Your AI Assistant Is an Advertiser’s Wet Dream. Should You Be Worried?
Your AI assistant knows what you want and what might change your mind. Who will pay more for that knowledge—you or an advertiser?
Say you ask your AI assistant to find a washing machine under ₹40,000. It knows how many people live in your house, how much space you have, and that your last machine sounded like it was preparing for take-off.
It compares the options and recommends three. You buy one of them.
Now imagine finding out that this recommendation was actually an ad from the highest bidder. It wasn’t the best option for you. It was the option that worked better for the company behind your assistant and the advertiser.
That leaves me with a few questions:
- Is the assistant choosing what is best for me, or what it gets paid to recommend?
- If it recommends only three washing machines, how will I ever discover a new brand outside those three?
- If brands and sellers pay to be recommended, will those three places go to whoever pays the most?
- What if I enjoy looking through the options myself, like choosing clothes or a holiday home?
- If I pay a subscription, will the assistant work only for me? Or will advertisers still get a say?
There are two different things here: showing an ad beside your assistant’s advice, and paying to influence the advice itself. The opening example is about the second.
To explore these questions, let’s first look at what advertisers pay for today.
Why advertising makes so much money
The same person can be worth very different amounts to an advertiser. Someone watching reels may discover a product they had never considered. Someone searching for an emergency plumber could become a customer within minutes. What they are doing changes the value of reaching them.
An assistant could show fewer ads and still earn substantial revenue if those opportunities lead to more business. Knowing your constraints and helping you reach a decision could make access to you valuable.
Advertisers then compete for that access. The highest bid does not automatically win: Google’s auction also considers quality and context. Better conversion gives an advertiser room to bid more; competition helps determine how much the platform actually earns.
The scale of the existing businesses shows why this opportunity attracts so much interest:
| Advertising business | Q3 2026 revenue | Growth over Q3 2025 |
|---|---|---|
| Google advertising | $82.01 billion | 10.6% |
| Meta advertising | $63.06 billion | 25.9% |
| Amazon advertising services | $19.64 billion | 10.9% |
Sources: Alphabet · Meta · Amazon. US dollars; growth calculated from reported revenues. Google’s total includes Search, YouTube ads, and Google Network.
Meta’s results show both sides of the model at work: impressions grew 10%, while average price per ad rose 14%.
Google is already dealing with the transition
Google’s advertising revenue has not collapsed. It grew year over year in its latest quarter. What slowed was the rate of growth.
On Alphabet’s Q3 2026 earnings call, management acknowledged that increased usage of AI Mode affected Search growth because AI Mode monetized at a lower rate. Foreign exchange was the largest driver of the slowdown from Q2, and advertising changes in particular industries also contributed.
Google is moving users into an experience that currently earns less from advertising.
Google plans to close that gap by expanding advertising in AI Mode. Whether it can do that is worth watching.
What happens to the brand that comes fifth?
Several brands may make a washing machine that fits your needs. Many sellers may offer those machines. Your assistant has to decide both what you should buy and who you should buy it from.
A new manufacturer needs its product considered. A retailer selling the same machine needs its offer chosen.
On a marketplace, coming fifth still gives you a chance. When an agent makes the purchase, coming second may mean you don’t exist.
Being evaluated by software is different from being discovered by a person. Even a purchase you do not make today can introduce you to a brand you remember tomorrow.
That is where bidding could become more intense. If the assistant sells access to a much smaller set of recommendations, eligible sellers have fewer places to compete. And if a recommendation is more likely to become a purchase, they may be willing to pay more for it.
Not everyone has to win for the platform to make money. Scarcity can make the winning position more valuable.
A seller still cannot sustainably pay more to acquire customers than those customers are worth. Higher prices per recommendation also have to compensate for fewer opportunities overall.
Different budgets, locations and preferences can produce different winners. A new brand could benefit if the assistant finds a better product that you would never have noticed yourself. But that depends on how the assistant decides which products to consider—and whether payment affects who gets considered.
Nor does every assistant have to return a single answer. ChatGPT’s shopping experience already includes visual browsing and side-by-side comparisons. The more we delegate the final decision, the stronger this concern becomes.
What exactly is the advertiser paying for?
