French-made, AI-powered: Another AI Product 'Tops' the US Overall Rankings

[baoshi.rao]

On April 1, 2024, an AI dating product named Ava topped the US free download chart.

The most impressive part is that Ava's entire team consists of no more than 5 people, and they only hired their first product manager and first product designer a week ago.

In the screenshot above, apart from Ava, which most people have never heard of, the remaining 7 products are all familiar names to users and entrepreneurs on the US charts.

So how did Ava manage to surpass these popular products and top the US download charts? The analysis mainly points to the following three reasons:

1. Despite its low profile, Ava has strong backing. According to Crunchbase data, Ava has secured a total of $5.8 million in seed funding from renowned investment institutions, including OpenAI and Sequoia Capital.

2. Unlike the common focus on competing over models, tools, and efficiency, Ava is a social assistant app. Before users respond to greeting messages from people they're interested in, Ava acts as a wingman, facilitating conversations to help users get to know each other better. 3. The most crucial aspect is to skillfully utilize node marketing. Observant readers might have noticed that we bolded "April 1st" at the beginning of the article—April 1st being April Fools' Day.

Indeed, there was never any truth to Ava topping the U.S. charts. It was merely Ava's founder, Clara Gold, capitalizing on the psychological aspect of April Fools' Day—where people are more forgiving of harmless lies—to execute a successful holiday marketing campaign.

As people grow older, April Fools' Day has faded from the minds of many, becoming just another Monday at work for most. However, pulling off this stunt wasn't easy for Ava either: First, to make the message appear as authentic as possible, Clara employed substantial data to underscore the challenges behind this "successful climb to the top."

"After 15 months of exploration and effort, we've reached the #1 spot on the App Store's overall chart. I can finally say with pride that AI can improve people's love lives. To achieve this goal, we conducted 603 user interviews, redesigned the product 3 times, pushed 246 versions to the App Store, got rejected by Apple 57 times, fixed 2,354 bugs, and attracted 1.6 million users to join the application."

By listing detailed data and common startup challenges, Clara further enhanced the credibility of her message and captured attention—demonstrating her mastery of social media engagement strategies. Additionally, Clara specifically chose 8:59 PM as the timing, when most people had already left work and might not specifically check the rankings again. Moreover, 'mature adults' might not have remembered that today was the 'legitimate' day for jokes.

Besides the X platform, Clara Gold also shared the good news of topping the charts on LinkedIn. The post received congratulations from over 30 investors and various corporate executives.

This interesting April Fools' Day marketing campaign has once again brought AI Dating back into the public eye. Although founder Clara's tweet was a joke, the three product design iterations it mentioned were real.

Ava wasn't initially positioned as an AI wingman.

At least when receiving OpenAI's seed round investment, Ava was still telling the story of being a 'virtual friend for teenagers'.

In June 2023, the first-generation Ava was launched, and the product was so simple that it hardly resembled a social product from 2023. Users could create a personalized avatar and freely select and set clothing for it. Once set up, they could interact with Ava through text and voice messages.

From the author's actual experience in July 2023, the avatars and clothing options in the product were poorly designed and unattractive. Additionally, the interaction and chat experience did not feel smooth or genuine, lacking the expected qualities of a digital companion. The experience was not significantly different from chatting with ChatGPT 3.5.

Even when discussing the product with a virtual social entrepreneur, they jokingly remarked, 'OpenAI must still have a lot of money.' Clara Gold chose to shut down the product in October 2023. Shortly after the shutdown, the second-generation Ava was officially launched in November.

The most significant change in Ava before and after the revamp is the shift from the original 'human-AI' model to a 'human-AI-human' model. Users can now not only interact with Ava but also chat with potentially suitable users selected by AI algorithms.

The second-phase Ava still focuses on the traditional dating model, where users upload their photos and like other users they are interested in. Ava even uses algorithms and users' selected interests to pick out the 'best date of the day' for them. This "photo feed + AI interaction" format persisted until January 2024. Starting January 28, 2024, Ava officially entered the 3.0 era, adopting the "instant video recording + vertical swipe + AI wingman" mode.

In the new version, Ava emphasizes that users don't need to be perfect or overly elegant in this community. Therefore, the system doesn't allow users to upload their own photos or videos. Users can only use the platform's built-in camera to record a short, unedited front-facing video without beauty filters. This video will serve as the user's profile picture and main page display. Users can swipe up and down on the homepage to interact with other profiles. They can directly like users they're interested in or simply swipe up to pass. Ava officially refers to this interaction mode as 'TikTok Style'.

In fact, Ava isn't the first dating product to adopt TikTok's classic interaction pattern. Previously, dating apps like Lolly and Snack have also used this as their selling point.

