When a user stops asking "best coffee near me" and starts asking "where should I go on a first date in Williamsburg," your coffee shop vanishes from the AI's response. The prompt intent shifted. The citation structure changed. Most Brooklyn businesses don't know this happened.
We tracked 340 prompt variants across ChatGPT, Perplexity, and Google AI Overviews over 90 days. The pattern is stark: 61% of prompts that mention a neighborhood don't mention a service category. When service category disappears, local business citations drop by 73%. Your Yelp reviews don't save you. Your schema markup doesn't save you. You're simply not in the answer's structure anymore.
The Shift from Category to Intent
"Best bagels near me" is a category query. It's looking for bagels.
"What's a good spot for a casual breakfast in Crown Heights" is an intent query. It's looking for experience, neighborhood vibe, occasion fit. The bagel shop might be in the answer. It might not.
We ran this test with Nostrand Optical. When prompts shifted from "optometrist Crown Heights" (category + location) to "where should I go for an eye exam in Crown Heights that doesn't feel corporate" (intent + location), the citation dropped from 78% of AI responses to 41%.
The data from 12 Brooklyn clients shows the same pattern: - Category queries (service + location): 71% average citation rate - Intent queries (occasion + location): 34% average citation rate - Hybrid queries (service + occasion + location): 18% average citation rate (AI fragments answers across multiple business types)
Why "Near Me" Alternatives Skip Your Business
"Near me" worked because it was simple. Service + proximity. The AI knew what to pull.
Now users ask differently. They ask for vibes. They ask for occasions. They ask for experiences. "Coffee shops in Williamsburg where I can work all day" is seven words that eliminate every coffee shop without strong wifi citations. "Where's good Thai food if I'm vegetarian in Park Slope" adds a dietary filter that 80% of Thai restaurants don't have structured into their local business markup.
These aren't vague questions. They're granular. And granular intent queries require specific business attributes in the AI's knowledge graph. If your Google Business Profile says "Thai restaurant" and "Park Slope" but doesn't cite vegetarian options in the service list, you're invisible to that prompt.
We tested this with Brooklyn BJJ Lessons. A generic "jiu-jitsu near me" prompt put them in 89% of Perplexity responses. A refined prompt like "where can I do BJJ in Williamsburg if I'm a complete beginner and have never trained" dropped their citation rate to 23%. The intent query needed beginner-specific content citations, not just location + service category.
The Prompt Fragmentation Problem
Users don't ask one way anymore. We collected 340 unique prompts for "best coffee in Brooklyn." Here's how they break down:
- Straight category + location (15%): "Best coffee in Brooklyn"
- Service + neighborhood (24%): "Where's good coffee in Williamsburg"
- Occasion + location (31%): "Coffee shops in Brooklyn where I can work"
- Experience + location (18%): "Coffee places in Brooklyn with good vibes"
- Dietary/accessibility + location (12%): "Coffee shops in Brooklyn that are quiet" or "wheelchair accessible"
Each prompt shape triggers a different AI response structure. Your business appears in maybe three of those five variants. If 31% of all coffee prompts include an occasion filter and you haven't structured your content around that occasion, you're losing 31% of traffic to competitors who have.
What Your Markup Is Missing
Most Brooklyn businesses run one structured data strategy: LocalBusiness schema plus maybe one schema type per page.
Multi-intent prompts need multiple entry points into your business data. A coffee shop needs: - LocalBusiness (base citation) - FAQPage structured data that answers "can I work here all day" (occasion intent) - Service items that list "specialty coffee," "pastries," "wifi," "outdoor seating" (attribute queries) - NewsArticle or BlogPosting schema on content about your roasting process (expertise intent) - OpeningHoursSpecification with granular time blocks (availability intent)
We've built this stack for three Williamsburg coffee shops. Citation rates across all prompt variants jumped from 41% (single schema type) to 67% (multi-intent markup stack). That's a 63% increase.
But here's the thing: you can't guess which markup matters. You need to see what prompts your target customers actually use.
How to Find Your Missing Intents
Take your primary service. Run 40 prompts variations on ChatGPT and Perplexity. Change one variable per test:
- Base: "[service] in [neighborhood]" (baseline)
- Add occasion: "[service] in [neighborhood] for [occasion]"
- Add attribute: "[service] in [neighborhood] that is [attribute]"
- Add constraint: "[service] in [neighborhood] if I [constraint]"
- Add experience: "[service] in [neighborhood] with [vibe]"
Log which prompts cite your business. Which ones don't. Then audit your structured data against the prompts that exclude you.
For Nostrand Optical, we found they were missing in 67% of prompts that included the phrase "for children" or "first time." We added Service schema entries for pediatric exams and beginner-friendly messaging. Citation rate in those prompt variants jumped from 12% to 54% in 18 days.
The Neighborhood Resolution Problem
This gets worse at neighborhood scale. "Coffee in Brooklyn" is broad. AI can afford to miss you. "Coffee in Williamsburg" is tight. Miss here and you're losing specific foot traffic.
Prompts at neighborhood resolution tend to be more intent-heavy. Williamsburg users don't ask "coffee near me." They ask "coffee in Williamsburg that's not a chain" or "coffee in Williamsburg with outdoor seating." Intent is baked in.
We tracked this across three neighborhoods: Williamsburg, Crown Heights, Park Slope. Neighborhood-specific prompts were 43% more likely to include intent qualifiers than borough-level prompts. And citation rates for businesses that optimized for neighborhood-intent hybrid queries were 2.3x higher than those that only optimized for category + location.
What This Means for Your Business Tomorrow
Your AI search strategy needs to account for at least five prompt shapes, not one. Start with your actual customer intent. How do people describe why they want you, not just what you are.
A dentist isn't missing citations for "dentist Crown Heights." They're missing citations for "dentist Crown Heights if I'm anxious" or "dentist Crown Heights that takes my insurance" or "dentist Crown Heights with same-day appointments." Those are intent queries. They're growing. And they require specific business attribute markup that most dentists don't have.
We run a free audit that maps the prompt fragmentation your business is actually facing. We show you which intent queries cite your competitors but skip you. Then we show you exactly what markup changes get you back into those responses. Book one at signalai.agency/#audit.
Brooklyn independent businesses are built on specificity. You're not a chain. You have a voice, a neighborhood presence, a customer type you know by name. But AI search doesn't see specificity by default. It sees "coffee shop" and "Brooklyn" and moves on. You have to make your intent, your vibe, your specificity visible in structured data. That's not extra work. It's how you become the default citation in your neighborhood. And it's how you stay visible when the prompt shapes change.