Mass-Producing Virtual Products with AI: $3,700 a Month on One Xianyu Account. How to Cash In on This Under-the-Radar Side Hustle

You want a side hustle, but something’s always in the way. Hoarding inventory ties up your cash. Shipping means handling customer service. Shooting videos means showing your face. Building a social media presence means fighting for traffic. Each path works on its own, but stack them together and most people quit before they start.

There’s one business that sidesteps every one of those roadblocks: no inventory, no shipping, no face on camera. You make one asset and sell it as many times as you want. That’s virtual products. In 2026, the heaviest part of this business, creating the content, has been slashed by AI to almost nothing.

First, the Market: How Big Is This?

Xianyu’s numbers get cited across the industry: the platform sees over 300,000 daily transactions in virtual goods (industry-cited data, not an official Xianyu figure). “Knowledge-based” and “template” products have the highest repurchase rates. What are virtual products? E-books, templates, courseware, asset packs, scripts, interactive mini-games. Build once, copy infinitely, and selling one more copy costs almost nothing.

Drill into a specific niche and there’s a “pre-class interactive mini-game” asset pack that, by industry accounts, moved 28,000 orders (this is an industry case study, not independently verified). One thing to get straight, though: that volume was stacked up years ago with sheer manpower. It proves the demand is real, not that AI will let you replicate those numbers overnight. What AI changes is the cost of supply. It doesn’t conjure demand out of thin air. These niches look unremarkable and the price points are low, but demand is steady and repurchase rates are high. One operator runs a single Xianyu account, spends about 2 hours a day on maintenance, and brings in around ¥3,700 a month. That’s not a top-tier outlier; it’s the baseline a normal account can hit when run seriously, and it’s the average across natural swings, not a guaranteed flat income every month.

The make-or-break variable here has never been “make one copy.” It’s “make a hundred copies that still sell.” The old bottleneck was human content production being too slow to keep up. AI has broken that barrier, but only the creation half. It hasn’t touched the threshold of actually selling. Those are two different problems, and they need separate answers.

How Far AI Cuts the “Creation” Step

Creation first. A few years ago, building a single teaching asset pack, nailing the topic, sourcing images, formatting, adding animations, took a skilled worker about 3 hours. The widely cited figure is that AI course-creation tools compress that to roughly 15 minutes. That’s outline-level generation, not a delivery-ready product. The mainstream tools on the market don’t stray far from these five:

• ChatPPT: the closest fit for education. One-click import across 40+ formats, turning a Word lesson plan straight into a slide deck.
• Gamma: strong dynamic visuals, good for public lectures and high-impact presentations.
• Kimi: turns long documents into slides. Paste the text and it builds the deck without losing the logic.
• Zhipu GLM: tight academic logic, built for STEM formulas and technical terms.
• Focusky: classroom interactivity with a micro-lesson feel.

In plain terms: building a “pre-class interactive mini-game” pack used to mean knowing design, animation, and interactive scripting. Now you just need to know how to prompt. Feed Kimi something like: “You are a classroom activity design expert. Design 10 five-minute icebreaker games for third graders. Each must include the name, rules, props, duration, and teacher-student interaction script. Output as a table.” Then have it source images and lay it out. One rough draft turns into ten, in about half an hour.

But AI only hands you the rough draft. To charge a real price and get repeat purchases, you still run a human pass: fact-check the content, polish the layout, test the interactive pieces to confirm they work. AI crushes the time from 0 to 60 down to almost nothing. Getting from 60 to 90 is on you. That’s the real shape of the cost reduction, not “cost going to zero.”

It’s not just courseware. AI can mass-produce a wide range of virtual products: Jimeng, Kling, and Midjourney for phone wallpapers, avatars, and sticker packs; ChatGPT for viral Xiaohongshu headline banks and copywriting templates; AI for holiday posters, invitations, and certificate templates; AI for industry resource packs, certification question banks, and SOP manuals. The logic is the same: AI pushes content production cost toward zero, and you handle product selection, packaging, and distribution. Let AI do the heavy lifting. You make the calls that actually matter, who to sell to and what pain point you’re solving.

How to Sell: Xianyu for Search, Xiaohongshu for Recommendations

You’ve built the products. Now turn them into cash. Two battlegrounds: Xianyu for search traffic, Xiaohongshu for recommendation traffic.

Xianyu comes down to one rule: keep listing new products. List something every day, without fail. Refresh your listings daily with one click. Reply to messages promptly and set solid auto-replies. Grind the Xianyu coin tasks for in-app exposure. For fulfillment, use Xian Guanjia (built by Xianyu’s official partner), around 100 yuan a year; it auto-delivers from your digital-key warehouse so you don’t have to man the phone. Once the model works on one account, clone it into a matrix of accounts. One catch: a single ID can verify at most 3 Xianyu accounts. To go wider, you borrow family members’ identities. Don’t plan on running ten accounts on day one.

