AI Visibility for Meta Muse
Meta Muse is an AI agent that shops, books, and buys for people. Centium tracks it as a seventh model: see how Muse recommends your brand, and what AEO and GEO mean when the buyer is an agent.
a seventh model.
Centium now measures AI visibility in Meta Muse alongside ChatGPT, Claude, Gemini, Perplexity, Grok, and Google AI Mode. It is included on every plan, with no change to your prompts, your categories, or your price.
Muse is Meta's personal AI agent. It launched on September 8 and reached No. 1 on the US App Store ten days later, ahead of ChatGPT, Gemini, Claude, and Instagram.
Centium measures Muse Spark, the model underneath the agent, through Meta's developer API. This gives you insight into the intelligence and search behavior powering this increasingly important model.
depth over frequency.
A new model makes your reading deeper. Every Centium plan runs 120 prompts across all seven models, which totals 840 AI responses behind every report and the highest volume of any entry plan on the market. Plans start at $99 a month, and that price has not moved as we've added new models.
Depth is what we believe in. AI models are non-deterministic: ask the same question twice and the answer changes, so any single response is a sample rather than a fact. Data drawn from over 800 responses is steady enough to act on and to compare against your longer range trend, without day-to-day noise.
The alternative is frequency, and most of the AEO category sells it. Twenty-five prompts on one model, refreshed daily, shows you the typical AI data fluctuation. We believe that is noise, and does not give business leaders the right data to act upon. It does not tell you where you stand in your market, because it never asked enough of the market to know.
the moment we were built for.
Centium was built on a premise that future buying decisions will pass through AI, and that we should treat these models like our future customers. That premise assumed agents were coming, and Muse is that future manifested.
Connected to a person's apps and accounts, Muse sends emails, books travel, and makes purchases for them. Meta trains the model for long-horizon agentic work and for computer use, where it operates an application it was never pointed to by reading the screen. The distance between a recommendation and a completed transaction is now one step, and the customer may never see the options that were passed over.
Answer Engine Optimization and Generative Engine Optimization, AEO and GEO, come down to one question for Muse: when an agent asks about your category, are you named? The work behind it is the work behind every model, which is earning the sources it reads and being the clear answer to the question your customers actually ask. What changes is the stakes, because here the answer is acted on.
more data, automatically
Muse appears everywhere the other models do: your recommendation rate, Model Visibility, Sources, Position, Fan-Out, and your export. The Methodology tab lists it as Muse Spark 1.3.
Expect your headline rate to move on the first seven-model update, in either direction. It now averages seven models rather than six, so the blend shifts even when your market has not. Your per-model trends are unaffected, and the blended line is comparable again from that update forward.
So far, we've noticed Muse has a strong grasp on brands and makes recommendations similar to leading models. However, it has behaved differently on citations by reading widely but crediting sparingly. It searches on nearly every prompt, then writes answers that name few of the pages behind them. Its data in Sources will be thinner than ChatGPT's or Gemini's.
hide or turn off any model.
Seven is a lot of models, and not all of them matter to every brand. If your customers are not on Meta's apps, or your team reports on ChatGPT and Gemini alone, you can now take a model off your dashboard.
Open your dashboard menu and choose Select Models. Each model has three states:
- Tracked is the default. We measure it and show it.
- Hidden keeps measuring the model and takes it off your dashboard. Your recommendation rate, Model Visibility, the trend, Sources, Prompts, Fan-Out, and Methodology recalculate without it immediately. Nothing is lost, so you can bring it back whenever you want.
- Off stops measuring the model from your next update forward. Past updates keep their data, and turning it back on resumes collection on the following update.
Competitors, Position, and the written AI summaries combine every model into a single number, so a hidden model still counts toward those until your next update. Off clears it from those as well, once that update has run.
The setting belongs to the brand, so everyone with access sees the same dashboard. Team members with manager access can change it.
Select Models
Choose which AI models this dashboard tracks and shows.
ChatGPT
Claude
Gemini
Perplexity
AI Mode
Grok
MuseHidden continues tracking a model and takes it off the dashboard. Off stops tracking in future dashboard updates.
Some features like Competitors, Position and the AI summaries blend model data and must be off to fully remove them.
Try it — nothing here is saved
questions, answered.
No. Every brand runs seven models automatically, with no change to your prompt count, your categories, or your cadence. Brands processed since the rollout already include Muse, and every other brand picks it up on its next update.
No. Plans start at $99 a month and include every model we track. When we add a model your reading gets deeper at the same price: 120 prompts across seven models is 840 AI responses per report.
Meta's personal AI agent. It launched on September 8, 2026 and reached No. 1 on the US App Store ten days later. Connected to a person's apps and accounts, it sends emails, books travel, and makes purchases on their behalf rather than only answering questions. It runs on iOS, Android, WhatsApp, and the web.
The model. Centium queries Muse Spark, the model underneath the agent, through Meta's developer API, the same way we query every other model we track. That keeps the measurement consistent across models and reproducible from one update to the next, which an app driven by one person's connected accounts cannot be.
Centium measures it for you. Every plan runs 120 category prompts through Muse alongside six other models, then reports how often Muse names your brand, which competitors it names instead, where you land in the answer, and which sources it cites. There is nothing to configure: Muse is included, and your next update covers it.
The same as for any other model, with one difference that matters. Answer Engine Optimization and Generative Engine Optimization both come down to being named when someone asks about your category rather than about you, and Muse acts on that answer: it shops, books, and buys. The shortlist is the outcome rather than a step toward one. The work is earning the sources the model reads and being the clear answer to the questions your customers ask.
It can move in either direction on your first seven-model update. Your headline rate now averages seven models rather than six, so the blend shifts even when your market has not. Your per-model trends are unaffected, and the blended line is comparable again from that update forward.
Hidden keeps measuring the model and takes it off your dashboard, so you can bring it back at any time with its history intact. Off stops measuring it from your next update forward, so those updates hold no data for it. Past updates keep what they collected either way.
Most of it, immediately: your recommendation rate, Model Visibility, the trend, Sources, Prompts, Fan-Out, and Methodology all recalculate without it. Competitors, Position, and the written AI summaries combine every model into one number, so they include a hidden model until your next update. Turning a model off clears it from those as well, once that update runs.
Muse searches on nearly every prompt but credits sources sparingly. We have noticed it pulls snippets from many search results, and may use this data without ever dropping citations. Its column in Sources will be thinner than ChatGPT's or Gemini's. That is a property of the model rather than a gap in the measurement.
More guides
Which Writers AI Cites in Your Category
Two new features are live. Influential Voices names the people behind the pages AI cites, and Content Freshness shows how old that content is.
Expanded Sentiment Tracking in Brand Perception
Brand Perception shows the themes AI models use when they talk about you. Each theme now expands to explain what the models actually said, in their words, with sentiment attached.
Content Amplification: Own Your Wins
Media companies are blocking AI crawlers, and the coverage you worked hardest to earn can go invisible overnight. Content amplification keeps those wins where the models can still find them.