Compared to the harshness of directly clicking 'dislike', swiping away can also indicate rejection but appears more graceful and gentle. Users can directly click to start chatting with the Ava of their interested user. Typically, the conversation will proceed in three rounds. Ava initiates with a multiple-choice question, and only after the user answers the first question will Ava proceed to the second question.

When the user completes three rounds of responses, the system will provide "three preferences and three disliked behaviors" of the counterpart. Ava will then automatically exit the conversation, leaving the subsequent interaction to be conducted directly between the two human users.

It's worth noting that Ava is very meticulous in collecting users' personality traits and information. It encourages users to select personality tags from various categories such as "social life, romantic life, hobbies, serious topics, personality traits," while also prompting users to identify behaviors they dislike or cannot tolerate. Ava emphasizes that the more detailed the user tags are selected, the more obvious Ava's role as a wingman will be.

From the author's personal experience, Ava's greatest role is to save time and improve efficiency. When using apps like Tinder and Bumble, users often encounter situations where there are many matches but few conversations.

One reason is the presence of bots designed to inflate numbers, and another is missing the optimal interaction time. Regardless, not being able to interact with potential dates that one has carefully selected is a significant waste of time and emotional investment. Ava's three questions automatically filter out robots and users who are not genuinely interested, allowing users to engage only with those who show strong interactive intent.

In simpler terms, you must first pass the assistant's screening to have a chance to discuss cooperation with the general manager.

The approach is polite, but the issue is that Ava's model actually requires a certain user base. During a full day of testing by the author, none of the 17 users of interest resulted in a successful conversation with a real person, including 4 users who had already passed Ava's screening. According to data from Diandian, Ava's iOS downloads have not exceeded 20,000 in the past three months. Even with a 20% monthly retention rate, the probability of matching with active real users remains extremely low.

Therefore, Ava's most important task at present must be growth.

Unlike many foundational models and Bot-type products that struggle to find suitable application scenarios, Ava's applicable scenarios are very clear. Based on our past observations of the dating track, escaping the "ghost" phenomenon is a demand for the vast majority of users. It can even be said that ending endless swiping and establishing real connections with matched users is a need for almost all dating users. AI wingman products, including Ava, represent an attempt to improve matching efficiency, though we cannot yet definitively conclude whether they are effective.

Since the explosive growth of ChatGPT in late 2022, AI wingman products have emerged in large numbers.

From the perspective of AI wingman service providers: There are dating apps like WooPlus and Grindr that have independently introduced wingman services; service providers like WingAI and Rizz that don't focus on a specific platform but offer dating guidance to users; AI bots created by users themselves using tools like GPTs; and wingman services provided by dating app influencers who combine their dating experience with AI.

From the perspective of AI wingman service offerings:

These services include but are not limited to optimizing dating app profiles, crafting opening lines, providing personal image guidance, suggesting conversation topics, answering specific questions interactively, and facilitating the dating process - all aimed at improving matching efficiency, online response rates, and offline meeting rates across different stages. There are many options available, but few AI wingman products are actually familiar to users and can truly deliver value. At least in the top 50 AI traffic product rankings, AI wingman products have never made an appearance. Here are the reasons for this:

1. Users can directly use large model products like GPT to solve most daily problems without additional costs.

In online interactions, users have ample time to find the optimal solution to a problem. However, in offline interactions, when users need an AI wingman, they often cannot access it in time. 2. Overall, conversations on dating apps remain relatively private. Most current wingman platforms require users to screenshot profiles or interaction records to provide targeted suggestions, which creates an invisible barrier for privacy-conscious international users.

3. The intelligence of AI wingmen is still insufficient. After testing multiple wingman products, the author's direct impression is that they wouldn't use these products again unless for testing purposes. The AI fails to accurately capture users' personalities and language characteristics, thus unable to fully function as a dating assistant. It neither enhances the user's appeal in the eyes of their matches nor effectively captures useful information about the other party, making it ultimately ineffective.

Moreover, and more importantly, human interactions often rely on a certain "chemistry," but current AI wingmen seem unable to successfully create an online atmosphere suitable for romance or dating. Of course, we do not deny that some AI wingman products have achieved decent profitability through small-scale user bases, but most are based on forced subscription models—essentially 'one-time deals' rather than sustainable businesses.

As for whether the continuous development of AI and targeted training of AI wingmen can change the current situation, we can only wait and observe.

Finally, it must be said that Ava's non-annoying April Fools' marketing campaign has shown us an interesting new approach to growth, while also prompting AI entrepreneurs to reflect once again on what problems AI+ is actually solving and whether it is truly addressing them.