Xiaohongshu runs on the opposite logic. It’s driven by posts, not search. And there’s a hard line to know upfront: Xiaohongshu aggressively punishes off-platform routing. Dropping a Xianyu link in your post is close to asking for a shadowban. The insider move is to keep links out of posts entirely. Tease the offer in your bio, send the link by DM, and embed stealth watermarks in your visuals to funnel people over. The posts themselves should show real use cases, surface customer feedback, and target pain-point keywords. One viral post can send hundreds to thousands of visitors to your Xianyu store.

Two fronts: Xiaohongshu builds the hype, Xianyu closes the sale, automated fulfillment covers the rest. Orders come in while you sleep.

A few traps nobody likes to talk about

1. Platform policy is the bigger landmine, not copyright. Xianyu keeps tightening the rules on virtual goods. Virtual categories are flagged as sensitive, so new accounts get close to zero organic traffic for the first 30 days, and you live under the constant threat of delisting or throttling. The platform’s own rulebook is less controllable than a copyright strike. Walk in knowing that.
2. Stay off copyrighted material. Don’t upload anything with real names, institutions, or brand logos. One complaint from the rights holder and you’re looking at anything from a banned account to a lawsuit. The test is simple: if the name isn’t yours or the assets come from someone else’s paid course, it’s a no.
3. Refund disputes eat 10 to 20% of your profit. Refunds are the real trap with virtual products. The moment a card key or access link goes out, a buyer can screenshot “won’t open” and support will usually side with them. That’s standard loss for a virtual account. Price it in.
4. One account won’t conquer the market. A single Xianyu account has a hard traffic ceiling. To scale you need a matrix of accounts and paid traffic, cash for eyeballs. Two to three accounts per person is the starting line. Don’t put all your eggs in one basket.
5. Test new categories with AI first. Don’t stockpile dozens of asset packs on day one. Use AI to spin up three to five listings, put them up, and watch the clicks and inquiries. If demand is there, scale. The biggest upside of the AI era is that testing costs almost nothing.
6. Homogenization is the real rival. A lower barrier means anyone can use AI to produce 30 packs. The day you drop a winner, competitors copy your title and main image within three days and cut the price in half. Your moat isn’t “knowing how to make it.” It sits in three things: a six-month early window, drilling into one vertical audience (say, early-childhood transition only, not every subject), and the discipline to keep listing new products. Those three aren’t easy to take from you.

A 7-Day Launch Checklist, Without the Hype

If you want to move today, here’s a checklist you can actually run:

• Day 1: Lock in a hyper-specific niche and scenario. Brand-new teachers, training center instructors, HR icebreakers, parents entertaining kids, streamer audience interactions. The tighter the angle, the easier it cuts. Skip “all teachers.” Public school teachers have shared resource pools from their teaching research groups and free Seewo whiteboard lesson plans; they are not your buyers. Your buyers are the rookies with zero accumulated resources.
• Day 2: Use AI to pump out 5 test assets. Feed a prompt like this to Kimi as your starting point: “You are an expert in classroom activity design. Design 10 [5-minute] pre-class icebreaker mini-games for [3rd graders]. Each must include the name, rules, props, duration, and teacher-student interaction scripts. Output as a table.” Then have it generate images and format everything into a polished, finished product. Don’t try to pass off half-baked work.
• Day 3: List your items on Xianyu. Stuff the titles with long-tail keywords (low search volume but highly precise terms, like “pre-class interactive mini-games icebreaker warm-up class activities 3rd grade”). Price them between 9.9 and 19.9 initially to drive volume.
• Day 4: Set up Xian Guanjia for automated fulfillment. Configure your auto-replies and completely delete “staring at your phone to reply to messages” from your workflow.
• Day 5–6: Post 3 to 5 scenario-based notes on Xiaohongshu. Remember: never drop links in the main text. Drive traffic through your profile page and DMs. Test which angles actually get clicks.
• Day 7: Read the data. For categories with inquiries and sales, use AI to copy and scale. For dead listings, cut them and pivot. Don’t get sentimental.

You probably won’t make real money in 7 days. But you’ll have wired up the whole loop: AI product creation, listing, traffic, closing the sale. Once that loop runs, the rest is copy and scale.

Bottom Line

The barrier in the digital product game was never tech. It’s execution and judgment. AI took down the tech barrier. Xianyu and Xiaohongshu flattened distribution. Xian Guanjia absorbed fulfillment. What’s left is “do you actually want to start” and “are you willing to drill into one audience.”

One person, one laptop, a few AI tools, turning unremarkable little assets into a steady side income. It’s happening every day in 2026. The ones who move first eat the window.