For most of its two decades, the search bar at the top of Facebook has been one of the least useful tools on the internet. On Monday, Meta Platforms turned it into the company’s most important new product surface. AI Mode, rolling out to Facebook users, lets people ask questions and receive summarized answers drawn from the billions of public posts, groups and Reels on the platform.
Instead of showing a list of links, Meta AI will answer questions directly, with sources cited from public Facebook content, the company said. A user can ask what restaurants people in their city recommend, what the reaction to a product launch has been or what groups are saying about a news event, and the model will synthesize an answer from the posts. The feature is a search engine for the social graph, and it has no direct equivalent on any other major platform.
Analysts see a large prize attached to a simple interface. Forbes reported Monday that analysts project the AI search tool could generate $10 billion in annual revenue for Meta as it begins showing ads alongside AI answers, a figure that would make AI Mode one of the company’s fastest-growing businesses within a few years. The revenue would come from the same machine that already powers Facebook’s ad business: attention, data and placement.
The launch is the clearest sign yet of how Mark Zuckerberg intends to pay for the company’s artificial-intelligence ambitions. Meta has disclosed a capital-expenditure plan that analysts estimate at $380 billion, spending aimed at building the data centers, chips and models that AI products require. The market has treated that number with suspicion, and Monday’s launch is Meta’s answer: here is what the money is for.
The company also added AI editing tools and a virtual try-on feature for clothing ads, moves designed to show advertisers that the AI can do more than answer questions. AI editing lets creators polish images and video in the feed, and the try-on feature lets shoppers see clothes on a model of themselves. Both are aimed at the same goal: more content, more engagement and more ad impressions.
The hiring of Alexandr Wang, the chief executive of Scale AI, to help build Meta’s models underscored the shift. Wang is one of the most sought-after figures in AI infrastructure, and his arrival at Meta signaled that the company intends to build frontier models in-house rather than rent them from rivals. Zuckerberg’s harder task now is selling the output, and AI Mode is the first serious test of whether those models can make money.
The competitive stakes are considerable. Google has dominated search advertising for two decades, and TikTok has been eating into the attention that used to flow through Facebook and Instagram. AI Mode gives Meta a product that sits between the two: the utility of search with the social data that neither Google nor TikTok can match. If users begin treating Facebook as a place to ask questions, the ad inventory that follows could be enormous.
There are reasons for caution. AI answers can be wrong, and a social network that summarizes its own posts will inherit every rumor and conspiracy circulating in its groups. Meta has promised to cite sources and let users report bad answers, but moderation at this scale has burned the company before. European regulators are also watching, and the feature’s use of personal data to generate answers will draw scrutiny under privacy rules that have already cost Meta billions.
The financial stakes are simpler to calculate. Meta’s core ad business is a machine that generates tens of billions in quarterly revenue, and every new surface for ads on Facebook and Instagram compounds it. If AI Mode reaches even a fraction of the company’s three billion users and keeps them engaged, the $10 billion revenue estimate will start to look modest. If it fails to catch on, it becomes another costly experiment in a company that can afford experiments.
The mechanics of monetization are still being worked out, but the direction is clear. Meta is expected to follow the same playbook Google used for its AI search summaries: show an answer, then show sponsored placements around it, with advertisers bidding on the questions users ask. The difference, analysts note, is that Google’s AI answers draw on the open web while Meta’s draw on the social graph, which means the ads can be targeted with the platform’s own data rather than inferred intent.
Regulators are a wild card. The feature’s reliance on personal data will face scrutiny in Europe, where Meta has already been fined billions for privacy violations, and the company’s decision to launch in the U.S. first reflects that calculation. For all the revenue potential analysts see in AI Mode, the product will succeed or fail on whether users actually ask Facebook questions, a habit that two decades of search bar neglect did nothing to build.
For Zuckerberg, the pattern is familiar. He has spent the past two years telling investors that AI would reshape Meta, and the market has rewarded the promise while questioning the price. AI Mode is the first product that tries to turn the promise into revenue, and the reception over the next few months will tell Wall Street whether the biggest capital plan in the company’s history is building a business or a museum.
