The Assistants: Where Attention Might Go NextNarrow moat
Meta Platforms (META) — moat facet
The first serious claim on attention that has no feed to copy, and therefore no format Meta can fast-follow.
Every previous competitor for Meta's users offered a variation on the same thing: a stream of content, ranked by an algorithm, interrupted by ads. Meta beat most of them by copying the format and deploying it at a scale nobody else had. AI assistants are the first serious claim on attention that does not work that way at all.
The threat is not that people stop using Facebook or Instagram. It is that a meaningful share of the time and intent that currently flows through a feed — looking things up, deciding what to buy, asking for recommendations — moves into a conversation with no advertising surface. Meta's monetisation depends on impressions: it delivered 12% more of them in 20251. A destination that serves one answer instead of a scrolling column of them generates almost none.
Meta's position is better than it looks. It has distribution no assistant can match, it is spending on the scale described in the Future Bets, and conversational interfaces have so far proved additive to media consumption rather than substitutive. The company's own AI is embedded across apps that 3.58 billion people use daily.
Watch time spent per user rather than user counts. Users leave slowly and time leaves fast, and time is what the auction actually sells. A sustained decline in engagement while daily users hold steady would be the earliest evidence that attention has found a new home.
Assistants have grown enormously without measurably reducing time spent in Meta's apps, which is the only thing that would matter. Meta is also spending heavily enough to be a participant rather than a bystander. Stable for now — with the honest caveat that this is the first competitor in a decade whose format Meta cannot simply copy into a feed.
Meta monetises impressions; an assistant that returns one answer instead of a scrolling column generates almost none. So far assistants have proved additive to media consumption rather than substitutive. Watch time spent per user rather than user counts — users leave slowly and time leaves fast.
Source: Meta Form 10-K, FY2025 ↗- ReportedMeta delivered 12% more ad impressions in 2025, and Family daily active people reached 3.58 billion.Meta Form 10-K, FY2025 — Family daily active people 3.58 billion on average for December 2025 (+7%); ad impressions delivered across the Family of Apps increased 12% year over year in 2025; average price per ad increased 9%; total revenue $200,966M, with revenue disaggregated by customer address of United States and Canada $78,866M, Europe $46,569M, Asia-Pacific $53,817M and Rest of World $21,714M; United States revenue $74.78 billion — FY2025 (ended December 31, 2025) · publ. January 29, 2026 · source ↗