Anyone can ask a chatbot for startup ideas. GapSonar reads hundreds of real complaints from the communities where your buyers already are, checks each pattern against live Google volume, difficulty and competition data, and hands you the gaps that survive. With quotes, permalinks and numbers you can defend.
Free reports need no account. The engine has a free tier.
Where the data comes from
Four independent data layers feed every gap. If a claim can't be traced to one of them, it doesn't ship. That is the whole product.
We harvest discussions from Reddit and Hacker News and extract typed signals: pains, feature requests, and the gold standard, costly workarounds. "I built a spreadsheet for this" beats any survey. Every signal keeps its permalink and date.
Monthly volume, keyword difficulty, CPC and 12-month trend for every keyword a gap touches, from a licensed commercial data provider. Not scraped, not estimated by a model, not last year's CSV.
For each gap we pull the live Google results and classify what actually ranks: a working tool, a blog post, a forum thread, an abandoned product. Three purpose-built tools in the top 3 is a wall. Three forum threads is an open door.
A curated taxonomy of real-world business verticals tells us how dense a market is, which makes directory and local plays measurable instead of hopeful.
The order matters. The AI never invents an idea and then defends it. Evidence is collected first; the model's only job is to describe what the data already proved. Ideas that fail verification get rejected, and we tell you how many did.
Three real signals from the live corpus. Each one is a person already paying for a missing product with their time.
"I got a Python script that does what I need"
"I tried a bunch of technologies like RPA tools but they are a bit expensive, so I'm looking for a more affordable choice."
"I have to do a lot of largely the exact same editing to make the data useful and reformat it for our software to read"
Free gap reports
Each issue takes one market, reads hundreds of its complaints, and profiles the gaps that survived verification. The free scan shows the top opportunities. The full report goes deeper on every one.
The engine
Reports are snapshots. The engine is live. Describe what you want to build and it answers from evidence, not imagination.
"Something that helps accountants stop retyping numbers." Semantic search finds gaps that match the meaning, not the keywords.
Verified or emerging, n= behind it, monthly search volume, the channel it sells through. Model brainstorms are labeled as exactly that.
If the corpus only has 3 real matches, you get 3, not 10 padded ones. Thin evidence says so on the card.
A cheap, layout-adaptive tool that extracts line items from any invoice PDF without regex or huge automation budgets.
Recurring subscription that automatically extracts line items from any invoice format, with a manual review step.
Dictate complex markup using voice with structured commands.
Pricing
One bad product bet costs months. One report costs less than a dinner.
Payments are being wired up. Until then the bundle button starts a free account and the report button reaches a human.
Questions
A chatbot generates plausible text. GapSonar collects evidence first: real complaints with permalinks, licensed search volumes, classified competition. The AI's only job is to describe patterns the data already proved. That order is the difference between research and fiction, and it's why every number in our output links to a source.
A verified gap has at least 3 independent people voicing the same problem plus hard evidence: someone already hand-rolling a workaround or explicitly willing to pay. Patterns with 2+ voices but no hard evidence yet are labeled "emerging" and we say so everywhere they appear. Single complaints get rejected as anecdotes, and each report states how many candidates didn't make it.
Demand signals come from public discussions on Reddit and Hacker News, quoted briefly and always linked to the original. Search metrics come from a licensed commercial provider of Google Ads and SERP data. Competition classification runs on live Google results. We never republish anyone's post and we never invent a number.
Then the report says so. A scan that finds 3 real gaps delivers 3, with the rejects counted. Padding a thin market to a round number is exactly the practice this product exists to replace.
Yes. Bundle customers can request verticals and we run the sonar on them as new issues. Email hello@gapsonar.io with the market you care about.
The next report is generating tonight. Get the free scans and see the difference evidence makes.
Get the free reports