Meta is drastically restructuring how it manages two of its biggest pain points: user support and content moderation. In an announcement timed for a global rollout, the company revealed that “Meta AI” is no longer just a creative chatbot but is now being integrated as a functional support assistant across Facebook and Instagram. Simultaneously, the company is shifting its moderation strategy to lean more heavily on advanced AI systems, aiming to reduce human exposure to graphic content while increasing the accuracy of scam detection.
However, given the many complaints users have generally had with their AI support system systems for business accounts and ad campaign setup, many are very skeptical of how this will in actual work.
The most immediate change for users is the Meta AI support assistant. Unlike previous iterations of “Help Centers” that relied on keyword searches and static articles, this new tool is designed to take direct action within the app. Meta claims the assistant can respond to most queries in under five seconds, a significant benchmark for a platform with billions of users.
Beyond answering questions about privacy settings or new features, the assistant is being granted permission to execute account-level changes. Users can now ask the AI to reset passwords, update profile settings, and manage notification toggles directly. Crucially, it also handles the “paperwork” of moderation, allowing users to report impersonation accounts and track the status of content appeals through a conversational interface rather than a series of nested menus.
On the back end, Meta is deploying more sophisticated AI models to handle content enforcement, moving away from a primary reliance on third-party human review teams. The company shared several data points from early testing that suggest these models are outperforming human moderators in specific, high-velocity areas:
- Scam Mitigation: The AI currently identifies approximately 5,000 scam attempts per day that were previously missed by human reviewers.
- Impersonation: Meta reports an 80% reduction in user reports regarding celebrity impersonation since the AI began flagging these accounts at the point of creation.
- Contextual Awareness: The system is now capable of identifying “account takeovers” by analyzing clusters of behavior—such as a password change combined with a new login location and profile edits—that might look harmless to a human reviewer when viewed in isolation.
One of the most significant technical leaps is in language support. While previous systems covered roughly 80 languages, the new AI models are trained to understand cultural nuances, slang, and regionally specific code words across languages spoken by 98% of the online population.
The company is also targeting “spoofing” more aggressively. In one test case, the AI was able to detect fake retail sites by identifying the use of legitimate logos paired with suspicious web addresses and “unusually low” pricing. This specific application reportedly drove down views of scam-related ads by 7%.
You might also like
More from Apps
Mzansi’s online grocery sector… the trends, challenges and approaches
As the country undergoes a digital transformation, the online grocery delivery sector is experiencing revolutionary changes. The shift from in-person …
All PSL Matches Now on Showmax For R69
The Premier Soccer League (PSL) has partnered with SuperSport to bring all PSL matches, including the prestigious Betway Premiership, to …
DStv Stream contract customers can now save R100 p/m
Yes you read that correctly, DStv has today announced a reduced price for Its Stream Premium contract customers, this is …
