0.3 Four Business Models: Pick One Before You Learn Any Tool

There is no such thing as "building an AI influencer". There are four businesses that all look like that from the outside, and they need different tools, different budgets, different amounts of your week, and different tolerance for risk.

Most people learn the tools first and discover the business model by accident. That is the expensive order. A Model D operator does not need a chat team; a Model B operator does not need a brand deck. Buying the wrong stack is not a small mistake — it is months.

The four models

Model A — the brand-safe persona

A persona with an aesthetic: travel, fitness, fashion, food. You build an audience, and brands pay to appear in front of it.

The economics are attractive and the evidence is mixed. Virtual influencers show engagement rates of 5.67% against 1.89% for human creators, and 57% of Gen Z follow at least one virtual persona, with 35% having bought something through one. [UNVERIFIED] — those figures come from reports adjacent to the vendors selling avatar tooling, and there is no independent panel behind them.

The counterweight is trust. The same body of work puts trust in a virtual persona at 42 out of 100 versus 73 out of 100 for a human, and finds the human advantage reaching 2.7× in some categories. Brand spend on virtual personas was roughly $1.37 billion, about 4.2% of influencer marketing, and growing at 243% year on year. The money is real; the trust deficit is real too, and it decides which categories will hire you.

Highest ceiling, longest runway. You will not have a brand deal in month one, because brand deals require an audience and an audience requires months. Budget six to twelve months before meaningful revenue.

Choose A if: you can wait, you have visual taste, and you are willing to be explicit about the persona being synthetic — because in this model, the brands you want to work with increasingly require it.

Model B — AI OFM and the subscription business

OFM means the management side of a paid-subscription and chat business. A persona posts on Instagram, TikTok, or X, and converts attention into a subscription on a platform built for it, where the recurring revenue engine is private conversation, not the subscription itself.

Fastest cashflow. Highest operational and reputational cost.

The platform data is solid: Fanvue reached $200 million ARR by May 2026, up from $100 million at the end of 2025, with roughly 325,000 creators and a 20% commission. [AS OF 2026-10]

The practitioner data is where you should look closely, because it disagrees with the marketing:

Read that table twice. The same community that produces "$4,230 a month" produces "most accounts make under $500". Both are true. The difference is not the tool stack — it is whether the person works the funnel daily.

The operational cost is the part the marketing never mentions: chat is not a feature of this model, it is the model, and it runs every day. Two independent practitioners reported more than two thirds of revenue coming from conversations. That is a job.

Choose B if: you will show up in DMs daily and you are comfortable with a high ban risk — platform bans follow device and IP, not content, and OnlyFans prohibits AI personas outright (see 11.2). If either of those makes you flinch, pick a different letter.

Model C — AI UGC for brands

You do not build an audience. You sell creative assets to brands that need content but cannot shoot it: product shots, short-form video, ad creative, lifestyle scenes.

B2B, smaller reach, steadier revenue. Your buyer is a marketing manager with a budget line and a deadline, not an algorithm. That changes everything about how you work: deliverables are scoped, invoices are predictable, and a ban on your Instagram account is an inconvenience rather than an extinction event.

The demand side is documented: roughly 48% of brand content is now AI-generated in some form. [UNVERIFIED] — the figure comes from agency-adjacent reporting and is probably inflated. What matters for you is the direction, not the decimal.

Pricing here varies far too much to give one number honestly. Rates depend on deliverable count, usage rights, and whether the brand wants exclusivity, and they are covered with sources in Part IX rather than guessed at here.

Choose C if: you can handle client work, you can hit a brief, and you want revenue that does not depend on an algorithm deciding to like you.

Model D — product, education, and persona licensing

You monetize the knowledge and the character, not the attention. Courses, templates, service work, licensing an established persona to a brand, or selling the persona's identity as a productized asset.

This is the model the book you are reading runs on. The publisher built a persona capability, then sold the method rather than the following. That is not a rhetorical flourish; it is the structural reason this model has the best risk profile of the four. When your revenue comes from teaching the method, a vendor disappearing is a paragraph to update rather than a business to rebuild, and a platform changing its policy is a chapter to revise rather than a revenue line to lose.

The ceiling is lower and the work is different — you are now a writer, a teacher, or a product manager as much as an operator. The floor is far higher, because you are not competing for attention against millions of accounts doing the same thing.

Choose D if: you already know something worth teaching, or you are willing to find out by doing the work in Models A to C and documenting it.

Side by side

The ranges below are planning estimates assembled from the case studies in Part IX. They are not measurements and not benchmarks. They exist so you can budget, not so you can compare yourself to them.

Choosing

Answer these in order and stop at the first one you can say yes to.

1. Do you already have distribution? An audience, a mailing list, a client base, a community. If yes, Model D is nearly always the right answer, because you can sell to people who already trust you.
2. Can you do conversation every day for a year? If yes and you need money fast, Model B. If no, cross B off permanently. It does not work part-time.
3. Do you have six to twelve months of patience and a visual eye? Model A.
4. Do you want clients and predictable invoices? Model C.

If you cannot say yes to any of them yet, the answer is not to pick a harder model. It is to take Model D and build the persona in public while you learn, which converts the learning period into an asset instead of a cost.

Where this goes wrong

Running two models at once. A persona doing brand-safe travel content and selling private subscriptions is a persona with no position, and both audiences can see it. Every practitioner who reported a reach collapse had drifted across models.

Choosing B because it is fastest. It is fastest only for people who will do the chat. For everyone else it is the most expensive way to learn a lesson about their own tolerance for repetitive work.

Choosing A because the ceiling is highest. The ceiling only matters if you reach it, and Model A has the longest runway of the four. A practitioner's summary of this is worth quoting directly: AI influencers are overrated unless you already have distribution figured out.

Letting a purchase pick the model. If you have already bought a video tool, you will naturally conclude you need the model that uses it. That is backwards, and Chapter 3.1 exists to undo it.

What you leave with

A chosen model, written down, with the two numbers that go with it: your monthly stack budget and your weekly hours. Chapter 0.4 turns those into the single number that decides whether the project is viable.