· Shop Locally · NYC Spots · 5 min read
Why AI Has Never Heard of the Best Halal Noodles in Queens
Old Sport Food in Forest Hills is a 4.7-star, hand-pulled noodle spot with 400+ reviews. Yet ChatGPT thinks it's a kebab restaurant in Flushing. Here's the anatomy of that failure.
This one is personal. It started with a birthday dinner. My wife wanted something comforting, noodle soup, the kind that warms you from the inside on a cold night. I asked ChatGPT for recommendations in Queens. I went down the rabbit hole: looking up each spot on TikTok, watching influencer reviews, cross-referencing vibes.
That’s how I stumbled across Old Sport Food. Not from ChatGPT or Google. From an influencer’s page, deep in related content, half by accident. I bookmarked it. We didn’t make it there for the birthday. But a week later, we went anyway. It’s a trek from the South Bronx. Worth every minute of the commute.
Fresh hand-pulled noodles, Lanzhou Beef Noodle Soup. The real deal, served with broad hand-torn noodles soaking up the braising liquid. A Muslim-friendly space with a 4.7 rating across 408 Google reviews and genuine word-of-mouth buzz in the halal food community. TikTok videos with thousands of likes. The kind of spot that earns loyalty.
And yet: ask ChatGPT for “halal noodles in Queens” and Old Sport Food will not come up. This is a textbook case of the exact problem Shop Locally exists to solve. Let me break it down.
The Five Reasons AI Can’t Find Old Sport Food
1. The website is describing the wrong restaurant
This is the biggest issue. The Old Sport Food website describes itself as offering “authentic Halal cuisine” featuring “tender kebabs, savory biryanis, flavorful wraps, and spiced curries.”
That’s Middle Eastern food. It’s not what Old Sport Food serves.
ChatGPT has categorized Old Sport Food as a Middle Eastern restaurant on its map results. When someone searches “halal noodles Queens,” the AI has no strong signal that this place serves noodles at all, because the restaurant’s own website says it doesn’t.
This happens more often than you’d think. The website copy was probably written to sound appealing and comprehensive rather than precise. But AI models treat your website as the authoritative description of your business. If your website describes kebabs, you’re a kebab spot.
2. The domain name plants the business in the wrong neighborhood
The domain is oldsportfoodflushing.com. The restaurant is in Forest Hills not Flushing. These are different neighborhoods, different zip codes, different customer bases.
ChatGPT placed Old Sport Food in Flushing on its map. When someone is in Forest Hills asking for nearby food, the model deprioritizes a result it believes is in a different part of Queens. Location signals matter enormously for AI recommendations. A mismatched domain compounds every other signal problem.
3. The website actively blocks AI crawlers
When I tried to fetch the site, it returned a 403 error. The server actively rejected the request.
That includes GPTBot, ClaudeBot and any other AI crawler trying to index the content. If the site can’t be read, it can’t be cited, summarized or recommended. A Google Business Profile and TikTok buzz are helpful but they’re not enough to compensate for a website that refuses to be indexed.
4. The social proof lives on TikTok, not on the restaurant’s own site
Multiple TikTok creators have posted about Old Sport Food. The content has real engagement, thousands of likes, comments from people asking for the address, the kind of organic buzz that money can’t buy.
But AI models lean heavily on a restaurant’s own website as the authoritative source. Social proof that lives only on TikTok doesn’t transfer into LLM recommendations. The model can’t easily attribute viral TikTok content to a specific business, verify the information or cite it with confidence. That social proof is essentially stranded on a platform AI can’t reliably use.
5. The menu is buried behind an ordering system
The categories Soup Noodles, Lagman, Fried Noodles, Dumplings are only accessible through the online ordering flow, not as readable HTML text on the website.
AI crawlers can’t read dynamic ordering pages. The menu that would prove this is a noodle restaurant, the exact content that would match a search for “hand-pulled noodles Forest Hills” is invisible to every bot that visits the site.
What an AI-Ready Site Would Look Like
If Old Sport Food were a Shop Locally client, here’s what would change:
- Rewrite the homepage copy to lead with “halal hand-pulled noodles” and “Lanzhou beef noodle soup”. The exact phrases people search
- Fix location signals everywhere: Forest Hills, Queens, Austin Street, 11375. The domain name alone is costing them recommendations
- Add a static HTML menu page with dish names and descriptions that Google and AI bots can actually read
- Allow crawlers: remove the 403 block or at minimum allow GPTBot and ClaudeBot through
- Add an About page that tells the restaurant’s story in natural language, the Xinjiang-style halal tradition, the hand-pulling process, what makes this spot different from the kebab shop down the block
None of this requires a viral campaign. None of it requires ads. It requires structured, readable, accurate information on the restaurant’s own web presence.
The 4.7 rating and 408 reviews mean the product is excellent. The invisibility is entirely a web presence problem and a digital presence problem is exactly what Shop Locally fixes.
If you know someone running a spot like this, somewhere great that AI should know of, send them our way.
Get a free AI Readiness Audit →
The Original Research: ChatGPT on Halal Noodles in Queens
This is the actual conversation that started it all — the ChatGPT session where Old Sport Food didn’t appear once.