—an interesting recursive challenge that highlights how fundamentally these technologies are reshaping software creation and distribution.

The future they envision is one where creation becomes increasingly democratized, with personal software (built for audiences of one) and disposable applications (with intentionally short lifespans) becoming commonplace alongside more traditional consumer applications.

Timestamp: [18:31-19:45]

• AI video is finally becoming viable with three new models (Hyo, Clling, and Sora) making the Gen AI 100 list, with Google's V2 model expected to push capabilities even further
• "Vibe coding" products like Cursor and Bolt are gaining mainstream traction, enabling new concepts like "personal software" and "disposable software"
• The new "Brink List" reveals both established players that temporarily fell off the rankings and rising stars poised to break through
• There's a noticeable lag between model capabilities and consumer applications, with voice, screen-aware assistants, and agentic browsers not yet widely adopted
• Models like Deep Research offer "magical" capabilities but remain primitives awaiting specialized consumer implementations
• Developer tools have achieved remarkable saturation, with reports of 95% of Y Combinator companies using AI coding assistants
• AI companion products remain surprisingly popular, with three such applications in the top 10 rankings
• Despite the industry's rapid evolution, 16 companies have maintained their positions across all four editions of the report
• Traffic patterns show more people creating than consuming on platforms like Lovable, indicating massive untapped desire for software creation
• The proliferation of AI-generated apps will eventually require AI solutions just to navigate the resulting "chaos" in app stores

Timestamp: [08:48-19:45]

AI Video Models:

• Hyo - Chinese AI video model that made the Gen AI 100 list
• Clling - Chinese AI video model that made the Gen AI 100 list
• Sora - OpenAI's video model released after being announced over a year ago
• V2 - Google's upcoming video model expected to push capabilities further

Vibe Coding Platforms:

• Cursor - Agentic IDE for technical users that made the rankings
• Bolt - Platform allowing text-to-web app creation for non-technical users
• Lovable - Rising coding platform that appeared on the Brink List

Tools & Models:

• Deep Research - Powerful information retrieval model with untapped potential beyond market research
• Gemini Flash - Screen-aware AI assistant mentioned as upcoming technology
• OpenAI Operator - Model that can complete browser-based tasks independently
• Know Your Meme - Website referenced as being significantly outperformed by Deep Research for tracking meme origins

Concepts:

• The Brink List - New addition to the Gen AI 100 showing the five companies just below the cutoff
• Personal Software - Software designed for an audience of one, newly economical with AI tools
• Disposable Software - Applications with intentionally short lifespans created for specific moments or jokes

Companies Mentioned:

• Y Combinator - 95% of YC companies reportedly using AI coding tools
• Runway - Previously ranked company that appeared on the Brink List
• Otter - Previously ranked company that appeared on the Brink List
• UMAX - Previously ranked company that appeared on the Brink List
• Korea - Rising company on the Brink List
• Grock - AI companion with voice capabilities and distinctive aesthetics
• Character.AI - Companion platform mentioned as having added voice mode

Timestamp: [08:48-19:45]

ChatGPT has maintained its dominant position at the top of the Gen AI 100 list across every iteration for both web and mobile. However, the traffic patterns have not followed a consistent upward trajectory as might be expected.

"It was basically flat for a while which I think was surprising to a lot of people. Between February 2023 basically for a whole year through February 2024, it was essentially flat in monthly visits to the website."

During this plateau period, approximately 50% of ChatGPT's traffic came from students using it for essays and homework problems. Meanwhile, many other potential users—including the a16z team themselves—had not yet integrated the tool into their daily routines or found compelling everyday use cases.

This changed dramatically in recent months, with ChatGPT experiencing extraordinary growth:

"They 2X'ed the number of visits on web since then. They actually made their own announcement too where they counted across web and mobile, and in the past 6 months they grew from 200 million to 400 million weekly active users."

This doubling of scale is particularly remarkable because it occurred after ChatGPT had already reached massive scale—when growth typically becomes more difficult, not easier. The team attributes this resurgence to the release of new models with expanded capabilities, including reasoning models, multimodal functionalities, and advanced voice features.

