Source: a16z|Dalei
On November 30 last year, ChatGPT, a chatbot program developed by OpenAI, was officially released, and in the past 10 months, it has become one of the fastest growing applications in history with data of more than 100 million monthly active users.
At the same time, the rapid development of ChatGPT also announced that a new era of generative artificial intelligence has arrived, so a large number of new generative AI (GenAI) products have burst out. In addition to ChatGPT, there are also Midjourney, You.com, Bard and other products abroad, and in China, there are also many well-known Internet companies entering the AI track, and the more well-known ones are Baidu Wenxin Yiyan, Alibaba Tongyi Qianwen, Huawei Pangu and so on.
Although the AI boom is in the limelight, from the perspective of application scenarios, generative AI (GenAI) products are more oriented to the B-side, that is, reconstructing the product architecture and business model through large language models to improve the core competitiveness of enterprises.
So, focusing on each consumer, how do they interact with generative AI products? Of all the products today, who is likely to be the next big winner?
Not long ago, a16z, a well-known private venture capital firm in the United States, ranked the top 50 GenAI network products in the market by traffic data (as of June 2023) based on SimilarWeb's traffic data.
Note: This list is generated based on global desktop and mobile web access to data from SimilarWeb as of June 2023. However, for companies on the list that also have mobile apps, a16z added an estimate of their app "traffic" (MAUs x per MAU session per month), as well as data from Sensor Tower as of June 2023, as well as web traffic to determine where they are on the list.
This list does not include app-only companies, nor does it include activities that occur within Discord servers.
In this regard, Morketing came to six conclusions based on the list and the data provided by a16z:
1. Most generative AI (GenAI) products are less than 1 year old2. ChatGPT is far ahead with the advantage of faults 3.AI Companion and AI content generation tools are growing the fastest4. Product categories are clearly differentiated and competition is intensifying5. Consumers are willing to "pay" for top products6. Mobile apps are still emerging as GenAI platforms
Most generative AI (GenAI) products
Born no more than 1 year ago
In this list, ChatGPT is the earliest generative AI (GenAI) product, which has only been released for 10 months. Most of the products on this list are 80% new. That said, while many companies are eager to enhance their products with AI, at least the consumer experience is relatively new.
Meanwhile, 5 of the 50 companies on the list are products of existing big tech companies: Bard (Google), Poe (Quora), QuillBot (Course Hero), Pixlr (123RF), and Clipchamp (Microsoft). In the remaining list, up to 48% of the products are not backed by external funding.
A16z says one of the main differences between companies with external funding and being fully self-reliant is the technology stack. After all, depending on the size of the model, it can cost millions of dollars to set up and train accordingly.
According to industry insiders, at least 100+ projects are working to commercialize AI agents, and nearly 100,000 developers are building autonomous generative AI (GenAI) products.
If you divide the 50 companies in the list, you can roughly divide them into three categories: first, there are trained proprietary models (such as ChatGPT); Second, it fine-tunes its own large language model (such as WizardMath); Third, consumer user interfaces (e.g., "GPT-4") are built on top of existing models.
It is worth noting that half of the top 10 products on the list are building their own exclusive models, 4 are fine-tuned, and only one belongs to the "packaging" category.
That doesn't include ChatGPT, whose parent company, OpenAI, raised $11.3 billion, compared to $98 million on average from other proprietary companies. This compares to $20 million for companies fine-tuning the open source model, compared to $9 million for "packaging" companies.
ChatGPT is far ahead with the advantage of faults
From the list, we can see that ChatGPT accounts for 60% of the monthly traffic in the entire list, with an estimated 1.6 billion monthly visits and 200 million monthly active users (data as of June 2023). This undoubtedly makes ChatGPT the 24th most visited website in the world.
Although CharacterAI is gaining momentum and has firmly ranked second on the list, it is about 21% of ChatGPT's size. Especially in the mobile device space, according to Sensor Tower data, CharacterAI is one of the strongest early players, with a DAU that is fully comparable to ChatGPT and a significantly higher retention rate.
It has also been previously reported that chatbot startup CharacterAI is negotiating a valuation of more than $5 billion. But even so, ChatGPT leads the way with 1.5 billion monthly active data.
But when viewed across the mainstream consumer market, when web and mobile app traffic is combined, ChatGPT ranks about the same size as Reddit, LinkedIn and Twitch, but far below platforms like WhatsApp, YouTube, and Facebook.
