What do ads on ChatGPT tell us about the AI race?

Admatic Comment: This is really interesting by Hugh Linehan. AI searches are becoming much more common notably by younger users and a consequence of that, will be to reduce clicks on paid Search for Google (such as PPC). Open AI is a Google threat to Google’s search advertising monopoly. And now Open AI is taking ads…..that’s a higher threat. Be aware too, that Google Search is not going to perform as good as, because of AI searching. 

 

From Irish Times:

In May 2024, speaking at Harvard University, OpenAI chief executive Sam Altman described the prospect of combining advertising with artificial intelligence as “uniquely unsettling”. He went further, characterising ads as “a last resort” for his company’s business model. “I kind of hate ads just as an aesthetic choice,” he said on another occasion.

Twenty months later, that last resort has arrived.

Last week, OpenAI announced that US users of ChatGPT’s free tier, along with subscribers to its new $8-a-month ChatGPT Go plan, will soon find contextual advertising appearing at the bottom of their queries. The ads will be “clearly labelled”, the company promises, and will “never” influence ChatGPT’s responses. Your conversations will remain private from advertisers.

More troubling is the company’s plan to leverage ChatGPT’s “memories” feature – which stores information from conversations to provide personalised responses – to deliver hyper-targeted advertising. According to the Financial Times, this tranche of user preferences represents a powerful resource for advertisers. OpenAI says personalisation can be turned off. If you believe that will solve potential problems, I have a bridge in London you might like to buy.

Anyone familiar with Cory Doctorow’s theory of enshittification will see this as inevitable. The writer has documented how digital platforms follow a predictable doom loop: innovation, customer acquisition, market dominance and degradation of the service in the interest of maximising profit. Just as Google went from placing ads in a right-hand column to cramming them above organic search results – quietly dropping its “don’t be evil” motto along the way – so too will ChatGPT probably evolve from helpful assistant to hawker of goods.

Others will wonder whether this represents an early warning sign about the AI industry’s underlying economics. If the world’s best-known AI platform needs to resort to advertising, despite 800 million weekly users, is that evidence that cracks are appearing in the eye-wateringly optimistic assumptions underpinning the AI investment boom?

OpenAI has committed to spending roughly $1.4 trillion (€1.2 trillion) on computing infrastructure over the next decade. Despite annual revenue reaching $20 billion in 2025, it faces operational expenses that dwarf its income. The company expects cumulative losses of $143 billion before achieving profitability around 2029. One analyst at Deutsche Bank put it bluntly: “No start-up in history has operated with losses on anything approaching this scale.”

Competitors have been quick to distance themselves. Google’s DeepMind CEO Demis Hassabis, at Davos last week, said he was “a little bit surprised they’ve moved so early into that”. While acknowledging that ads “funded much of the consumer internet”, Hassabis raised a fundamental question: “In the realm of assistants … there is a question about how ads fit into that model. You want to have trust in your assistant, so how does that work?”

Using a chatbot, Hassabis argued, is very different from search. Chatbots are meant to become digital assistants that know about you and can help with many aspects of your life. That requires a fundamentally different relationship – one that advertising will almost certainly corrode.

We should take these expressions of ethical concern with a large pinch of salt. Google, as one of the world’s most profitable companies, can afford to subsidise its own AI service, Gemini, through its existing advertising empire. After being rocked by the launch of ChatGPT three years ago, it has regained ground in the AI race and looks well placed to integrate AI into its search, email and other services in a host of lucrative ways.

Some analysts suggest OpenAI could achieve ad revenues of $25 billion by 2030. Competitors will not ignore that opportunity forever. As hundreds of millions of users replace traditional search with AI for everything from price comparisons to travel planning, Google’s “principled” stance will undoubtedly shift accordingly. The company’s vice-president of global ads was notably careful to say there are “no current plans” for Gemini ads, leaving plenty of wriggle room.

It would have been naive to expect AI’s free tiers would remain genuinely free forever. For users, the old adage remains as relevant as ever: if you’re not paying for the product, you are the product. That’s been true of digital platforms for decades.

The promise of AI assistants was that they would provide unbiased, helpful answers tailored to your needs. Advertising fundamentally compromises that proposition. Messages placed beside personalised guidance are likely to be more influential than the same ads seen while browsing a webpage. When OpenAI positions ChatGPT as a “super assistant” for everything from health queries to financial advice, the line between guidance and persuasion disappears.

Altman, in announcing the change, now says he “loves Instagram ads” and has “found stuff I never would have found” through targeted advertising. This is quite the journey from describing AI advertising as uniquely unsettling. The shift reflects the brutal economics of the AI race along with a warning about the sustainability of the entire enterprise.

History offers a template. Social media platforms spent years building trust before pivoting to aggressive monetisation. Netflix famously insisted it would never run advertisements, then introduced an ad-supported tier. The tech industry’s promises about user experience tend to have a shelf life that inevitably carries an expiration date.

OpenAI insists it will maintain guardrails. Ads won’t appear near topics such as health, mental health or politics. Users under 18 won’t see them. The company will “prioritise user trust over revenue”. These are the same kinds of assurances we heard from social media platforms two decades ago, before algorithmic engagement took over and the dopamine economy swamped public discourse.

Some thought AI would be different. That illusion lasted about three years.