Mastodon Isn't One Place — It's a Constellation of Communities

The first thing newcomers ask about Mastodon is which server to join, and the question carries a low hum of anxiety. Pick the wrong one and you'll be stuck with the wrong people, the wrong rules, the wrong vibe. The mainstream platforms trained us to think this way. On Twitter or Threads or Bluesky, the platform is the place — there's one of it, and you join it. So when Mastodon shows up with a server-picker page and several hundred options, it reads like several hundred ways to make a mistake.

That framing is misleading, and it’s probably the main reason people give up on the network before they really get it. Mastodon isn’t a single place you sign up for; it’s a protocol that links communities, and those communities are what you’re actually choosing between.

A constellation, not a product

Each Mastodon server is an independent community with its own admins, its own funding, its own moderation policies, and its own culture. They share a protocol — ActivityPub — that lets accounts on any server follow, reply to, and share posts with accounts on any other server. From a user's perspective, the network behaves like one social graph. From a governance perspective, it behaves like a federation of small towns that happen to share a road system.

That separation between protocol and community is the structural fact most newcomers miss. ActivityPub is not the property of any one company; it's a published standard that any compatible server can speak. When you sign up on a particular instance, you're picking the community whose timeline, rules, and admins shape your local experience — but the protocol underneath means your account can follow anyone, anywhere, on any compatible server. The road system is open. The neighborhood is what you're choosing.

The implication is that "which Mastodon should I use?" is the same kind of question as "which neighborhood should I move into?" The answer depends on what kind of community you want to be part of — what they care about, how they fund themselves, who runs them, and what they've decided is and isn't allowed in their corner of the timeline. Three patterns show up consistently across the major instances, and once you can see them, the constellation organizes itself.

For the broader fediverse primer that this post takes as background — what ActivityPub is, what the wider network of services around Mastodon looks like — see the earlier post What the #$! is a Fediverse?*.

Communities run on donations, not advertising

The flagship instance, mastodon.social, is operated by Mastodon gGmbH — a German non-profit. Its funding mix is donations, sponsorships, and grants, with no advertising and no venture capital pressure to extract value from the user base. The same model holds across most major instances. mastodon.world publishes its budget on OpenCollective. Hachyderm runs on community contributions. Fosstodon takes patron support. mas.to and mstdn.social ask their users to chip in.

This isn't decoration. The funding model is the governance model. When the people paying for the server are the people on the server, decisions about moderation, federation, and product direction get made on different timelines and against different incentives than they would inside an ad-supported company. There's no growth target that requires algorithmic engagement. There's no quarterly pressure to monetize attention. The user base is much smaller, much cheaper to keep happy — and it's the same group of people producing the content that makes the place worth being on.

The downstream effects show up in places that take a while to notice. Feeds are chronological because there's no advertiser case for ranking them. Account verification works through your own domain because there's no central authority that needs to monetize blue checkmarks. Server admins post about budget decisions because the people they're answering to are the people funding the server. None of this is utopian — donation funding is fragile, instances can shut down or downsize when contributions dry up, and "the people are the product" doesn't translate into the kind of polish that a billion dollars of operating budget produces. The interface is rougher. Mobile apps are uneven. Discovery is weaker than what algorithms deliver elsewhere. What you get in exchange is a category of social space that simply doesn't exist on platforms designed to maximize user-time-spent.

Place and language still anchor identity

mstdn.jp is one of the largest Japanese-language Mastodon instances and has been a significant home for Japanese tech and fandom culture since the early days of the network. mastodon.uno hosts a significant Italian-language community. mastodon.scot is small, donation-funded, and culturally Scottish in a way that no global platform allows for. mastodon.london serves people who want London to be a unit of identity rather than a location tag. sfba.social is the Bay Area's regional commons — relevant precisely because the Bay Area is where so many of the technical decisions about the future of social networks get made. mstdn.ca runs as a Canadian generalist instance with both English and French activity.

Mastodon has not flattened geography the way mainstream networks have. A Japanese-speaking user can land on mstdn.jp or pawoo.net and find a timeline shaped by Japanese conversational norms, Japanese-language content, and Japanese cultural references — without giving up the ability to follow anyone on any other server. A Scottish user can join a community that treats Scottish civic life as the center of attention rather than as a side topic. The wider fediverse remains accessible; it just isn’t the default lens.

This is one of the reasons "pick a community" maps better than "pick a server." Regional and language instances aren't worse versions of the global flagship — they're communities organized around an identity that the flagship can't credibly serve. Part 2 of this mini-series will go deeper into these communities, and into the niche and topic-focused instances that work the same way at the level of craft instead of place.

What a community thinks about AI is now part of its identity

mastodon.art bans AI-generated content outright. Its public rules name generative imagery as out of scope for the community, and this isn't a quirk of one server — it reflects the community of working illustrators, photographers, and concept artists who chose mastodon.art as their refuge from a wider internet they feel is being flooded with synthetic work. The ban is the community's policy; the policy is what attracts and retains the membership.

