Hours of scrolling through other people’s Reels, copying hooks into a spreadsheet, saving the videos that worked. In content marketing, research is the least rewarding and most time-consuming job. Alex Olim, who by his own profile is building consumer apps through UGC, now claims on LinkedIn that a new agent from Meta does this work for his team. “Meta's new AI agent can scroll Instagram for you,” he writes at the start of the post, which collected over 800 comments within a few hours. Almost all of them contain a single word, Muse. It is the keyword on which Olim sends out a link to his guide.
Muse is real, and Meta launched it on 8 September 2026 as a personal AI agent that, in the company’s words, “doesn't just answer questions, it actually does the work”. It handles emails, books trips, fills in forms and negotiates bills. It keeps working after the app is closed and asks for approval before sensitive steps. It runs on Meta’s own Muse Spark model inside an isolated, secure computer with its own browser, which Meta calls Muse Secure VM. The basic tier is free with a usage limit, and the premium plans, Power and Maximum, cost $20 and $100 a month.
The most downloaded app in the US App Store
On 18 September Muse reached first place in the US App Store, and a day later the top of Google Play. According to Sensor Tower, the app gathered 2.8 million downloads in its first two weeks, with downloads growing by an average of 55 percent a day between 8 and 17 September. For comparison, ChatGPT grew by 24 percent a day in its first ten days. By 25 September the app had passed 3.4 million downloads. Meta also paid for the product with the attention of its own ad inventory, where it ran its own advertising for Muse from 9 September, and according to Sensor Tower the number of daily active users rose by 27 percent on the Wednesday after the Meta Connect conference.
For now, Muse is available only in the US. European creators and brands are therefore watching it from across the Atlantic, not from their own experience.
On Instagram, it so far turns a recipe into a shopping list
This is exactly where what Meta delivered parts ways with what sounds good on LinkedIn. The only example of working with Instagram that Meta gives is that the agent can turn a saved recipe Reel into a shopping list, suggest a menu to go with it and remember dietary restrictions. From 27 September, Muse can be added directly to an Instagram profile, where it appears under the bio. Meta’s head of AI called it a “superintelligent sidekick”.
For creators, Meta promises more concrete things. The agent is meant to flag collaboration offers and compare them with the market average, supply summaries of account activity with reminders, and propose a content plan for Instagram. Because Muse is connected directly to Instagram and Facebook, according to Social Media Today, it can draw on official data from the platform. For creators that is a crucial difference from external tools, which estimate reach from the outside.
The six workflows Olim describes are not among Meta’s documented features. They include ranking other people’s Reels by the ratio of views to follower count, extracting text from videos, tracking five competitors every Monday, a hook bank sorted into groups, and verbatim mining of comments under viral videos. It is the author’s proposed use, a guide built around the agent, not a feature Meta promises. Whether Muse can handle them at the quality Olim claims will only be shown in practice.
Marketers let go of research before the creative core
What says the most is which task Olim picked. He did not hand the agent the script or the edit. He gave it the research, the work a team does before every script. “Most teams copy the winning video, when the thing worth copying is the reason it won,” he writes in the post, aiming at the most tedious, hard-to-bill part of preparation.
It shows where agents will reach first. They target the invisible layer beneath creativity: data collection, transcripts, sorting and pattern recognition. Whoever masters an agent that handles this layer gains a research lead over the competition. And when that agent is Muse, connected to official Instagram data, Meta holds that layer. Meta is building more tools around creators at once, and in September it also launched the Creator Marketing Hub globally, which turns a creator’s post into an ad draft.
Trust remains an open question. Social Media Today calls people’s willingness to entrust real tasks to a bot the biggest obstacle to adoption. Meta comes to this with a tarnished history. This August it agreed to a multistate settlement of $18 billion over the impact of social networks, and in 2019 it paid the FTC $5 billion for privacy violations. According to the company, Muse cannot see passwords or payment details and does not share conversations with the advertising system. It is a company offering creators that it will read, on their behalf, a platform it owns itself.
For the European market, what matters for now is that an agent doing research for a marketer has stopped being a conference hypothesis and is being downloaded by millions of people. Once it arrives in Europe, the main question will shift from whether it can scroll Instagram to who owns the data it reads from.
Comments