Project Background: Turning an AI Companion Product into Clickable Content
BabeChat is an AI companion platform built around original characters and immersive interactive storylines. The campaign targeted Traditional Chinese users on X, using anime-style characters, persona settings, romantic emotion, and plot reversals to drive users to the website and encourage them to try character conversations and hidden storylines.
The creative direction relied on polished anime-style character visuals, with a card-style landing page taking over after the click. This project behaved like a cold-start acquisition campaign for a content consumption product: users were not searching with explicit intent. They had to be interrupted in the feed by character appeal, story tension, and emotional value.
So the core task was not simply explaining the product. It was translating the product into content assets that users were willing to stop for and click: character personas, story suspense, interaction promises, and a clear CTA.
Strategy Breakdown: Test Character Assets, Then Scale Interest Audiences
The campaign followed a path of character asset testing -> interest audience scaling -> landing page handoff -> remarketing recall.
During cold start, we prioritized tests across different character personas and copy hooks to identify which expressions produced higher CTR and lower registration cost. During validation, budget shifted toward high-engagement assets while we started building audiences from people who had visited the site, clicked cards, or shown interest without registering. In the stable phase, remarketing ads brought users back and pushed them into the conversation experience.
On the audience side, we split targeting into three groups:
- Anime, manga, gaming, romantic interaction, and virtual character interest audiences
- Broad AI audiences around AI tools, AI companion, chatbot, and AI girlfriend topics
- Remarketing audiences built from website visitors, ad clickers, and users with prior interactions
On the creative side, we kept the validated framework of strong character visual + high-emotion headline + short CTA, then expanded into multiple test directions such as tsundere contrast, immersive story, healing companion, hidden plot, and limited-time activity angles.
Account and Budget Setup
| Campaign module | Budget share | Objective | Creative / audience focus |
|---|---|---|---|
| Cold-start interest audiences | 45% | Test click and registration performance across anime, romantic interaction, and AI companion interest groups | Character visual + plot headline + CTA |
| Character creative A/B tests | 25% | Compare how different personas, art styles, and story hooks affect CTR and registration rate | Mage / schoolmate / healing character / contrast persona |
| Website visit remarketing | 20% | Bring back users who clicked ads but did not register, or browsed the site without starting a conversation | Continue the character storyline and emphasize unlocking more |
| Interaction and account growth | 10% | Build account engagement and a base for later organic spread | Reposts, character topics, story snippets |
Performance: From Character Clicks to Stable Registrations
| Stage | Focus | Impressions | Clicks / visits | Registrations / valid interactions | Core metric |
|---|---|---|---|---|---|
| Cold start | Basic tests across character personas and interest audiences | 1,280,000 | 18,560 | 1,042 | CTR 1.45%, CPC $0.18 |
| Validation | Keep high-click assets and expand similar interest audiences | 2,460,000 | 42,311 | 2,886 | Registration cost $2.64, CVR 6.82% |
| Stable phase | Scale high-converting personas and add remarketing | 4,150,000 | 78,942 | 6,324 | CPA $2.21, CVR 8.01% |
| Total cycle | Build a character-led content acquisition model | 7,890,000 | 139,813 | 10,252 | Overall CPA $2.38 |
Takeaway: Content Products Need a Reason to Pause First
This campaign showed one important lesson: on X Ads, an AI companion product cannot rely only on feature claims. It first needs to give users a reason to stop. Character visuals create the first moment of attention, story headlines create the motivation to click, the landing page turns interest into trial, and remarketing brings back users who looked but did not start a conversation.
For content-led products like BabeChat, cold start is not about finding the perfect audience in one step. It is about using enough character assets and emotional hooks to continuously identify which content directions can scale.