Peak social? The data says… probably

• GWI data (via the FT) shows global time on social peaked in 2022 and was ~10% lower by end-2024; the drop is sharpest in teens/20-somethings.

• At the same time, platforms are doubling down on AI-generated video (OpenAI’s Sora app; Meta’s push into AI video), raising brand-safety and quality questions.

• Europe looks notably cooler on social time than the global average (EU+UK ~1h48m/day vs 2h23m globally).

What happened

On 3 October 2025, the FT argued we may have passed “peak social media”, citing GWI’s multi-year panel showing a smooth rise then decline in daily social time since 2014, with usage in 2024 almost 10% below 2022. This lands the same week OpenAI launched Sora, an invite-only social video app for short AI-generated clips, while platforms (including Meta) lean hard into AI video formats.

Context & data

• Time spent: The “typical” user spent 2h23m/day on social in 2024 (GWI), with EU+UK at 1h48m, well below the global average.

• The peak: GWI’s long-running tracker shows a post-2022 decline in global social time, especially among younger cohorts, per the FT’s analysis. Note: GWI updated its time-spent methodology in Q4-2024, so trend interpretation needs care.

• UK angle: UK adults spent 4h20m/day online across devices in May 2024 (not just social), with 18–24s at ~6h/day—suggesting overall screen time is resilient even if social time cools.

• Product moves: OpenAI’s Sora (30 September 2025) adds a TikTok-style feed of AI-generated 10-second videos; Meta is also pushing AI-video creation natively across its surfaces.

My take

If time on social is rolling over while total online time holds up, attention is fragmenting toward utilities (messaging, commerce, CTV/streaming, search/assistants) and away from “feed scrolling”. That’s a headwind for the social ad unit as a pure time-harvest business: lower incremental minutes, noisier feeds (AI slop), and rising brand-safety costs can compress effective CPMs and weaken LTV/CAC for creators and SMB advertisers. The strategic bet from platforms—flood the zone with creation tools and AI-video—is logical to prop up DAU/time-spent, but risks a quality → engagement → monetisation spiral if feeds degrade.

From an investment perspective, I’d underwrite mix-shift risk rather than a cliff. Dollars aren’t leaving digital; they’re rotating into retail media and CTV, where signal quality and closed-loop attribution are stronger. For social-native assets, I’d look for moats in (i) verified identity/consent layers for AI remixes, (ii) superior trust & safety tooling that protects brand adjacency (higher CPMs, better take-rates), and (iii) commerce surfaces with real conversion (not just view time). If management can show stable or improving revenue per minute (RPM) as minutes plateau, there’s still a case—otherwise multiple compression feels rational.

Risks & watch-list

• Measurement noise: GWI changed questions in Q4-2024; trend lines need normalisation before making hard calls on “peak”.

• Regulation & reputational risk: Deepfakes/AI content raise legal and brand-safety exposure; any high-profile incident can hit ad fill and pricing.

• Regional divergence: North America remains an outlier with rising usage; if that persists, global “peak” becomes a Europe-led story rather than universal.