Chatbots may hallucinate and sprinkle too flattering on users. “That’s an attractive question!” Someone recently told me – but at least the subscription model that supports them is healthy for our well-being. Many of us pay around 20 USD per month to use Openai’s ChatGpt, Google’s Gemini Pro, or the premium version of Claude of Humanity. As a result, the product is designed to provide the largest utility.
Don’t expect this situation to continue. There are limits to subscription revenue, and Anthropic’s new $200$200 “Max” tier suggests that even the most popular models are under pressure to find new revenue streams.
Unfortunately, the most obvious is advertising. This is the most successful business model on the web. AI builders are already exploring ways to connect more ads to their products, which is suitable for the final line, but means they are looking to see a new chapter in the attention economy that has promoted the internet.
Researchers at Cambridge University called this a future “intentional economy” in a recent paper
If the descent to social media engagement baits is a guide, the outcome is profound.
One cost is addiction. A recent study shows that young office workers rely on AI tools to help them write emails and digest long documents, while Openai says that a cohort of “problematic” ChatGPT users are obsessed with the tool. Advertisements in ChatGpt, which currently has over 500 million active users, will not spur the company to help businesses reduce product use. It’s exactly the opposite.
Advertising is why companies like Mark Zuckerberg’s metaplatform designed algorithms to drive engagement, designed to keep users looking more ads and driving more revenue. That’s the reason behind what is called “enshittification” on the web. This is where ClickBait and social media posts are now filled and is causing anger. Burning such incentives into AI will almost certainly help designers find ways to cause more dopamine spikes. Perhaps it complements users further, asks personal questions, speaks longer, or even develops emotional attachment.
Beguile
Millions of people in the Western world see chatbots as friends and romantic partners on apps such as characters, chai, talkie, replicas, and botii. Imagine how convincing such software is when users are captivated. Imagine someone who says they feel AI is depressed and a system that recommends affordable holiday destinations and medications to address the issue.
That’s how ads work in chatbots? The answer is the subject of many experiments, and companies are actually experimenting. For example, Google’s ad network recently began posting ads in third-party chatbots. A romance and friendship chatbot, Chai spends an average of 72 minutes a day serving pop-up ads in September 2024. The AI answer engine Perplexity displays sponsor questions. For example, after answering a job search question, “How can I actually use it to enhance my job search?”
Read: Google’s AI is personal, aggressive and premium
Aravind Srinivas, CEO of Perplexity, told a podcast in April that the company is aiming to enable what it called “hyper personalized” ads by building a browser and tracking what it calls “what hotels and which restaurants go” to “get data even outside the app.”
For some apps, that may mean weaving the ads directly into the conversation. You can also use intimate details that users share to predict them, manipulate them to want something, and sell their intentions to the highest bidder. Researchers at Cambridge University called it a future “intentional economy” in their recent paper, where chatbots piloted conversations for brands and direct sales. As evidence, they pointed out Openai’s 2023 blog post seeking “data expressing human intentions” to train models, Meta’s similar efforts, and Apple’s 2024 developer framework to help work with SIRI to “predict what actions someone might take in the future.”
As for Openai’s Sam Altman, there’s nothing to say that he’s “building an advertising business” that would hire people who built the delivery app Instacart in an ad powerhouse. Altman recently poached CEO Fidji Simo and helped Openai “scale as it enters the next stage of growth.” In Silicon Valley terminology, “scaling” often means quickly expanding your user base by offering free services in advertising.
High-tech companies inevitably argue that advertising is a necessary part of democratizing AI. But we have seen how “free” services sacrifice privacy and autonomy for people, and even mental health. And AI knows more about us than Google or Facebook. Details about health concerns, relationship issues and work. In two years they have also built a reputation as trustworthy peers and true arbitrators. For example, in X, users frequently confuse AI models with grok and if the post is fake, they permeate the conversation with flags.
Read: Microsoft Adopts AI Diversity
When people trust AI that much, they are more vulnerable to targeted manipulation. AI ads should be regulated before it becomes too ingrained, or repeat mistakes made on social media. Only after damages occur can we scrutinize the fallout of a favorable business model. – (c) 2025 Bloomberg LP
Get breaking news from WhatsApp’s TechCentral. Sign up here.
Don’t miss it:
Drugs generated by the first AI could be sold by 2030
