Original Summary

I picked my SaaS idea from a spreadsheet, not from a shower thought. I took a list of 749 IndieHackers products in the $20k–100k/month range. First surprise: only 30 were Stripe-verified. The other 719 report their own revenue, and a lot of it is junk: round numbers, about 30 near-identical spam pages, and calculator sites claiming $100k a month. So the list is a place to start looking, not proof. Six categories survived. The one that passed question 3 best for me was tools that track how a brand shows up in AI answers (ChatGPT, Google AI Overviews, Gemini, AI Mode). I've done SEO for about 7 years, and lately half the conversation is "are we in the AI answer or not?" The project is called Sightable. It's a working name, and I haven't done the trademark search yet. What it does, in one line You enter your brand, your competitors and the questions your customers ask. Every day it asks the AI engines those questions and reports how often you're mentioned, who gets mentioned instead, and which sources the AI cited. The numbers, before writing any code My benchmark is Otterly. This month their plans are $29 for 15 prompts, $189 for 100 and $489 for 400, run daily on 4 engines. Volume = prompts × engines × 30 days: Plan Prompts Queries/month $29 15 1,800 $189 100 12,000 $489 400 48,000 Cost per query depends on how you get the answers. These are rough estimates, not tested at scale: Data providers (scraper/SERP style): about $0.001–0.005 per query, so gross margin is around 90%. Official search-enabled APIs: about $0.01–0.03 per query, so margin is 0–35%. At the top of that range, the $29 plan costs about $54 a month to serve. So compute isn't the hard part. Getting customers and keeping them is. Is there a moat? Not a technical one. An MVP with 4 engines through data providers looks like 2–4 weeks of work. Matching Peec (another competitor, 6 engines) is harder, because a couple of engines have no data provider and need custom browser scrapers, and those break. The bigger threat is Semrush and Ahrefs adding AI tracking to tools people already pay for. What could become a moat later is history. If you start tracking today, nobody can sell you the last 6 months. Benchmarks across many brands and agency workflows could help too. But at $5–20k MRR, distribution matters more than a moat. How I'm building it: one person, an AI team I build at night with Claude Code. Instead of one giant prompt, I set up a manager plus 14 roles (engineering, SEO, copy, legal, design and so on), and each role has its own playbook. I didn't write every playbook from scratch. I pulled MIT-licensed Claude skills from GitHub, reviewed them and threw a lot of them out. I kept 18 and dropped the ones that were generic or invented results in their examples. Nobody had a ready-made skill for the core of the product, so I wrote 4 by hand: Data pipeline: gets the answers from each engine. Cost guard: spending caps and a kill switch, because a bug

  • 情报分类：开源项目与落地
  • 分类依据：内容涉及项目实践、创业、副业或变现
  • 信息来源：Reddit · SaaS
  • 发布时间：2026/10/19 15:37:51