Reddit research questions, answered

GummySearch wound down at the end of 2025 over Reddit’s API licensing. This pipeline is an alternative built on Reddit’s public data, so it carries no API-licensing risk; the only cost is the per-run AI classification, which stays inexpensive.

No. It reads Reddit’s public data — the same data your browser loads — with nothing to authenticate and no quota approval. That is the design choice that makes it resilient to the API changes that shut other tools down.

A typical run classifies a few hundred threads for just a few dollars. Cost scales with how many comments you pull per thread, and a cap lets you set the ceiling before you start.

Fetching and filtering take a few minutes and are free; classifying ~300 threads with AI takes about 10–15 minutes. Runs are resumable, so a dropped connection or a rate-limit doesn’t lose progress.

No. You start from a plain-English claim, and the tool suggests relevant subreddits and search queries, which you can edit before running.

We tested 100 AI-suggested subreddits against Reddit itself: 90 were live public communities and only one was hallucinated. Every suggested subreddit is validated against Reddit before it reaches you, so invented names are dropped automatically.

A ranked markdown report: each thread scored 0–100 for pain and bucketed into a willingness-to-pay tier, with verbatim quotes and links back to the original discussions so you can read the raw evidence.

Reading by hand gives you stories, and the most dramatic thread sticks — not the most common problem. The pipeline reduces every thread to the same comparable fields, so you rank by signal across the whole conversation instead of by what you happened to remember.

Willingness to pay is a first-class field. Each thread is bucketed high / medium / low / none based on whether people describe paying for a fix, asking for one, or just grumbling — so the report sorts on demand, not just complaints.

It has no dependency on the paid API to change. Because it runs on Reddit’s public data, the API-licensing risk that ended GummySearch doesn’t apply.

Classification runs on a fast, general-purpose large language model. Each thread takes only a few seconds to classify, and the step is inexpensive.

Yes. Each idea is its own claim, so you can run several and compare which has the strongest underlying demand.

For multi-word phrases the pipeline tries an exact-phrase (“quoted”) search first and falls back to loose matching only when the exact phrase is too rare. In testing, about three in four multi-word keywords returned enough exact matches to use directly.