That's one position. There are three others worth knowing, and together they form a spectrum that any creator using AI-assisted tools should understand before picking where to post.

The spectrum of community AI policies

At one end is the outright ban. mastodon.art is the clearest example, and it isn't the only one — several smaller art and craft instances have similar rules. If you post AI-generated images on a server with this kind of policy, expect removal and, depending on the instance, an account suspension. The point of the policy is to protect the community of human creators that the server was built to serve.

In the middle is the disclosure-required position. Fosstodon's code of conduct allows AI-assisted content as long as it's disclosed clearly. The community accepts AI as a tool but expects honesty about its role. This is becoming a common middle position across tech-leaning instances — content is welcome, the audience just gets to know what they're looking at.

Adjacent to that is consent-based for training. infosec.exchange explicitly prohibits training AI or machine-learning systems on user content without consent. Hachyderm publishes infrastructure documentation stating that user data will not be used for AI or LLM training. These instances are less concerned with what you post and more concerned with what gets done to what other people posted. The position protects the community's data even from outside companies that might scrape the public timeline.

At the other end is silent or permissive. Most of the major general-purpose instances — mastodon.social, mastodon.online, mas.to, mstdn.social, mastodon.world — haven't published bright lines on AI generation or training. That's a signal too, just a quieter one. It means the default applies: post what you want, with the broader Mastodon community-norms as your only soft constraint.

What this means if you publish AI-generated content

The instance you choose decides what you can post and how it has to be framed. A creator who runs a YouTube channel using AI-generated thumbnails or AI-narrated voiceovers can share that work without friction on a permissive general-purpose server. The same creator posting the same work on mastodon.art will be removed. On Fosstodon, the work is welcome but should carry a clear disclosure tag in the post itself.

This is not just a moderation question — it shapes reach. When you post on a server that aligns with your work, your audience grows naturally because the people who follow you don't object to what you're sharing. When you post on a server that doesn't, you spend energy fighting the community's culture instead of building on it. Federation makes the friction visible: a post that survives on your home instance can be muted, blocked, or reported by users on other instances who have stronger preferences. Repeated reports can prompt your home admin to ask you to leave, or prompt other servers to defederate from yours.

The tactical version of this advice is short. Read the code of conduct before you sign up, and look for explicit language on AI generation, AI assistance, and disclosure. If you produce AI-assisted work — images, video, narration, text — prefer instances that either welcome it openly or accept it with disclosure. Avoid instances that ban it outright unless you're committed to posting only your non-AI work there. If your output mix is split, consider running two accounts on two different instances, with each account scoped to the kind of work the host community welcomes.

The training question is separate from the generation question

A point that's easy to miss: a community can be permissive about AI-generated content while being strict about AI training on user data. Hachyderm welcomes posts of all kinds but won't allow its admins or third parties to train models on the timeline. That distinction matters even for creators who don't use AI tools at all — your posts are content that other systems can ingest, and the instance you sit on affects whether that ingestion is allowed.

If you care about your work being scraped to train someone else's model, instances with explicit no-training policies (infosec.exchange, hachyderm) offer protection that general-purpose servers don't. The protection isn't perfect — public posts are public, and policies don't stop bad actors — but it does give your admin grounds to act when violations are documented, and it signals to compliant scrapers that your data is off-limits.

Part 3 of this mini-series will treat AI policy as its own subject, and will close with the cautionary opposite case: Truth Social, a platform built on Mastodon's open-source code that deliberately removed federation and exited the network entirely.

For now, the takeaway is that "what does this community do about AI?" has become a basic question to ask when picking where to land — alongside funding, language, and culture.

Touring the big tents

The general-purpose instances are the natural starting point for most people, and they share a job: serve a broad, multilingual, mainstream-leaning user base without committing to a tight niche. They're not interchangeable, though, and the differences are easier to feel than to describe.

mastodon.social

The flagship. Operated by Mastodon gGmbH, English-dominant, very large, very visible, and the default landing place for journalists and FOSS-adjacent users who arrived during the Twitter migrations. Its culture is closest to "Mastodon as default identity" — when someone hands you a Mastodon handle without specifying a server, the muscle memory says it lives here.

mastodon.online

The sister instance, also gGmbH-operated, broadly similar in feel. It exists in part to give the project a second on-ramp when mastodon.social signups are paused or capacity is constrained.

mas.to

Fast, friendly, well-federated, and the server that gets recommended a lot when someone asks "where should I go that isn't mastodon.social?" Its user base leans technical and online-commentary-flavored. Tech and platform writers like Dare Obasanjo post there.

mstdn.social

Large, generalist, and known for transparent admin posts about the state of the instance. It's the answer for users who want a big server without joining either of the two flagships.

mastodon.world

Part of a wider fediverse infrastructure ecosystem, funded openly through OpenCollective, and pitched as a sustainable on-ramp for mainstream users. The funding transparency is part of the brand — you can see the budget, the contributors, and the financial reports.

universeodon.com

A self-described "space for all humans who have a little curiosity," operated by the team behind mastodonapp.uk via ATLAS Media Group. It's where George Takei posts.

mastodonapp.uk

The UK-linked general server best known for having served as Stephen Fry's entry point into Mastodon during the post-Twitter migrations.

