Onira is a mobile app that helps users decode and explore their dreams through AI‑driven, scientifically backed interpretations. With an affordable monthly subscription, the platform makes expert dream insights accessible to a broad audience.

With a low monthly subscription price and correspondingly low AOV, Onira had to keep customer‑acquisition costs (CAC) extremely low to remain profitable. Traditional Google and Meta campaigns could drive traffic, but never at a cost‑efficient enough level. At the same time, we discovered that Reddit was full of high‑intent conversations: across multiple dream‑related subreddits, users were sharing problems and actively seeking scientifically backed dream‑interpretation advice the exact value Onira delivers.

The challenges were:
‍
• Paid ads were bringing traffic but not at the target CAC required by the low‑AOV subscription model
• Joining every relevant Reddit thread manually simply did not scale
• We needed to insert Onira into conversations in a helpful, non‑spammy way that converted

We built a Reddit AI agent fine-tuned to Onira’s tone of voice and user base. In the first month, we focused on testing different types of replies short vs. long, casual vs. expert, direct vs. indirect CTAs to learn what worked best.

Once we identified the sweet spot, we deployed the agent at scale:
• Scraping Reddit posts using semantic keyword clusters related to dreams, interpretation, and mental wellness
• Writing natural, helpful replies using Claude and GPT models
• Adding a soft CTA only when contextually relevant
• Operating via a single warmed‑up Reddit account to stay under the radar and avoid moderation flags

We also implemented tracking to monitor Reddit-driven traffic and conversions, giving us visibility into what was working.

We tracked every conversion driven by the Reddit AI agent. Above is the revenue it generated in the first 90 days only from our Reddit AI chatbot.  

400+ paying users added in under three months
+240 % increase in on‑site conversion rate, from 0.5 % → 1.7 %
After 4–6 weeks of testing, we doubled down on the top‑performing reply styles and scaled the posts per day.