The data suggests that continued model improvements unlock new use cases, simultaneously attracting first-time users while deepening engagement with existing ones, driving ChatGPT's renewed growth trajectory despite its already dominant position.

Timestamp: [19:40-21:31]

The dramatic growth in ChatGPT usage directly correlates with the release of new features and models that unlocked broader applications and addressed previous limitations. These innovations transformed occasional users into daily active participants by enabling multiple distinct use cases.

"It's both bringing in new users who never tried it and then taking people like me who honestly was maybe a weekly if not less than daily active user. And now I'm a daily active but across several use cases now. Like some days I'm driving and talking to it voice mode. Some days I'm working on a memo and I'm generating something with Deep Research. Some days I'm doing some random other project and I'm brainstorming ideas with it."

The expansion of use cases spans several key areas:

1. Improved coding capabilities that now serve developers more effectively
2. Enhanced data analysis for more reliable information processing
3. Reasoning models that dramatically improve accuracy and reliability

The significance of reasoning models cannot be overstated:

"It's hard to overestimate because in the past you couldn't even rely on ChatGPT to tell you how many Rs were in 'strawberry' accurately. So it was hard to feel good about really tasking any sort of delicate or serious work to it."

This leap in reliability has unlocked a "long tail" of use cases that were previously impractical. Many users who had experimented with AI but couldn't rely on it for meaningful work are now migrating those tasks to these improved models, having gained confidence in their capabilities.

The pattern suggests that continued model improvements will further expand the range of practical applications, bringing more users into the ecosystem while deepening engagement with existing ones.

Timestamp: [21:31-22:47]

Unlike traditional markets where the number two player typically commands around 10% market share with proportionally reduced quality, the AI assistant space has evolved with more nuanced competitive dynamics.

Claude, for instance, occupies an interesting position as a specialized alternative rather than a diminished version of the market leader:

"Claude is not a traditional number two player. Typically the number two player has 10% of the market share and just sort of 10% of the product quality. Instead, Claude sits in this very interesting place where it seems like it's more beloved by a smaller number of people. It's better at creative writing. It seems to have more of a personality, which is interesting because at least I think it's designed to be more constrained. And then it's also strangely much, much better at coding."

This suggests a market where different models serve distinct use cases and user preferences rather than competing solely on scale and general capabilities.

The most dramatic competitive story, however, is DeepSeek's meteoric rise. Despite having only 10 days of data in January (due to launching at the end of the month), DeepSeek immediately captured the number two position on web with approximately 10% of ChatGPT's scale. Its mobile performance was similarly impressive:

"On mobile it had even less than that, 5 days. And it was number 14. And if it had had five more days, it would have been number two. And the gap is even narrower there between DeepSeek and ChatGPT."

DeepSeek's retention metrics are also strong, showing 7% Day 30 retention compared to ChatGPT's 9%. This performance is particularly notable since DeepSeek enjoys structural advantages in markets where ChatGPT is restricted or unavailable:

"If you look at DeepSeek usage, a lot of it is the US but a lot of it is China and other countries where ChatGPT is either—you can't use it or they try to make you not use it and you can only get by it with a VPN. And so in those markets, it's not ChatGPT versus DeepSeek versus Perplexity. It's like DeepSeek versus nothing."

These dynamics suggest a more complex competitive landscape than initially anticipated, with multiple viable players serving different market segments and geographic regions, rather than a winner-take-all scenario centered around ChatGPT.

Timestamp: [22:47-25:30]

The emergence of AI video capabilities represents one of the most significant developments in the latest Gen AI 100 report, with several revealing patterns in this nascent but rapidly evolving space.

A particularly notable trend is the strong performance of Chinese video models, with two such models appearing in the rankings:

"Two of the video models were Chinese video models, which is super interesting. The models are less copyright sensitive in their training data and—a great euphemism—they're maybe more realistic and more prompted here in the outputs as a result."