Back to the washing machine. An agent could ask merchants to improve their offers for you. Or its owner could charge merchants for influence over the recommendation. The difference becomes obvious when you follow the money.
A transaction fee could also pay for completing a purchase. But would the assistant still consider merchants that pay it nothing?
The difficulty is that you hired the assistant to do the comparison. If you have to repeat that work to check whether its advice was commercially influenced, part of the convenience disappears.
The more decisions we delegate, the more we depend on the assistant’s incentives. A small “sponsored” label can identify an ad. Establishing that money did not affect the surrounding advice is a harder task.
Companies are taking different positions. Anthropic’s explanation for keeping Claude ad-free argues that advertising can create pressure to steer conversations toward transactions and longer engagement.
OpenAI describes a separate advertising system designed to keep ads from influencing ChatGPT’s answers. In October 2026, it reported reaching $1 billion in annualized advertising revenue run rate less than 200 days after launch.
That is a company-reported run rate, not a full year of revenue or proof of profitability. It shows why AI advertising cannot be dismissed, although ads in a chat window are different from an assistant making purchases in the background.
And even if you find out that advertisers influence your assistant, will you stop using it? What if it already knows your preferences, manages your tasks, and saves you hours every week? Losing some trust may not mean losing the customer. How much influence will people tolerate before the convenience stops being worth it?
Sometimes browsing is what you came to do
There is another assumption worth challenging: that the shortest path to a transaction is always the best experience.
I might not even know which kind of holiday I want until I see the homes. Comparing them helps me decide what I want; it is not simply a delay before checkout.
Consumer research has long distinguished practical and enjoyable aspects of shopping. The same person may delegate a work trip and enjoy spending an evening choosing a family holiday.
The same problem appears when we treat scrolling through reels as an inefficient way to reach an outcome. Watching and discovering are often the activity itself. An agent summarizing the evening’s entertainment may have solved a problem the user did not have.
In its Q3 2026 shareholder letter, Airbnb reports that more than 75% of guests use maps when choosing where to stay, over 90% browse multiple listing photos, and 70% read reviews before booking.
Those numbers show how people choose a stay; they do not tell us how much of that browsing they enjoy. Airbnb’s proposed agent combines maps, photos, comparisons, calendars, and itineraries, with rollout planned to begin in the first half of 2027.
This corrects another assumption: agents do not have to be screenless. They can work inside a visual experience, handling the tedious parts while leaving you in control of the enjoyable ones.
That could preserve much of today’s browsing behaviour. It could also leave established platforms with a significant advantage: they already have the inventory, the audience, and the experience people return to.
YouTube makes the subscription argument harder to dismiss
If advertising creates these tensions, perhaps users will pay. YouTube shows how substantial that business can become.
Sources: Alphabet’s 2025 annual filing · Q4 earnings release. The subscription figure is inferred, not separately disclosed; it also includes services such as NFL Sunday Ticket.
YouTube shows that advertising and subscriptions can both bring in serious money. But these subscription revenues include TV, Music and Premium; they are not all payments to remove ads.
On YouTube, paying can remove the interruption. With an assistant, the question goes further: does paying also remove the advertiser’s influence? An ad-free screen and independent advice are two different promises.
Assistants could offer free access supported by ads, paid access without ads, or subscriptions with more capability and higher usage limits. They do not have to use one business model for everyone.
But the fee still has to cover the work. One user asks a few questions. Another delegates hours of research, browsing, and document creation. More usage can mean more computing and tool costs.
Can subscriptions cover that work? And if the company also earns commissions from purchases, will paying users still get recommendations from merchants that pay nothing?
Where does this leave us?
So when your assistant says, “This is the best option for you,” how will you know why it chose that option?
Was it the best product for your needs? Or did the seller pay more to be recommended?
What happened to the new brand that might have offered something better? Did the assistant compare it and reject it, or did it never get a chance?
And what if you wanted to look through the options yourself? Will the assistant help you explore, or keep pushing you towards a purchase?
If you pay a subscription, can you expect it to work only in your interest? Or will you be paying for a service that also sells access to you?
I don’t think we know how this will settle yet. There is money to be made here. The question is whether these companies can make it while building an assistant we continue to trust.
For more on AI, technology and business, follow me on X, LinkedIn and Instagram, or subscribe to my YouTube channel.