Hands-on with the Meta AI chatbot: fails at basic search queries, stinks at counting, excels at editing existing paragraphs, quickly creates images, and more
Despite Mark Zuckerberg's hope for the chatbot to be the smartest, it struggles with facts, numbers and web search.
Context & Ripple Effects
Meta had already positioned its assistant across WhatsApp, Messenger and Instagram after its earlier rollout alongside celebrity-based AI characters, following BlenderBot experiments that used public feedback to test conversational and web-search capabilities. This review shows the gap between broad distribution and dependable general-purpose assistance.
The product's strength in rewriting and image generation contrasts with weak factual, numerical and search performance. That distinction matters as Meta moves from a single assistant toward user-created AI chatbots on Instagram, where quality control becomes harder to centralize.
First-order effects
- Users can get fast help editing text and creating images, but must verify answers involving facts, counts, or web-derived information before relying on them.
- Meta's claim to offer a leading assistant is immediately constrained by visible failures in core query handling, even as it distributes the product through its social apps.
Second-order effects
- Meta has reason to steer usage toward bounded creative tasks, where the review finds clearer utility, rather than present search-like answers as inherently trustworthy.
- As Meta enables custom chatbots for creators, inconsistent reliability in the base assistant raises the importance of labeling, testing, and setting expectations for AI-generated responses.
Third-order effects
- The episode points to a split in consumer AI: assistants may gain adoption first as embedded creation tools, while factual retrieval remains a credibility bottleneck.
- If platforms keep expanding customizable assistants before reliability improves, differentiation will increasingly depend on safeguards and task-specific performance rather than personality or distribution alone.
The trend: Consumer AI platforms are shifting from novelty chatbots toward embedded, customizable assistants whose value depends on proving reliability in the tasks users delegate to them.