Beyond training data differences, China's technological ecosystem provides other advantages for video model development, including easier access to human captioning labor and a higher concentration of researchers focused on image and video processing.

These advantages have enabled Chinese video models to achieve impressive capabilities despite having raised relatively less capital than their Western counterparts:

"Sora in some ways was a little bit disappointing for some people, whereas the Chinese video models were maybe better than a lot of people expected given the relative lack of capital that they've raised."

The model landscape is becoming increasingly specialized, with each video model excelling at particular types of content:

"Each model is kind of known for being good at specific things like shots of people, shots of landscapes, anime, hyperrealistic."

This specialization is leading to growing demand for aggregation platforms like Krea (which appeared on the Brink List), which provide unified access to multiple models:

"In so far as we live in this sort of multi-polar world of models—image models, video models, language models—there'll be a role for aggregators like Krea to put them all together in a thoughtful way."

Such aggregation services become increasingly valuable as subscribing to individual models can quickly become expensive, with costs potentially reaching hundreds of dollars monthly for users who need access to multiple specialized tools.

Beyond model differentiation, video applications themselves are becoming more opinionated and specialized in their features:

"The application choices that they're making, that even the model companies are making, are becoming more opinionated. If you've used Runway or Cling or something, you can now prompt basically the camera angles or the wideness of the shot or all of these things a human cinematographer would do."

Among the standout examples is ID (Ideogram), which offers unique capabilities for text generation and aesthetic control that distinguish it from competitors:

"I still think Ideogram is one of the most unique models for what it does, what it's great at, which is text generation, sort of aesthetic that it has. It just sits in a very unique place in the ecosystem."

The development of these increasingly specialized and capable video models and applications suggests that the AI video space is maturing rapidly, with distinct market segments and use cases beginning to emerge despite the technology's relative youth.

Timestamp: [25:30-29:10]

The distribution and success of AI applications differs significantly between mobile and desktop platforms, with distinct patterns emerging based on use contexts and input methods.

"Video actually in general tends to be more of a mobile-first phenomenon. We see tons of, even before AI, tons of applications that focus on creators being able to edit and splice video."

This observation sparked exploration of the key differences between successful AI applications on mobile versus desktop platforms. The team identified several driving factors for mobile success:

1. On-the-go utility: Applications that users want to access while away from their desks naturally gravitate toward mobile.

2. Device-native assets: Mobile apps excel when they leverage data already captured by phones:

"A lot of the things that are working on mobile are either things you want to use on the go or where the asset, the underlying asset you're working with, is easily captured by the phones. The avatar apps will work on mobile because you have, you know, 10 selfies of yourself sitting on your phone."

3. Voice-first interfaces: Voice-based AI products tend to find more success on mobile where voice interaction feels more natural:

"A lot of the voice-first consumer products that we are seeing working actually are on mobile versus web because it's easier and more natural."

These patterns suggest that successful AI applications are increasingly aligning their platform strategies with the natural advantages and limitations of mobile versus desktop environments, rather than attempting platform-agnostic approaches.

The differentiation between mobile and desktop experiences will likely continue to sharpen as AI applications become more specialized and optimized for their specific use contexts and input modalities.

Timestamp: [29:10-29:56]

• ChatGPT maintained flat traffic for approximately a year (Feb 2023-Feb 2024) before experiencing a dramatic 2X growth in the past six months
• Early ChatGPT usage was dominated by students (50%+ of traffic), but new models and features have unlocked broader use cases
• ChatGPT grew from 200 million to 400 million weekly active users in just six months, defying the typical pattern of slowing growth at scale
• Reasoning models have been transformative by dramatically improving reliability, enabling users to trust AI for "delicate or serious work"
• Claude has established itself as a specialized alternative rather than a diminished version of ChatGPT, excelling in creative writing and coding
• DeepSeek achieved extraordinary penetration in just 5-10 days, reaching the #2 position on web and potentially #2 on mobile with strong retention metrics
• DeepSeek enjoys structural advantages in markets where ChatGPT is restricted, essentially competing against "nothing" in those regions
• Chinese video models are outperforming expectations due to different training data approaches and China's research concentration in image/video
• AI video models are becoming increasingly specialized, with different models excelling at specific content types (people, landscapes, anime, etc.)
• Aggregator platforms like Krea are emerging to provide unified access to multiple specialized models, addressing the cost burden of multiple subscriptions
• Video applications are becoming more "opinionated," offering cinematographic controls like camera angles and shot widths
• Ideogram (ID) stands out with unique text generation and aesthetic capabilities that distinguish it from competitors
• Mobile AI success is driven by on-the-go utility, device-native assets (like selfies), and voice-first interfaces that feel natural on phones

