Artificial Intelligence
AI Companions To Earn $120M In 2025 As AI Replaces Historians & Gen Z Workers
Updated on Thu, Aug 14, 2025
Share
Social media isn’t the only way people are forming new connections. Today, AI companion apps have quietly become a global phenomenon. From virtual friends and soulmates to anime characters and even snarky foxes, these AI apps offer conversations that feel personal, and users are engaging with them for emotional, financial, and personal advice.
Fresh figures from app intelligence firm Appfigures reveal a boom in AI companion apps that goes well beyond giants like ChatGPT and Grok. Out of the 337 earning revenue today, nearly 40%, or 128 to be exact, were launched in 2025 alone.
Collectively, these apps raked in $82 million in just the first half of 2025 and are on pace to top $120 million by year’s end. Downloads are surging as well, with 60 million in the first six months alone, an 88% jump from last year, bringing total installs to 220 million worldwide.
Revenue is outpacing download growth. By July, global consumer spending on these apps reached $221 million, a 64% jump from the same time in 2024. The market is heavily top-loaded, with the top 10% of apps pulling in 89% of all revenue, and about 33 titles surpassing $1 million in lifetime earnings.
Well-known players in the space include Replika, Character.AI, PolyBuzz, and Chai. Romance dominates the categories with 17% of active apps featuring “girlfriend” in their name, while only 4% use “boyfriend.” Other themes like fantasy, anime, and soulmate are also less common.
Naturally, Big Tech is not ignoring the trend. In July, xAI’s Grok introduced its own AI companions, while OpenAI sparked strong reactions with GPT-5 before bringing back its 4o model after user pushback. Google joined the action last year, hiring Character.ai founder Noam Shazeer, even as the app itself continues to draw controversy and tens of millions of active users.
This boom in AI tools isn’t limited to social interactions; it also raises questions about how AI is impacting professions and work expertise.
Microsoft’s recent study on jobs most likely to be augmented by AI ranked “historian” second, a placement that left many in the field uneasy. However, as one veteran researcher discovered, AI is still far from matching the depth and accuracy of human expertise.
The author has spent over a decade researching the movies watched by U.S. presidents. Several AI chatbots, including OpenAI’s GPT-5, Microsoft Copilot, Google Gemini, Perplexity, and xAI’s Grok, were put to the test using questions drawn from rare sources such as the White House projectionist logs.
GPT-5 came up empty. Copilot guessed wrong, even in its 3,500-word “Deep Research” mode. Perplexity made the same incorrect guess, Gemini offered no answer, while Grok got it right only by pulling the information from the author’s own 2019 tweet.
These missteps highlight an important reality: AI frequently guesses, misinterprets, or relies on incomplete information. Despite bold promises from tech CEOs, AI tools remain unreliable for niche or archival research.
As the author points out, “These tools still need human babysitters because they get so many things wrong so frequently.”
These limitations of AI sparked a broader discussion about how emerging AI tools will reshape careers and opportunities for younger generations.
OpenAI CEO Sam Altman painted a strikingly optimistic picture for Generation Z in a recent podcast, calling today’s young people “the luckiest kid in all of history,” even as he acknowledged looming job displacement from AI.
Altman said, “this always happens, and young people are the best at adapting to this. I’m more worried about what it means, not for the 22-year-old, but for the 62-year-old that doesn’t want to go retrain or reskill.”
Altman highlighted how tools like GPT-5 make entrepreneurship and innovation more accessible, allowing individuals to achieve feats once reserved for large teams. He stated that “half of the entry-level white-collar workforce will be replaced by AI” in the next five years but expects new, exciting, and well-paid jobs to emerge.
He also discussed the challenge of distinguishing real from AI-generated media, noting that society has always adapted to gradual shifts in technology: “the threshold for how real does it have to be to consider to be real will just keep moving … media is always a little bit real and a little bit not real.”
Altman urged people to use AI tools, saying humility and adaptability are essential in a fast-changing world. “Just using the tools really helps,” he said.
