Inspiration

Competitive research is often a collection of tabs, search results, and screenshots. Teams need to know which companies matter, what changed on their public pages, and where each claim came from. We built RivalCurrent to turn AI search into a source-linked competitor radar.

What it does

Start with your product website. RivalCurrent reads its public homepage and uses Telem search to suggest comparable company sites, filtering out your own domain and review articles. Choose a company to track and RivalCurrent fetches its homepage plus relevant pricing, product, feature, and changelog pages. The first scan creates a structured baseline with positioning, audience, features, published prices, and links back to the source pages. Later scans compare saved page content and surface detected changes separately from web search leads. You can also start a daily Telem watch, check its updates, and generate a source-linked email brief for review.

How we built it

Telem Search finds potential competitors and relevant public pages; Telem Fetch returns page content for snapshots; Telem Monitors provides daily watch results. Browser storage holds the interactive watchlist and snapshots, while D1 stores newsletter settings and drafts. An optional OpenAI Responses step extracts structured profile facts and shortlists newsletter items; each profile fact must cite text present on a fetched page. Keys stay on the server.

Challenges we ran into

Search results can mix official company sites with comparison posts, duplicates, and the user’s own pages. We added domain and brand filtering and combined category and brand searches so direct competitors are easier to find. Many app URLs lead to a sign-in screen, so discovery resolves those to a public company page. Some homepages do not expose useful links in fetched content, so tracking uses targeted search to find pricing and product pages. We also kept web search leads distinct from actual changes detected between scans.

Accomplishments that we’re proud of

A first-time user can go from a product URL to a curated watchlist and a source-linked competitor baseline. Every tracked fact has a page to inspect, and an initial scan is labeled as a baseline rather than a new announcement.

What we learned

Useful competitive intelligence needs both retrieval and restraint: finding pages is only the first step. Showing provenance and the difference between a search result and a detected page change makes the output more trustworthy.

What’s next for RivalCurrent

We want to compare structured pricing and feature facts between snapshots, confirm meaningful changes before alerting, and build a timeline and battlecards from those verified changes.

Built With

Share this project:

Updates

Submission history