Timestamp: [19:40-29:56]

AI Models & Platforms:

• ChatGPT - Market leader with 400M weekly active users after recent 2X growth
• Claude - Specialized competitor excelling in creative writing and coding
• DeepSeek - Rapidly rising model that captured #2 position on web in just 10 days
• Perplexity - Mentioned as competing with ChatGPT and DeepSeek in US markets
• Sora - OpenAI's video model described as "a little bit disappointing" compared to expectations
• Hyo - Chinese video model that performed well on the rankings
• Clling - Chinese video model that performed well on the rankings
• Runway - Video generation tool mentioned for its cinematographic controls
• Ideogram (ID) - Standout model for text generation and aesthetic control
• Krea - Aggregator platform on the Brink List that provides access to multiple models

Features & Capabilities:

• Advanced Voice Mode - New ChatGPT feature driving increased engagement
• Reasoning Models - Transformative improvement enabling reliable outputs
• Deep Research - Information retrieval capability for comprehensive research
• Multimodal Models - 4.0 models with expanded input/output capabilities
• Operator - OpenAI feature that can perform tasks directly on a computer
• Canvas - ChatGPT feature allowing for more natural writing interaction

Metrics & Data:

• 200M to 400M Weekly Active Users - ChatGPT's growth in the past six months
• 7% Day 30 Retention - DeepSeek's user retention metric
• 9% Day 30 Retention - ChatGPT's user retention metric
• 50%+ Student Usage - Early ChatGPT demographic composition

Geographic Information:

• China - Region with advantages in video model development and where ChatGPT access is restricted
• US - Major market for AI models with different competitive dynamics than restricted regions

Timestamp: [19:40-29:56]

The distribution of AI application usage across mobile and desktop platforms reflects both technological capabilities and user behavior patterns. Video-related applications, both pre-AI and now, tend to be mobile-first phenomena, given the creator-centric nature of video editing and sharing.

"Video actually in general tends to be more of a mobile first phenomena. We see tons of, even before AI, tons of applications that focus on creators being able to edit and splice video."

Several key factors determine whether AI applications find more success on mobile or desktop platforms:

1. On-the-go access requirements - Applications that users need while away from their desks naturally gravitate to mobile.

2. Native asset availability - Applications perform better on platforms where their required inputs are readily available:

"A lot of the things that are working on mobile are either things you want to use on the go or where the asset, the underlying asset you're working with, is easily captured by the phones. The avatar apps will work on mobile because you have, you know, 10 selfies of yourself sitting on your phone."

3. Natural interaction modalities - Voice-based AI applications have found greater traction on mobile since speaking into a phone feels more natural for most users:

"A lot of the voice-first consumer products that we are seeing working actually are on mobile versus web because it's easier and more natural to talk into your phone for language learning or for companionship or other use cases than it is to maybe talk into your laptop."

4. Context-appropriate use cases - Certain applications align naturally with mobile contexts, like homework helper apps that students can use on-the-go:

"Same with homework helper apps—those are really blowing up on mobile as compared to web."

These patterns suggest that successful AI products are increasingly designed with platform-specific advantages in mind, rather than pursuing platform-agnostic approaches that might compromise the user experience.

Timestamp: [29:23-30:16]

For the first time, the Gen AI 100 report included rankings based on revenue generation alongside usage metrics, revealing significant divergence between popularity and profitability in consumer AI.