These seven aren't the only general-purpose options, but they're the ones that show up most consistently in newcomer recommendations. The differences between them are real: the flagship-vs-community-run question, the funding visibility, the cultural concentrations of journalists or technologists or curious mainstream users. None of them is the wrong answer. All of them federate with each other. The choice you're making at signup is which on-ramp you want, not which network you're committing to.

When a public figure picks an instance, the instance changes

George Takei posts from a verified account on Universeodon. Stephen Fry's Mastodon presence is on mastodonapp.uk. Greta Thunberg's account lives on mastodon.nu. Three things follow from this.

First, federation means none of those choices restrict who can follow them. A user on mastodon.social, mas.to, fosstodon, infosec.exchange, or any other server federated with these three can follow all three from a single account. The network is still one network from the user's point of view. Centralized platforms made "the place" and "the celebrity" the same product. Mastodon doesn't.

Second, anchor accounts shape the visibility and stability of the instance they're on. Universeodon's profile rose noticeably when George Takei landed there. mastodon.nu became more visible to climate-engaged audiences when Greta Thunberg's account became the de facto reference point. This isn't quite the celebrity-launches-a-platform dynamic that mainstream networks chase, because the celebrity here doesn't bring exclusive distribution — but they do bring attention, and attention helps fund and sustain the instance.

Third, the existence of distributed anchor accounts is what proves the design works. A platform that depended on a single celebrity hub couldn't scatter its biggest names across separate communities and stay coherent. Mastodon can, because the social graph isn't owned by any one server. The famous people are on the same network everyone else is on; they just live in different neighborhoods.

Why a constellation outperforms a platform

The accusation against Mastodon, repeated often, is that it's harder to use than the centralized alternatives. That's true, and it's also the wrong measurement. Centralized platforms optimized for time-on-site and seamless discovery because they had to — the business model required it. Mastodon doesn't have that business model and doesn't need that optimization. What it offers instead is something the centralized platforms can't: communities that are actually accountable to the people in them, funded by the people in them, and structured around what those people care about.

The trade is friction for ownership. You give up the algorithmic feed and gain a chronological one. You give up the platform's growth machinery and gain a network where what you say is governed by people you can talk to. You give up the comfort of one-size-fits-all and gain a constellation of communities with real differences worth choosing between.

The good news for anyone still anxious about the server-picker page: the choice is much less load-bearing than it feels. Pick a general-purpose instance from the list above if nothing more specific calls to you. Migrate later if you find a community that fits better — Mastodon supports account portability and your followers move with you. The constellation isn't a maze. It's a map, and the map gets easier to read once you stop looking for the platform and start looking for the community.

Pick a community, not a server. The road system is the same either way; the neighborhood is what changes everything.

Ready to actually try Mastodon? The free Fediverse Quick-Start Checklist is a 12-step PDF that walks you through picking a server, setting up your profile, and finding your first follows. No long explanations — just go here, click this, type that.

What's next in this series

Part 2 of Exploring Mastodon is "Place, Language, and Craft: The Communities That Give Mastodon Its Texture" — a deeper look at the regional, language, and niche communities that work differently from the general-purpose servers covered here. Part 3, "Governance Is the New Differentiator: AI Policy, Moderation, and the Truth Social Cautionary Tale," closes the mini-series with the governance dimension, including the cautionary case of a Mastodon-derived platform that exited federation entirely.

After Exploring Mastodon, the Exploring the Fediverse umbrella continues with Federation Across Networks — a four-part mini-series on how Mastodon connects to (and disagrees with) Bluesky, Threads, Nostr, and the rest of the ActivityPub ecosystem (Misskey, Pleroma, Akkoma, Friendica, GoToSocial). For self-hosters and the technically curious, Run Your Own Instance lands later as the technical companion track on blog.federatedmind.com.

If you're new to the broader fediverse, the earlier post What the #$! is a Fediverse?* is the recommended primer — it covers the protocol and the wider ecosystem this post sits inside.

Sources and further reading

• Mastodon — Decentralized social media
• Mastodon documentation: Instance entity
• Mastodon documentation: ActivityPub specification
• mastodon.world OpenCollective budget
• Universeodon about page
• George Takei on Universeodon
• Fosstodon Code of Conduct
• Hachyderm community infrastructure docs
• How Mastodon makes money — Investopedia

Don't forget to subscribe so that you can receive the next article in this series - "Place, Language, and Craft: The Communities That Give Mastodon Its Texture"

Subscribe for the next one