LaunchRanked

Validate your startup idea

Describe your idea and AI searches the web for competitors, what they charge and people asking for a fix. Then score it on six factors and get experiments for the next seven days.

1. Research the market

Describe the idea. AI with live web search looks for competitors, what they charge, and people asking for a fix, then rates your idea from the outside. About 10 seconds.

Free, up to 5 runs every 10 minutes. What you type is sent to the AI model to search with; we don’t store it.

2. Score it with what you know

Adjust each rating, and tick “I have evidence” only for what you’ve seen or heard from real customers.

Score 50/100 · Unproven. See your 7-day plan

Problem · 20%

How painful is the problem for the people who have it?

3 = A real cost in time or money, a few times a month.

Urgency · 15%

How soon do they need it solved?

3 = On this quarter's list, below other things.

Demand signals · 20%

What shows people are already looking for a solution?

3 = Some searches, forum threads or a few waitlist sign-ups.

Willingness to pay · 20%

Will they pay, and do they control a budget?

3 = They pay for tools like it, but you haven't asked about price.

Reachable channel · 15%

Can you reach buyers cheaply and repeatably?

3 = You know the communities or keywords but haven't tried them.

Competition · 10%

Is there room for you?

3 = Several products, with clear complaints about them.

What the AI research finds

Describe what the product does and who pays. A web-search model looks for the closest competitors and what their own sites say they charge, then for public posts where people ask for a fix: Reddit and Hacker News threads, forum questions, complaints in reviews. It rates each factor from the outside, names the assumption most likely to sink the idea, and suggests a sharper angle based on the gaps it found.

Every demand signal links to a page the model actually read, and a competitor’s own marketing never counts as demand. Competitor links and prices only come from sites it read. No competitors found usually means you described the idea in words customers don’t use, not that the market is empty.

How the scorecard works

Rate six factors from 1 to 5. Each has written anchors for 1, 3 and 5 so a “4” means the same thing next week as it does today. The ratings are weighted to a score out of 100: problem, demand signals and willingness to pay count 20% each, urgency and a reachable channel 15% each, competition 10%. The weights are our judgement, and the tool says so.

The headline number is the evidence-backed score. Any rating you haven’t ticked “I have evidence” for counts as 3 at most. That stops optimism from doing the work. The gap between your two scores is the size of what you’re assuming.

What counts as evidence

  • Past behaviour, not opinions. “Last month I spent two days on this” is evidence. “I’d definitely use that” isn’t. Rob Fitzpatrick’s The Mom Test is the standard short guide to asking questions that get honest answers.
  • Workarounds. Spreadsheets, scripts, a hired assistant, a competitor they complain about. People who already spend time or money on a problem are your best early customers.
  • Money. A pre-order, a deposit or a signed letter of intent is the strongest signal you can get before the product exists.

The 7-day plan

We pick your weakest factors (the ones costing the most points, with unproven ratings counted as risk) and list concrete experiments with a pass mark you set before you start. Write the pass mark down first; it’s too easy to move it afterwards. Most of them need no code: conversations, a landing page, a price, one channel.

Useful next steps: the customer interview script for the calls, the keyword idea generator for search phrases to check, and the startup name generator once the idea holds up. When you’re ready to ship, the launch checklist covers launch week.

Questions

Can a tool tell me if my startup idea is good?

No, and be wary of any that claims to. The research shows you what already exists and what people say in public, and the scorecard makes the gaps obvious. The only real validation is people with the problem giving you their time, their data or their money. The 7-day plan is there to go and get that.

How does the AI research work?

Your description goes to Claude Sonnet 5.5 from Anthropic, which searches the web as it answers. It looks for direct competitors and their published prices, and for posts where potential customers ask for a fix (Reddit, Hacker News, forums, reviews). Demand signals are only shown when they link to a page the model actually read, and never when they point to a competitor's own site. Competitor links go to the page the model read, or to that company's homepage if it read a different page there, and a price is only shown when it came from a site the model read. It can still miss things or misread a page, so open the links.

Should I trust the AI's ratings?

Treat them as an outside view from public information: useful for spotting a crowded market or a missing buyer, blind to what only customers can tell you. Use them as a starting point in the scorecard, then change them based on your own conversations. They never tick the evidence boxes for you.

Why does the headline score cap ratings without evidence?

Because founders rate their own ideas generously; we all do. A rating you can't back up with something you saw or heard counts as 3 at most (plausible, unproven). Tick "I have evidence" only when you could show someone what it is: interview notes, threads, sign-ups, a payment.

How were the weights chosen?

They're our judgement, not research. Problem, demand and willingness to pay carry 20% each because an idea fails without them. Urgency and channel carry 15% because they decide how fast you can sell. Competition carries 10% because competitors mostly prove there's a market. Change the ratings to see which factor moves the score most.

Is my scorecard saved?

Yes, in this browser only (localStorage), with your last research. The scorecard is never sent to us. The research sends your description to the AI model to search with; we don't store it. Use Copy as Markdown to keep everything in your notes or share it with a co-founder.

Free launch

Launching something? Get a page that ranks.

Launch free on LaunchRanked: a permanent, SEO-ready product page, a weekly leaderboard slot and a followed link while our badge is on your site.

Already live? See what your site is missing on Google, free.

Related free tools

See all free tools