While AI opens new doors for innovation, it also shows the importance of reliability, especially when users depend on it for accurate information.
Even the most advanced AI models, like large language models (LLMs), aren’t always truthful. They can give answers that sound confident but are actually wrong. This problem is often called hallucination, but researchers say it goes beyond just making things up.
Jaime Fernández Fisac, an assistant professor at Princeton, and his team call this tendency “machine bullshit.” It’s not just about making things up, but AI can mislead through half-truths, vague answers, or even flattery that sounds convincing but isn’t accurate.
The researchers also found that common training methods, like reinforcement learning designed to make AI more engaging, can actually make the problem worse. To tackle this, Fernández Fisac’s team created a “Bullshit Index,” a tool to track how often AI misleads users. By measuring these patterns, developers can refine AI training and reduce misinformation.
As AI becomes part of everyday life, tools like the Bullshit Index are meant to keep models honest, helping users get answers that are not just convincing but actually reliable.
Even though AI promises convenience, tools still struggle to offer real-world comfort.
Mozilla’s latest Firefox update, version 141.0, brought a new AI-powered feature that automatically groups similar tabs and even suggests names. However, not everyone was happy; some users reported major CPU spikes, with one noting usage jumping from 0.05% to 130%.
Mozilla calls this “abnormal CPU spikes” and says it shouldn’t happen under normal use. While the feature is still experimental and can be turned off manually, it is rolling out gradually. Although it’s meant to make browsing easier, some frustrated users see it as unnecessary bloat, showing that AI in everyday tools isn’t always smooth, or necessary.
Do you think AI companions can really replace human connection? Can we trust AI tools more than human experts?
Let us know what you think in the comments section below!
Fresh figures from app intelligence firm Appfigures reveal a boom in AI companion apps that goes well beyond giants like ChatGPT and Grok. Out of the 337 earning revenue today, nearly 40%, or 128 to be exact, were launched in 2025 alone.
Collectively, these apps raked in $82 million in just the first half of 2025 and are on pace to top $120 million by year’s end. Downloads are surging as well, with 60 million in the first six months alone, an 88% jump from last year, bringing total installs to 220 million worldwide.
Revenue is outpacing download growth. By July, global consumer spending on these apps reached $221 million, a 64% jump from the same time in 2024. The market is heavily top-loaded, with the top 10% of apps pulling in 89% of all revenue, and about 33 titles surpassing $1 million in lifetime earnings.
Well-known players in the space include Replika, Character.AI, PolyBuzz, and Chai. Romance dominates the categories with 17% of active apps featuring “girlfriend” in their name, while only 4% use “boyfriend.” Other themes like fantasy, anime, and soulmate are also less common.
Naturally, Big Tech is not ignoring the trend. In July, xAI’s Grok introduced its own AI companions, while OpenAI sparked strong reactions with GPT-5 before bringing back its 4o model after user pushback. Google joined the action last year, hiring Character.ai founder Noam Shazeer, even as the app itself continues to draw controversy and tens of millions of active users.
This boom in AI tools isn’t limited to social interactions; it also raises questions about how AI is impacting professions and work expertise.
Microsoft’s recent study on jobs most likely to be augmented by AI ranked “historian” second, a placement that left many in the field uneasy. However, as one veteran researcher discovered, AI is still far from matching the depth and accuracy of human expertise.
The author has spent over a decade researching the movies watched by U.S. presidents. Several AI chatbots, including OpenAI’s GPT-5, Microsoft Copilot, Google Gemini, Perplexity, and xAI’s Grok, were put to the test using questions drawn from rare sources such as the White House projectionist logs.
GPT-5 came up empty. Copilot guessed wrong, even in its 3,500-word “Deep Research” mode. Perplexity made the same incorrect guess, Gemini offered no answer, while Grok got it right only by pulling the information from the author’s own 2019 tweet.
These missteps highlight an important reality: AI frequently guesses, misinterprets, or relies on incomplete information. Despite bold promises from tech CEOs, AI tools remain unreliable for niche or archival research.