"For the first time we actually ranked the top 50 by what Sensor Tower can measure as mobile revenue, which is typically in-app purchases and subscriptions, so probably not ads, and we ranked those separately from what has the most monthly active users, and there was only 40% overlap between the two lists."

While the primary categories generating revenue (photo/video generators, photo/video editors, beauty filters/enhancers, and ChatGPT-like assistants) matched those with the highest user counts, the specific companies leading each category differed dramatically between the two lists.

A particularly revealing pattern emerged when analyzing revenue per user against total user base:

"We plotted revenue per user versus number of users, and we found the apps that had smaller user bases out of this sample set were much more likely to be making significantly more money on a per user basis. So apps like Speak, apps like Otter, captions and video editing."

These high-revenue, lower-user-count applications typically fit the "prosumer" profile—professional-grade tools that cater to serious users willing to pay premium subscription fees. Many of these apps gate most of their functionality behind subscriptions, deliberately limiting casual usage while focusing on monetizing dedicated users:

"There are companies on here that might be making 50-100 million in ARR off of only a million users, 2 million users. So they wouldn't make the ranks, ironically enough, for monthly active users, but they rank really, really high on a revenue basis."

In contrast, many high-user-count mobile applications employ aggressive user acquisition strategies through app store ads and other low-cost channels. These companies, often operating as indie developers or international app studios, target different economics than venture-backed startups:

"If you're doing this as an indie developer or maybe an app studio running internationally, you're not looking for the 10x payback of acquisition costs that we might be looking for as venture investors. So if you make back one or 2x your money on a user, that's amazing."

This diversity of business models reflects the maturation of the consumer AI space, where companies are finding various paths to sustainability beyond the traditional venture-backed growth model. For founders, the appropriate monetization strategy depends largely on the nature of their product and its potential audience:

"Some of these markets are naturally maybe not mainstream behavior. Like one example of a category that did appear on the mobile revenue list but was not on mobile usage was several plant identification apps... If you're one of the like—I can think of a few relatives who love plants or love birds—totally they'll pay $100 a year for that and they'll use it every day or every other day."

The findings suggest that founders should optimize their business model based on the inherent characteristics of their product rather than pursuing universal growth metrics.

Timestamp: [30:16-34:06]

After four iterations of the Gen AI 100 report, some fundamental patterns have emerged that transcend specific technology trends. These insights offer valuable guidance for founders and investors navigating the consumer AI landscape.

The most profound realization is that despite all the technological sophistication of AI, consumer product success ultimately hinges on the same timeless principles that have always governed consumer markets:

"The biggest thing now, being a consumer investor for close to a decade now, it's almost like the more you know, the less you know in some cases, because it all just comes back to the product at the end of the day. Technologists or investors can have opinions on the best monetization strategy or the best growth hacks, but in the end if the product isn't capturing users' attention and isn't retaining them, the business is just going to be kind of a completely leaky bucket—users in, users out."

This principle creates a particular challenge for technical founders coming from AI research backgrounds. Technical sophistication doesn't automatically translate to consumer product success:

"Often we meet with these amazing PhD researchers, best-in-class in the whole world in terms of their technical understanding of a model or capability, and they can struggle building in consumer sometimes because often the more complicated thing is not actually the thing that is highest utility, most delightful, most helpful to a consumer user."

The team emphasizes that sometimes older models might deliver better user experiences than cutting-edge ones, or that limited AI integration might prove more effective than comprehensive AI deployment if the technology isn't sufficiently stable. These decisions should be guided by user experience requirements rather than technological showcasing.

The overarching recommendation for AI founders is to focus relentlessly on the fundamental value proposition—either the specific pain point being addressed or the unique experience being created—and let that guide technological implementation decisions:

"We never like to be prescriptive on consumer products, but in general we see when teams focus on either the pain point they're trying to solve or the unique experience they're trying to create and build towards that... In consumer, you really have to kind of let the data be your guide."