According to statistics, WhatsApp has 2.5 billion monthly active users, covering more than 180 countries and regions around the world, and is growing into a super application platform; YouTube, the world's largest video site, has about 3.2 billion monthly active users; In the second quarter of this year, Facebook's monthly active users (MAU) exceeded the 3 billion mark for the first time.
AI companions and AI content generation tools are growing the fastest
In the list, LLM chatbots account for 68% of total consumer traffic. In addition to ChatGPT, this category includes Google's Bard and Quora's Poe, both in the top five.
In recent months, AI companions (such as CharacterAI) and AI content generation tools (such as Midjourney and ElevenLabs) have grown significantly and are starting to gain attention. In the content generation category, image generation accounted for 41% of traffic, followed by professional consumer writing tools at 26% and video generation at 8%.
The reason is that AI companions can simulate dialogue and communication between humans, so they can be applied to virtual customer service, intelligent robots and other fields to provide a more humanized service experience. AI content generation tools are widely used in scenarios to improve production efficiency and save creation costs.
Another category worth mentioning is Model Center, which is capable of generating huge traffic. Examples include Civitai (a platform for sharing and discovering AI art resources designed to help users easily explore and use a variety of AI art models) and Hugging Face (an open-source library for building, training, and deploying state-of-the-art NLP models).
By visiting these websites, consumers are able to download models that are easy to run locally, so counting only network traffic can underestimate actual traffic.
Product categories are clearly differentiated and competition is intensifying
Despite the surge in interest in generative AI among most consumers, there are still many categories that have yet to be successful.
As shown in the figure, except for the companion category, the gap between content generation categories such as pictures and writing is often less than 2 times. At the same time, the different categories are clearly differentiated, and products built specifically for specific jobs are growing along with more general-purpose tools and gradually showing signs that they can also be successful companies.
In the image generation category, while Midjourney currently dominates the category, companies like Leonardo (which specializes in gaming assets) are growing rapidly in traffic and also have a lot of room for growth.
The image below shows data on Midjourney's Discord server with Leonardo's monthly visitors. Despite the varying scale of growth, Leonardo has been able to accommodate millions of users.
Consumers are willing to "pay" for top-of-the-line products
Over the past 5 years, many consumer apps have been caught up in the acquisition game. In the absence of a similar shift from the Internet to mobile apps, it is difficult to stimulate their interest in new products. At the same time, acquisition costs have been rising, which means that most companies have to consider things like lifetime value and customer acquisition costs as top considerations.
The emergence of generative AI (GenAI) products will be a game-changer. For example, most companies in the a16z list do not have a paid marketing model, because they can often get some referrals and organic traffic growth through X, Reddit, Discord, and email.
A quarter of the lowest generative AI (GenAI) products on the list generate 2% of paid traffic. By comparison, according to a16z's benchmark of 150 products, a quarter of non-AI-generating AI companies generate 70% of paid traffic.
Consumers are more willing to "pay" for generative AI (GenAI) products. 90% of the companies on the list are already profitable through subscription models. The products on this list earn an average of $21/month (for users of the monthly plan) and $252 per year.
Essentially, generative AI unlocks a new level of value, increasing consumers' willingness to pay.
Mobile apps are still emerging as GenAI platforms
So far, consumer AI products have been dominated by browsers rather than applications. Even ChatGPT took 6 months to release its first mobile app.
There are some exceptions, such as in the image generation category, where the "launch barrier" of applications is quite low due to third-party APIs, and products like Lensa and WOMBO will see sales rise sharply or drop dramatically.
So why aren't most AI companies moving into mobile? The main reason is that browsers are the natural starting point for reaching the widest range of consumers.
Since most AI companies started late and had smaller teams, they didn't want to split their attention and resources across the web, iOS, and Android. As a result, only 15 companies on the list currently have mobile apps, and less than 10% of their monthly traffic comes from mobile apps.
Of course, there are 3 exceptions to the list: consumer design studio PhotoRoom accounts for 88% of the total traffic, companion app CharacterAI accounts for 46% of the app traffic, and text-to-speech product Speechify accounts for 20%. These companies are a bit more engaged on mobile apps than their websites.
Given that most consumers now spend 36 minutes more per day on mobile devices than on PCs, more mobile-first GenAI products are expected to emerge as the technology matures.
Reference: How Are Consumers Using Generative AI?