NameDropped

Inspiration

A quiet shift is underway in how customers find local businesses. Instead of searching and scrolling, they ask an AI assistant — "best med spa in Frisco," "personal injury lawyer in Houston who speaks Spanish" — and get three names back. If your business isn't one of them, you didn't lose the click; you never existed. Most small-business owners have no idea whether AI recommends them, recommends their competitors, or confidently states things about them that are wrong. The SEO industry measures none of this.

What it does

NameDropped (namedropped.ai) is an AI-visibility audit service for small businesses. It asks Gemini the questions real consumers actually ask — cold, with no hints — and measures who gets named. Then it audits the business's website the way an AI reads it: entity clarity, structured data, crawlable pricing, the GEO (generative engine optimization) issues that keep AI from recommending it. For regulated verticals — med spas, real estate, law firms, financial services, supplements — it crawls the whole site and flags marketing claims that commonly violate advertising rules (FTC guarantees, fair-housing steering language, unnamed certifying boards), each flag quote-verified against the exact page it appears on and framed for review with a professional, never as legal advice. Every report ends with a scored gap analysis, a competitor share-of-voice breakdown, a 30-day action plan, and a ready-to-publish optimized page.

Revenue is live and layered: a free teaser audit feeds a $99 full report, $49/month weekly monitoring, and a $249/month agency tier with team seats, tracked portfolios, white-label exports, and an MCP API — agencies can drive audits programmatically from Gemini CLI or any MCP client. Accounts are passwordless magic-link; fulfillment is fully automated from payment to delivered report with zero human involvement.

How we built it

Every LLM call in the product is Gemini (gemini-3.6-flash for consumer simulation and classification, gemini-pro-latest for analysis, scoring, compliance, and drafting). The stack is Google Cloud throughout: Cloud Run (the service plus three batch jobs), Firestore, Cloud Build, Cloud Scheduler, Secret Manager, the Places API, and Google Ads with GA4 measurement. Repo created July 22, inside the contest window; we disclose reuse of our own generic Express/React boilerplate patterns.

What makes NameDropped an XPRIZE story is that the company itself runs as an agent pipeline. Every weeknight at 2am, a Cloud Run job discovers Texas businesses in our regulated verticals via the Places API, discards any without a lawfully usable published contact email, runs the complete audit on each survivor, ranks them by how compelling our email would be to them (visibility gap, a high-confidence compliance flag, review volume), and drafts personalized outreach citing the prospect's own results — "we asked AI assistants 8 questions consumers ask; you appeared once; here's your report, already generated." A digest lands in the founder's inbox by morning. The one human step is approval, from a phone, in the operator console. Approved drafts are sent by another cloud job from a reputation-warmed mailbox inside hard-coded guardrails: published-emails-only with provenance, permanent one-click suppression, a 30/day cap, business hours only, jittered sends. A third job triages replies with Gemini — unsubscribes auto-suppress forever; interested leads reach a human, because humans close deals.

Challenges we ran into

Keeping a model honest in a compliance product: Gemini can't be allowed to hallucinate a regulation into a customer's report, so it only ever sees curated rulepacks, and every flagged quote is verified verbatim against the live page server-side before it can appear. Cloud Run's CPU throttling silently killed any work that continued after an HTTP response — we rebuilt batch work as dedicated Cloud Run Jobs and made customer audits fully asynchronous after real mobile users hit two-minute requests that died when their phones locked. And cold outreach that respects people took more engineering than outreach that doesn't: provenance-checked published emails only, permanent suppression, caps, business hours, jitter — all hard-coded in the sender, not requested in a prompt.

Accomplishments that we're proud of

Real, honestly-reported revenue: per the rules we report arms-length customer revenue ($39 — our first paying customer, a founder-led in-person sale at our disclosed promotional price; we claim no ads or cold-outreach conversions yet) separately from related-party revenue ($0 — a $99 self-test of the checkout flow was refunded same-day and is disclosed in the ledger). The unit economics only work because of Gemini: a complete audit costs about $0.15–0.25 in inference against a $99 price. Our entire Google Cloud bill from launch through mid-August is $51.02, and all but about a dime of it is Gemini inference — the hosting, database, builds, and batch jobs ride the free tier. Full P&L, from the $377 in domains down, is in the repo.

What we learned

The hardest problems weren't AI problems. Getting Gemini to run a compliance crawl took a day; making sure it could never put a hallucinated regulation in front of a customer was the real design work. The same pattern held everywhere: guardrails hard-coded in the sender, not in a prompt; revenue buckets separated in the ledger, not in a footnote. Building a company on agents taught us the agent is the easy part — the honesty constraints around it are the product.

What's next for NameDropped

Scale the nightly prospecting beyond Texas and into more regulated verticals, grow monitoring subscriptions into the recurring base, and push the agency tier's MCP API so agencies run NameDropped audits from their own Gemini workflows. Five million U.S. small businesses are about to discover that AI assistants are their new front door — NameDropped makes that visible for the price of a tank of gas, and demonstrates the XPRIZE thesis literally: one founder, one phone, and a fleet of Gemini agents running a revenue-generating company overnight.## Inspiration

Built With

• cloud-run
• cloud-scheduler
• express.js
• firestore
• gemini
• google-cloud
• google-places
• graph
• node.js
• react
• resend
• secret-manager
• stripe