This perspective brings the Gen AI landscape back to first principles of consumer product development, suggesting that while AI enables new possibilities, the rules of creating successful, sustainable consumer products remain fundamentally unchanged.

Timestamp: [34:06-35:40]

• Video applications tend to be mobile-first, reflecting both pre-AI and current creator behaviors around video editing and sharing
• Success on mobile vs. desktop is driven by factors like on-the-go access needs, native asset availability, natural interaction modalities, and context-appropriate use cases
• Voice-based AI applications perform better on mobile where speaking feels more natural and contextually appropriate
• Only 40% overlap exists between the top apps by revenue and the top apps by user count, revealing distinct paths to business success
• Apps with smaller user bases often generate significantly higher revenue per user, particularly in the prosumer category
• Some companies achieve $50-100M in annual recurring revenue with just 1-2M users through subscription-based models
• High-user-count apps often employ different economics, with indie developers and international app studios targeting 1-2x return on acquisition costs rather than the 10x pursued by venture-backed companies
• Niche applications like plant identification apps can generate substantial revenue despite limited mainstream appeal by serving passionate users willing to pay premium prices
• Technical sophistication in AI doesn't automatically translate to consumer product success, as the "more complicated thing is not actually the thing that is highest utility"
• The fundamental drivers of consumer AI success remain product-market fit and user retention, despite all technological advancements
• Successful AI founders focus on the specific pain point or unique experience they're creating, rather than showcasing AI capabilities
• Sometimes older models or limited AI integration deliver better user experiences than cutting-edge implementations if they better serve the core value proposition
• In consumer AI development, "you really have to let the data be your guide" rather than following predetermined strategies or assumptions

Timestamp: [29:23-35:40]

AI Applications & Products:

• Speak - AI language learning app mentioned as having high revenue per user
• Otter - Transcription app with high revenue per user despite smaller user base
• Plant Identification Apps - Niche category that appeared on revenue rankings but not usage rankings

Metrics & Data:

• Sensor Tower - Data provider used to measure mobile revenue through in-app purchases and subscriptions
• 40% Overlap - Percentage of apps appearing on both the revenue and user rankings
• $50-100M ARR - Revenue range achieved by some prosumer apps with just 1-2M users
• 1-2x Return - Target return on user acquisition for indie developers and international app studios
• 10x Return - Target return on user acquisition for venture-backed companies

Industry Terms:

• ARR (Annual Recurring Revenue) - Metric used to measure subscription-based business performance
• User Acquisition - Process of gaining new users through paid and organic channels
• App Store Ads - Acquisition channel mentioned for mobile applications
• Prosumer - Category of applications that bridge professional and consumer markets with premium features
• In-App Purchases - Monetization method measured in revenue rankings
• Subscriptions - Primary revenue model for high-revenue-per-user applications

Product Categories:

• Photo/Video Generators - Category appearing in both usage and revenue rankings
• Photo/Video Editors - Category appearing in both usage and revenue rankings
• Beauty Filters/Enhancers - Category appearing in both usage and revenue rankings
• ChatGPT-like Assistants - Category appearing in both usage and revenue rankings
• Voice-First Products - Category more successful on mobile than desktop
• Homework Helper Apps - Category mentioned as "blowing up" on mobile compared to web
• Avatar Apps - Mobile-dominant category leveraging device-native assets like selfies

Timestamp: [29:23-35:40]

Report Availability:

• The full GenAI 100 list (4th edition) is available at: https://a16z.com/100-gen-ai-apps-4/
• The report includes rankings of top 50 web products, top 50 mobile apps, and for the first time, top 50 mobile apps by revenue

Podcast Information:

• This episode is part of the A16Z podcast series
• Subscribers get exclusive access to video content
• Hand-selected related videos are available on the A16Z channel

Featured Experts:

• Anish Acharya - a16z General Partner
• Olivia Moore - a16z Partner
• Both specialists in the consumer AI landscape with expertise in tracking application development and adoption patterns

Timestamp: [35:40-36:10]