3
2 Comments

I spent $600 on Quora ads before realizing 95% of my traffic was bots

I'm Ryan. I've been building RunCabin solo for 2 months — an AI website builder that produces real working sites with a backing database (the differentiator vs other AI builders is that you actually get the source code on GitHub when you publish, instead of being locked into a SaaS template).

Pre-revenue. ~3 months runway. Distribution has been the entire problem since week one.

Here's the story of one experiment that taught me more about paid acquisition than anything I'd read.

THE QUORA BURN

I picked Quora because the demographics looked right — small business owners, AI-curious developers, professionals researching tools. CPCs were low ($0.40-$0.80). I set up topic targeting across small business, web design, AI, programming, and indie hacker topics. Three creatives, three angles. $20/day budget.

Two weeks in: ~$280 spent, 700+ clicks, 0 signups. Activation rate showed as 0% across the board. I assumed the landing page was broken or the funnel had a leak somewhere. I rebuilt the landing UX, added quickstart prompt chips that auto-trigger a build, shipped streaming for the AI output so users see real-time progress. Still 0 signups.

After another week: $600 total spent. Still 0 signups.

I'd built an interaction-log table on the backend that captured every page hit with IP, user agent, action. I finally sat down and just queried it. The pattern was obvious in 5 minutes:

  • 26 of 36 sessions in one day had the EXACT same UA: "Mozilla/5.0 (Linux; Android 10; K) AppleWebKit/537.36" — no version variation, no Chrome version, no locale data. That's a fingerprint, not organic diversity.
  • Multiple visits from different IPs within the SAME SECOND. Two sessions from the same IP within 27 milliseconds.
  • iPhone UAs all on identical OS version 18_7 with no model variation.

It was a click farm. Quora's "Audience Network" placement (off-Quora ads on third-party sites and apps) was serving my budget to bots so I'd hit my daily limit and they could bill me.

When I dug into Quora's settings: yes, "Reach beyond Quora" was checked by default. Auto-enrolled into the network when I created the campaign. Buried under "Show advanced placement settings."

Real human visitors over 2 weeks: maybe 5-8 total. Cost per real human: ~$75. Real human activation rate: ~1 in 8. Effective cost per signup-shaped event: ~$600.

That's not a marketing failure. That's an inventory failure.

WHAT I'M DOING ABOUT IT

Pivoted to Google Search ads. Manual CPC at $2.50, narrow keyword targeting (small business website, profession-specific keywords like "website for plumbers"), no display network, no audience expansion. Total addressable bot market is much smaller on Google Search than on Quora's audience network.

Set up full conversion tracking with the gtag callback pattern so events fire before page-redirect, not after. Wired three conversions: "started a build" (activation), "bought tokens" (purchase intent), "bought publish" (high-value purchase intent).

First day of Google: 1 real activation (verified via referer header — came from Quora ironically, not Google yet, but a REAL human on a real Pixel device from Tyler Texas at 4am, clicked the karaoke chip, got a working site streamed back in 90 seconds, then left without signing up).

It's the first real production data point in 2 months.

WHAT I'M ASKING YOU

I'd love this thread to be useful for other pre-revenue indie founders, so a few specific questions:

  1. Has anyone here gotten Quora ads to work? If yes, what placement settings did you use, and was it before or after they aggressively pushed Audience Network?

  2. For B2B-ish SaaS targeting small business owners — what's actually worked for you? Google Search seems like the obvious answer but the CPCs on relevant keywords are brutal ($3-6 range).

  3. Open-ended: at what point did you stop iterating on the product and start iterating on the distribution? I keep wanting to add features instead of putting the work into outreach. That feels like procrastination dressed up as engineering.

WHAT THE PRODUCT IS, FOR CONTEXT

RunCabin's AI builder: https://runcabin.com/ai-build/

You type a sentence ("a karaoke night queue where friends add songs from their phone," "a website for my plumbing business with services and a contact form"), and Claude generates HTML/CSS/JavaScript plus a real database schema, streamed back live. 5 free prompts. Publish includes your domain + a private GitHub repo of the source code.

Stack: .NET 10 + EF Core + Postgres + Anthropic Claude (sonnet-4-6) + DigitalOcean droplets + Porkbun DNS + Traefik/Let's Encrypt SSL + ImprovMX email forwarding. All glued together by one person over 2 months.

Happy to answer anything. Roast the product or the marketing.

— Ryan
Baltimore, MD
[email protected]

on May 21, 2026
  1. 1

    This was a painful read but super useful. Quora Audience Network has burned a lot of people with low quality traffic.

    Switch to Google Search + strict negative keywords is usually the first real filter for real intent users. Also worth testing Reddit niche communities or direct founder outreach instead of scaling ads too early.

    Curious if you tried any organic channels alongside this?

  2. 1

    This is a strong founder problem because you are not just building another AI website generator. The useful difference is that RunCabin gives people a working site with real code, database structure, GitHub source, publishing, DNS, SSL, and less SaaS lock-in. That is a much stronger position than “AI website builder.”

    The one thing I would pressure-test hard before spending more on distribution is the name.

    RunCabin is memorable, but it does not immediately signal website creation, code ownership, or developer-grade publishing. In a crowded AI builder market, that creates extra work for the landing page because the name is not carrying the category or trust signal by itself.

    A name like Xevoa.com would fit this better as a broader AI web-building and publishing platform. It feels more like a serious product layer, and it gives you room to expand beyond prompt-to-site into code export, hosting, GitHub workflows, app generation, or small-business web infrastructure.

    Since distribution is already the pain, I would not treat naming as cosmetic here. If the first impression is unclear, every paid click becomes harder to convert.

Trending on Indie Hackers
AI runs 70% of my distribution. The exact stack. User Avatar 113 comments I'm a solo founder. It took me 9 months and at least 3 stack rewrites to ship my SaaS. User Avatar 98 comments Show IH: I'm building a lead gen + CRM tool for web designers targeting local businesses without websites — starting with Spain User Avatar 72 comments I built a URL indexing SaaS in 40 days — here's the honest story User Avatar 58 comments We could see our AI bill, but not explain it — so I built AiKey User Avatar 24 comments Creative Generator — create product-focused visuals and ad concepts faster User Avatar 11 comments