Jun 4, 2024 · 15m · a16z
The Top 100 GenAI Consumer Apps
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The a16z Podcast, consumer partner Olivia Moore breaks down the 'Top 100 GenAI Consumer Apps' report, analyzing web versus mobile consumer behaviors, viral companion platforms, and the global decentralization of AI innovation.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 21.9% of the talking time here. How this is scored →
The host as informed peer 4.2 Guest teaching 5.4 Guest disagreement 1.2 The host pushing back 1.0
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 7:15 Re-framing companion app use cases
The guest gently corrects the host's suggestion that companion apps are breaking out in coaching or therapy, clarifying that current viral growth is concentrated in uncensored consumer tools.
Hardest push from the host ▶ 4:17 Questioning model clarity for consumersThe host presses on the practical reality of App Store arbitrage by asking if average users can actually tell the difference between GPT-3.5 and GPT-4.
Biggest teaching moment ▶ 3:26 Explaining App Store optimization arbitrageThe guest details how opportunistic mobile developers capitalized on ChatGPT's delayed app rollout by creating thin wrapper products around OpenAI's API.
The host holds their own ▶ 14:27 Historical analogy on audio tech maturationThe host demonstrates deep domain perspective by comparing current AI audio infrastructure delays to the decade-long gap between early web blogs and dedicated podcast apps.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Introducing the Top 100 GenAI Consumer Apps Report | 3 | 5 | 1 | 1 | The host introduces the report and demonstrates awareness of the previous iteration from September. The guest explains the research methodology involving SimilarWeb and Sensor Tower to highlight the minimal overlap between top web and mobile AI products. | |
| Mobile Avatar Apps, Dating Keyboards, and ASO Arbitrage | 4 | 6 | 1 | 1 | The host asks a sharp question about whether regular consumers can distinguish between GPT-3.5 and GPT-4. The guest educates on how ASO arbitrage apps monetize free underlying models by targeting non-technical mobile users. | |
| Emerging Web Categories: ElevenLabs, Gamma, and Companion Apps | 4 | 5 | 2 | 1 | The host asks technically informed questions about whether companion apps rely on Llama or custom fine-tuned models. The guest explains how companion apps expand into uncensored domains and gamified mechanics like card trading. | |
| Character.ai and Exceptional User Engagement Metrics | 4 | 6 | 1 | 1 | The host seeks precise definition on engagement metrics, asking whether a 'session' means an individual message or full conversation. The guest breaks down Character.ai's extreme usage metrics that rival major social messaging apps. | |
| Discord Communities as a Leading Indicator for Web Apps | 6 | 5 | 1 | 1 | The host actively contributes context by citing a viral example of unstructured voice prompting and making a historical comparison to early web blogging. The guest outlines how tracking Discord invite traffic acts as a leading indicator for upcoming web platforms. |
Statements from this episode (18)
Only 5-6 GenAI apps overlap between top 50 web and mobile lists
“I think there was only five or six companies that made both lists.”
Every mobile app on a16z's GenAI ranking is brand new
“All the mobile companies were brand new to the list.”
OpenAI's delayed mobile launch cost ChatGPT crucial App Store optimization ground
“The one other interesting thing about mobile was ChatGPT was so slow to launch an app, they lost a little bit of ground just from, like, an, App Store keyword and optimization perspective.”
Multiple top mobile AI apps are copycat wrappers arbitraging OpenAI's models
“So there are quite a few apps on our mobile app list that are essentially copycat products and using OpenAI's own model, but are named slightly differently, and are just kind of playing this, like, maybe short to medium term arbitrage on App Store optimization…”
Consumers unknowingly pay third-party wrapper apps for free OpenAI models
“If you're a normal person who doesn't know the names of these models, you might not understand that, like, on this other app, you're paying for access to the free version of the OpenAI model that you would get on ChatGPT.”
AI presentation generator Gamma transitioned from consumer virality to workplace adoption
“Gamma was another one, which is an AI presentation document slide deck builder that I think is another example of one that kind of went consumer viral and now getting pulled into workspaces.”
AI companion apps on a16z's top web list grew from 2 to 8
“I think on the first list there were two companion products, and I think there were eight on the most recent one.”
Most top web AI companion apps rely on uncensored models
“Most of them were actually uncensored models. So not to say they were all NSFW, because that isn't necessarily the use case, but they allow you more freedom to talk about different topics, do different things than you can do if you were just trying to use Chat…”
AI companion app Talkie gamifies interactions with tradable digital cards
“There's a company called Taki that's almost appealing to, like, the gamer-trader mindset when it comes to companions, where you actually unlock, like, trading cards as you talk with companions for different amounts of time, and then you can kind of sell them i…”
Uncensored AI companion apps can reach 20 million monthly web visits
“They're growing, but maybe not the zero to twenty million monthly web visits that some of these uncensored companion apps can get.”
The average Character.ai user logs nearly 300 sessions per month
“Character AI has the average user close to 300 sessions per month, which again is like 10 sessions per day.”
The average session duration on Character.ai exceeds five minutes
“An average session for a company like Character is probably gonna be upwards of five minutes long.”
AI talent hubs and founder networks are decentralizing globally
“As AI matures, these mafias of talent or hubs of talent are spinning out of different research labs, different universities, and kind of building these different communities of founders around the globe, which is very cool.”
Discord invite traffic predicts which generative AI products will launch top apps
“Like, if you wanted to predict what's gonna be on the list in six months, Looking at the Discord data, you can even track, like, which Discord pages are getting the most invite traffic or things like that as a proxy for interest. And many of those companies th…”
AI music generator Suno debuted among top web apps via Discord
“And we've seen that be fairly predictive now of Suno was one of the highest ranking new companies that we saw on the list this time, and they just sprouted out of a successful Discord.”
Midjourney generates nearly 1% of all invite traffic across Discord
“Like Midjourney itself is driving, you know, close to a percent of all kind of Discord invite traffic, which is very cool.”