Ask ChatGPT to find you a dog daycare in Calgary that specializes in shy dogs. It will give you a list. Names, addresses, star ratings. Maybe a phone number.
What it won't tell you is that one of those daycares does individual introductions before the first group session, keeps groups under eight, and has two spots open this week. That information exists — it lives in the owner's head. No scraper has ever captured it, and no amount of better scraping ever will.
This is the local nuance gap. And it's the reason AI recommendations for local businesses feel generic, stale, and often wrong.
What scrapers actually get
Web scraping has become remarkably good at extracting public data. A modern crawler can pull a business name, phone number, hours of operation, and star rating from a dozen platforms in seconds. Structured data markup, Google Business Profile, Yelp, TripAdvisor — the public internet is well-indexed.
The problem is that none of this is useful for intent-driven queries.
When someone asks an AI agent to "find a physiotherapist in Tokyo who speaks English and has experience with runner's knee," the agent needs more than a name and a rating. It needs to know that Dr. Kato at Shinjuku Sports Physio is a former triathlete, treats running injuries specifically, speaks fluent English, and is accepting new patients this month.
That information doesn't exist on any scrapable page. It lives in the clinic's intake coordinator's memory, maybe in a brochure in the waiting room, maybe nowhere written down at all.
The boolean problem
Online platforms reduce businesses to checkboxes. A hotel is "pet-friendly" — yes or no. A restaurant has a "private dining room" — yes or no. A mechanic offers "loaner vehicles" — yes or no.
These booleans are technically accurate and practically useless.
"Pet-friendly" could mean a grudging $50 surcharge and a 25-pound weight limit, or it could mean no weight limit, no fee, a trail-adjacent location, and an on-call vet — one of the rarest combinations in Banff.
"Private dining room" could mean a curtained-off corner by the kitchen, or it could mean a sommelier-staffed private salon that seats twelve with a prix fixe weekend menu.
The boolean tells you nothing. The sentence tells you everything. And no scraper writes the sentence.
What AI agents actually need
When an AI agent handles a local query, it's not searching — it's matching. The user has expressed intent with specificity: this kind of business, this location, these attributes, right now.
To match well, the agent needs three categories of data that scraping cannot provide:
Live deals and availability. Not what was true when the website was last updated — what's true today. The bistro running a $65 prix fixe tonight. The daycare with two spots opening Thursday. The dealership trying to move last year's SUVs before Q2. These signals expire in hours or days. By the time a crawler finds them, they're gone.
Hidden inventory. Cars not listed on the lot website yet. A hotel room block released for the weekend that hasn't hit the OTAs. The groomer who just had a Tuesday cancellation. Businesses hold inventory back from public channels for all sorts of reasons — margin protection, relationship sales, timing. This inventory is invisible to scrapers because it was never published.
Seller intent. The most valuable signal and the most impossible to scrape: what the business actually wants right now. The restaurant that needs more weeknight reservations. The electrician who's looking for commercial jobs this month. The boutique hotel trying to fill a gap next weekend. Seller intent is the difference between a cold lead and a warm match — and it changes daily.
Why better scraping isn't the answer
The instinct is to build better scrapers. Scrape more sources, scrape more often, use LLMs to extract meaning from unstructured web pages.
This doesn't work, for a structural reason: the data isn't on the web to begin with.
A sales manager's knowledge about which inventory needs to move, what kind of customer they want to attract this week, and what hidden advantage their business has over the competition — none of that is published anywhere. It's not on the website, not on the Google listing, not on any social media post. It's operational knowledge that exists only in human memory.
You can't scrape what was never written down.
The alternative: sourced intelligence
The only way to get this data is to ask for it.
At Pawlo, we go directly to businesses and collect what we call nuance — the structured intelligence that lives inside a business and differentiates it from every competitor. We turn natural language descriptions into structured, queryable data that any AI agent can fetch in milliseconds.
Here's the difference in practice. A web scraper returns this:
name: Hillside Bistro
phone: 587-555-0172
hours: Mon-Sun 11am-10pm
rating: 4.6 starsPawlo returns this:
name: Hillside Bistro
hidden_strength: "Walk-ins welcome until 9pm"
wants_more_of: "Date night reservations"
tonight_signal: "Prix fixe menu, $65/head"
expires_at: 2026-02-28One of these helps an AI agent make a generic recommendation. The other helps it make the right recommendation, right now, for this specific person.
What this means for builders
If you're building an AI agent that handles local queries — travel planning, service booking, restaurant recommendations, anything involving real businesses in real places — you have a choice.
You can scrape public data, synthesize it with an LLM, and hope the result isn't hallucinated or stale. This is what most agents do today, and it's why most local AI recommendations feel like warmed-over Google results.
Or you can fetch structured, sourced intelligence directly from the businesses themselves — live deals, hidden inventory, seller intent — and give your agent data it can actually trust.
The gap between these two approaches will only widen. As AI agents become the primary way consumers find local businesses, the agents with the best data will win. Not the best model, not the best prompt engineering — the best data.
Scraping gives you the same data everyone else has. Nuance gives you the data nobody else can get.
What this means for businesses
If you run a local business, the nuance gap is both a problem and an opportunity.
The problem: AI agents are already recommending your competitors, and they're doing it based on whoever has the best-structured data — not whoever has the best service. The bistro with a detailed, current profile gets the AI recommendation. Yours, with a stale Google listing, doesn't.
The opportunity: the bar is still low. Most businesses haven't thought about AI-readable data at all. Getting your nuance — your real differentiators, your live availability, your current deals — into a structured format that AI agents can query puts you ahead of every competitor who's still waiting for Google to crawl their updated hours.
The businesses that structure their nuance now will be the ones AI recommends first.