How would you enhance the Meta AI chatbot experience within Instagram direct messages?
How to scope and prioritize improvements to Meta AI inside Instagram DMs when users are churning after early adoption.
This post is for our Paid Subscribers. If you haven’t subscribed yet,
Clarifying Questions
Before diving in, I would want to align on scope before jumping into solutions. A few questions I would bring to my stakeholders:
Personal use versus business or creator use? Are we enhancing Meta AI for everyday users chatting with the assistant for personal tasks, or are we focused on creators and business accounts who might use AI to help manage high-volume DM conversations? This changes the entire problem space. For this exercise I will assume personal-slash-creator use, everyday Instagram users interacting with Meta AI in their own DM thread.
Adoption, retention, or quality? Meta AI was launched to broad audiences in 2023, so there is a meaningful cohort of users who have tried it and churned. Are we trying to reactivate lapsed users, deepen engagement with active users, or lift the raw quality of responses for everyone? These require different solution bets. I will assume quality and retention of existing users is the priority, with adoption as a secondary outcome.
Scope of the DM experience? Are we confined to what happens inside the Meta AI thread itself, or can we also redesign the entry points that bring users into the thread in the first place? I will assume both are in scope, since entry points significantly influence the first impression of the product.
Platform priority? Instagram is used on iOS, Android, and desktop browser. Given that over 85 percent of Instagram sessions happen on mobile, I will design mobile-first with the assumption that desktop is a lower-priority adaptation.
Monetisation constraint? Meta AI is not directly monetised today. Should we avoid any solution that introduces a paywall or premium tier, to avoid fragmenting the experience? I will assume we stay free and focus on engagement metrics rather than revenue metrics for this cycle.
For this answer I will assume: personal and creator use, mobile-first Instagram DMs, with a mandate to improve the quality and retention of existing Meta AI users, and full freedom to redesign both entry points and the in-thread experience within the free tier.
Product Description
Meta AI is an AI assistant built on Meta’s own Llama large language model family, first announced at Meta Connect in September 2023 and rolled out progressively across Meta’s family of apps through 2024. It is embedded natively inside Facebook, WhatsApp, Messenger, and Instagram, with a unified identity across all four surfaces. As of early 2025, Meta has stated that Meta AI is used by over 600 million people monthly across its platforms, making it one of the most widely distributed AI assistants in the world by raw user count, even if engagement depth lags behind more intentional tools like ChatGPT, which crossed 100 million weekly active users in 2023.
Within Instagram specifically, Meta AI lives as a dedicated chat thread inside the Direct Messages inbox, accessible via a persistent entry point at the top of the DM list. Users can ask factual questions, request creative help (caption writing, Reels hook ideas, hashtag strategy), generate images using the “Imagine” command, get recommendations for restaurants or travel, and complete general knowledge tasks. Instagram itself has 2 billion monthly active users globally as of 2024, with a demographic skew toward 18 to 34 year olds who represent roughly 60 percent of the active user base. That younger cohort is already comfortable with AI tools and is the primary driver of AI feature adoption across the app.
Meta AI is not directly monetised today. It generates no subscription revenue and carries no in-thread advertising. Its strategic value is indirect: it increases time spent on platform, deepens engagement with Instagram surfaces like Reels, Stories, and Shopping, and positions Meta competitively against Apple Intelligence (announced at WWDC 2024), Google Gemini (integrated into Android), and OpenAI’s ChatGPT. Meta’s long-term thesis, articulated by Mark Zuckerberg publicly, is that Meta AI becomes the leading personal AI assistant by virtue of social graph context and scale, with future monetisation flowing through AI-powered advertising, commerce recommendations, and eventually a premium tier. The core tension today is that Meta AI sits inside one of the world’s most-used social apps but is largely invisible to the average user: most Instagram users open DMs to talk to friends, not to an AI, and Meta AI has to earn its place in that attention context without feeling intrusive.
Define Goal
The problem I want to focus on is this: Meta AI in Instagram DMs has strong distribution but weak habit formation. Users try it, find it generic, and do not return. The challenge is not awareness; it is depth and relevance of the experience.
The goal I want to optimise for is making Meta AI in Instagram DMs feel like a genuinely useful creative partner for creators, not a novelty chatbot, so that they build a repeatable weekly habit of using it.
My north star metric is: Weekly Active Users of Meta AI in DMs who complete at least 3 meaningful exchanges in a single session, where a “meaningful exchange” means a back-and-forth multi-turn interaction rather than a single query with no follow-up. I would track this as a 7-day rolling window.
I chose this metric over simpler alternatives like total sessions or DAU because it captures both habit (weekly return) and quality (multi-turn depth). A user who opens the Meta AI thread once and fires off one question is not building a habit and is not deriving real value. A user who has a 5-turn conversation about their next Reel idea and comes back Thursday to plan their next post is exactly the behaviour we want to scale. This metric is also hard to game with superficial product changes, which makes it a healthy north star. Secondary metrics I would track alongside it include session length in minutes, response satisfaction rate (thumbs up divided by total rated responses), and week-2 retention of first-time Creator Mode users.
User Segmentation
Instagram has 2 billion monthly active users, but not all of them interact with Meta AI the same way. I would break the addressable Meta AI user base into four segments:
Casual Explorers (18 to 35, high volume, low engagement): Users who opened the Meta AI thread once out of curiosity, asked something like “what’s the capital of Peru,” and have not returned. They represent the majority of Meta AI’s total user count but contribute almost nothing to engagement depth. Retention risk is existential: they have no mental model of what Meta AI is for on Instagram.
Content Creators, nano to mid-tier (18 to 34, 1,000 to 500,000 followers, high intent): Users who treat Instagram as a primary creative and professional channel. They post 3 to 7 times per week, track their analytics, and are already using third-party tools like Later, Notion, and ChatGPT for content planning. They have a clear job to be done and Meta AI has a real opportunity to become embedded in their workflow.
Social Planners (22 to 40, moderate engagement): Users who use Instagram DMs as a social operating system. They organise group trips, share restaurant recommendations, coordinate events. Meta AI could serve as a planning co-pilot embedded in their existing communication behaviour, but the use case is more intermittent than creators.
Students and Knowledge Seekers (18 to 24, high session frequency, shallow depth): Users who treat Meta AI like a casual search engine or study helper. Engagement frequency is high but sessions are short and single-turn. Image generation (the “Imagine” command) is a popular hook for this group. Conversion to deeper habit is possible but requires better onboarding to unlock utility.
I am going to focus on Content Creators. They visit Instagram daily, they have an explicit and recurring job to be done (produce great content, grow their audience, maintain their brand voice), and they are already paying for or investing time in third-party tools that Meta AI could credibly replace or augment. If we can make Meta AI indispensable to even 5 percent of Instagram’s estimated 200 million creator accounts, that is a meaningful retention and engagement win. Creators also have strong social proof potential: when a creator publicly attributes their content strategy to Meta AI, it drives organic discovery among their followers.
