Builder project · Built at NYU SPS X Pulse Foundry
Plate Rank AI
PlateRank is an AI search optimization agent that shows restaurant how often AI assistant recommend
- ✧ First Place

Tip: click @Pulse AI in the post to tag us.
From idea to reality
What we built
Diners now ask ChatGPT, Claude and Gemini where to eat. A restaurant the AI doesn’t name never gets the customer, and owners have no way to see it happening. PlateRank makes that visible and fixable. PlateRank asks AI the 36 questions real customers ask, with live web search, and measures who gets recommended. For Café Athena in Pacific Beach, San Diego, it found the restaurant is already the AI favourite: named in 83% of answers, ahead of Spitz at 56%. It also found a blind spot: 0% for happy hour and drinks, despite a daily 3–6pm happy hour and Greek-mythology cocktails. PlateRank traced the cause. The website never says it where AI looks, and none of the 10 verified reviews mention the bar. It then builds the fix across seven tabs: Housekeeping: reads the real website and drafts a visible FAQ, a better title and description, robots.txt rules for AI search bots, schema and llms.txt. Each shows before, after, why, and how strong the evidence is. Social media: verified local creators, with content briefs and outreach drafts. Business value: what verified reviews praise, complain about, and never mention. Data farming: Reddit posts in the owner’s voice, reviewed by two AI agents. A Promoter and a Skeptic debate each draft against seven rules, and code (not AI) issues the verdict, so the same draft always gets the same answer. To-do list: one ranked plan built from every tab. Re-check: measures whether the fixes worked. Built against hallucination. Every fact from the internet goes through an evidence ledger and is marked verified, unverified or rejected. Only verified facts reach a score or a draft, and every claim links to its source. Nothing is posted automatically: the owner approves every word.