Instagram chief Adam Mosseri is pushing back against the idea that artificial intelligence will hollow out the creator economy, in response to recent alarms sounded by YouTube star MrBeast. Addressing Bloomberg’s Screentime conference, Mosseri defended his idea that generative tools are more likely to expand the creator pool and increase production quality than supplant popular creators. At the same time, he admitted to a harsher reality: Society will have to adopt new habits in order to live with an internet where video alone is no longer proof of reality.
AI will widen the circle of creators, not replace them
Mosseri’s primary thesis is that AI extracts the financial value from content, in a similar way to how the internet has removed costs for distributing it. That turn could pave the way for people who have ideas and taste but not teams, budgets or technical chops. He cited the widespread use of “hybrid” workflows — color correction, clean-up, dubbing or rapid editing — where AI adds to human production rather than relying on what AI outputs from scratch.
For top-tier entertainers whose brand depends on scale, stunts or access — MrBeast is the canonical example — Mosseri argued that AI is a force multiplier and not a clone machine. It can aid in localization, personalization and iteration, but it won’t immediately fabricate the logistics, relationships or creative judgment that support blockbuster channels. That perspective is consistent with findings from the Reuters Institute that creators who lean into unique formats and audience communities are more resistant to algorithmic churn than those that compete on commodity clips.
Yet the industrialization of creativity is a thing. Generative video models and voice tools are improving rapidly and falling in cost, and some formats will get saturated. Taste, curation and narrative — places where AI helps but humans still set the brief — are the opportunity, Mosseri suggested.
It’s still messy to accurately label synthetic media
Mosseri acknowledged that Meta has struggled to effectively label AI-generated content. Early attempts to tag content automatically resulted in false positives when creators made videos and photographs using standard editing software that embeds provenance signals. Industry initiatives such as the C2PA (which is less focused on deepfakes) or Adobe’s Content Credentials can help but cannot cover every workflow or adversarial case, and not all tools attach trustworthy metadata.
He did not promise a silver bullet; he called for additional context to allow people to judge what they see. That could look like explicit labels when provenance information is available, visible disclosures from creators and crowdsourced context about questionable posts — a model pursued in Community Notes and used as the subject of academic research on misinformation. Metadata generation and usage (among others, Google’s SynthID watermarking) represent other promising tools, but issues related to the robustness of watermarks or cross-platform adoption are still open technical challenges.
The stakes are evident. The Reuters Institute Digital News Report finds that a majority worry about false information online, and Pew Research Center surveys show more Americans are concerned than excited about the impact of AI on daily life. Regulators in the U.S. and Europe have raised alarms over deepfakes aimed at voters; watchdogs like Sensity have observed steep climbs in the volume of synthetic media. Accuracy and explainability aren’t just features of the platform — they’re categories of trust infrastructure.
Media literacy as a basic digital survival skill
Mosseri stressed that the change is as much about a cultural shift as it is a technical one. Children — and adults — will need to approach videos less as recordings of events than as claims that must be defended or debunked.
- Who posted it?
- What is their track record?
- What incentives are at play?
These are the questions that echo guidance from organizations like the Stanford History Education Group, which has found that lateral reading and source vetting can drastically improve our ability to assess information online.
Default platform cues and educational campaigns can help minimize cognitive load. UX research also indicates that simple, consistent badges and short, plain-language explanations are more effective than complex disclosure pages. Then there’s a policy context: consumer-protection regulators have made clear that the failure to disclose synthetic endorsements or media that is materially altered could lead to enforcement, meaning standardized labeling is not just good hygiene but compliance.
Creator economics in the AI-native epoch
Mosseri is optimistic for a simple creator-economy reason: When production costs decrease, distribution gets noisier — but the best creators iterate quickly. We’re going to see more multilingual versions of videos, quicker A/B testing of thumbnails and scripts, and niche-community-specific cuts. Brands are already toying with synthetic voices for localization and AI-driven post-production in order to cut costs — which favors creators who can wield these abstractions with taste and retain the trust of their audiences.
None of that erases the risk or the downside. Generative tools make imitation and scam content more affordable. Platforms will require better reporting flows, higher-friction verification for high-reach accounts and incentives to accurately self-disclose. The potential payoff is huge: If audiences are convinced they’ll be able to detect what’s synthetic, they’re more likely to participate; if not, skepticism becomes the default, and that hurts all of us.
Competition and product direction at Instagram
Outside AI, Mosseri said Instagram’s short-term priorities mirror where users are spending time: short-form video and private messaging. A dedicated TV app is still in the works as viewing moves to living rooms. He also made the case that stout competition in short video has spurred Instagram to ship faster and iterate on ranking systems — an implicit allusion to how product velocity, not only policy, will decide whether creators stick around as the AI era remolds formats.
The throughline in Mosseri’s comments is practical: AI won’t level the creator landscape overnight, but it will collapse production costs, stress-test trust systems and raise the bar for media literacy. Platforms can cushion the transition with clearer labels and context; creators can compete on origination and relationship; and audiences will have to develop new senses for judging what is real. That’s not apocalyptic — it’s a recalibration.