As the author points out, “These tools still need human babysitters because they get so many things wrong so frequently.”
These limitations of AI sparked a broader discussion about how emerging AI tools will reshape careers and opportunities for younger generations.
OpenAI CEO Sam Altman painted a strikingly optimistic picture for Generation Z in a recent podcast, calling today’s young people “the luckiest kid in all of history,” even as he acknowledged looming job displacement from AI.
Altman said, “this always happens, and young people are the best at adapting to this. I’m more worried about what it means, not for the 22-year-old, but for the 62-year-old that doesn’t want to go retrain or reskill.”
Altman highlighted how tools like GPT-5 make entrepreneurship and innovation more accessible, allowing individuals to achieve feats once reserved for large teams. He stated that “half of the entry-level white-collar workforce will be replaced by AI” in the next five years but expects new, exciting, and well-paid jobs to emerge.
He also discussed the challenge of distinguishing real from AI-generated media, noting that society has always adapted to gradual shifts in technology: “the threshold for how real does it have to be to consider to be real will just keep moving … media is always a little bit real and a little bit not real.”
Altman urged people to use AI tools, saying humility and adaptability are essential in a fast-changing world. “Just using the tools really helps,” he said.
While AI opens new doors for innovation, it also shows the importance of reliability, especially when users depend on it for accurate information.
Even the most advanced AI models, like large language models (LLMs), aren’t always truthful. They can give answers that sound confident but are actually wrong. This problem is often called hallucination, but researchers say it goes beyond just making things up.
Jaime Fernández Fisac, an assistant professor at Princeton, and his team call this tendency “machine bullshit.” It’s not just about making things up, but AI can mislead through half-truths, vague answers, or even flattery that sounds convincing but isn’t accurate.
The researchers also found that common training methods, like reinforcement learning designed to make AI more engaging, can actually make the problem worse. To tackle this, Fernández Fisac’s team created a “Bullshit Index,” a tool to track how often AI misleads users. By measuring these patterns, developers can refine AI training and reduce misinformation.
As AI becomes part of everyday life, tools like the Bullshit Index are meant to keep models honest, helping users get answers that are not just convincing but actually reliable.
Even though AI promises convenience, tools still struggle to offer real-world comfort.
Mozilla’s latest Firefox update, version 141.0, brought a new AI-powered feature that automatically groups similar tabs and even suggests names. However, not everyone was happy; some users reported major CPU spikes, with one noting usage jumping from 0.05% to 130%.
Mozilla calls this “abnormal CPU spikes” and says it shouldn’t happen under normal use. While the feature is still experimental and can be turned off manually, it is rolling out gradually. Although it’s meant to make browsing easier, some frustrated users see it as unnecessary bloat, showing that AI in everyday tools isn’t always smooth, or necessary.
Do you think AI companions can really replace human connection? Can we trust AI tools more than human experts?
Let us know what you think in the comments section below!
First published on Thu, Aug 14, 2025
Enjoyed what you've read so far? Great news - there's more to explore!
Stay up to date with the latest news, a vast collection of tech articles including introductory guides, product reviews, trends and more, thought-provoking interviews, hottest AI blogs and entertaining tech memes.
Plus, get access to branded insights such as informative white papers, intriguing case studies, in-depth reports, enlightening videos and exciting events and webinars from industry-leading global brands.
Dive into TechDogs' treasure trove today and Know Your World of technology!
Disclaimer - Reference to any specific product, software or entity does not constitute an endorsement or recommendation by TechDogs nor should any data or content published be relied upon. The views expressed by TechDogs' members and guests are their own and their appearance on our site does not imply an endorsement of them or any entity they represent. Views and opinions expressed by TechDogs' Authors are those of the Authors and do not necessarily reflect the view of TechDogs or any of its officials. While we aim to provide valuable and helpful information, some content on TechDogs' site may not have been thoroughly reviewed for every detail or aspect. We encourage users to verify any information independently where necessary.
Loading comments...