Overview
Greg Isenberg hosts directory specialist Frey Chu for a deep technical walkthrough on using Claude Code to build data-rich niche directories. The episode opens with a game where Frey reveals traffic and monetization details for three real directories -- a funeral home comparison site generating over 60,000 monthly visitors, a senior living directory attracting 824,000 visitors per month, and GasBuddy pulling 1.1 million monthly visits with a MasterCard debit product as its primary revenue engine. These examples establish a core thesis: the most durable directories help people save time, save money, or make money, and price transparency in opaque industries is a massive untapped opportunity.
The heart of the episode is Frey's seven-step process for building a luxury restroom trailer directory from a raw CSV of 71,000 scraped Google Maps listings. He walks through removing junk data with Claude Code, installing Crawl4AI as an automated web crawler, using it to verify and classify businesses by website content, and then layering enrichment passes to capture trailer inventory, images, amenities, and service areas. The full build took four days and cost under $250 -- roughly $100 for a Claude Code Max subscription, $100 for data acquisition, and $50 in Claude API credits for image verification via Claude Vision. The same work done manually would have taken an estimated 2,000-plus hours.
Frey also addresses the business case honestly. Directories are not a quick-money play; SEO takes months to build authority. But for anyone wanting to learn Claude Code, SEO, and lead generation simultaneously with a low-cost structured project, the directory model provides a practical and high-leverage vehicle. The episode closes with a discussion of how AI search and LLMs affect the directory model going forward, with Frey arguing that niche directories serving users in the decision-making phase -- particularly in high-stakes categories like health, legal, and finance -- remain structurally strong even as discovery shifts toward LLM-based interfaces.
Why This Matters
The workflow Frey demonstrates is genuinely novel in its economics. Pre-Claude Code, enriching 725 business listings with inventory, images, amenities, and service areas required either months of manual work or expensive developer-built scraping pipelines. The combination of Claude Code for logic and prompting, Crawl4AI for concurrent web crawling, and Claude Vision for image classification reduces this to a multi-day automated process costing tens of dollars in API credits. That cost compression changes what is viable at the small-operator level.
The data moat concept is worth internalizing beyond directories. Any vertical where publicly available data is messy, scattered, or non-comparable creates an opportunity for someone willing to do the enrichment work. Price transparency in boring service niches -- porta potties, ADA contractors, water quality, restroom trailers -- has barely been touched. The same methodology applies to any business model requiring a high-quality proprietary dataset as its foundation.
The directory-to-vertical-SaaS progression shown by Parting.com (funeral directory to cremation software) is a recurring pattern worth tracking. Traffic and trust built through a useful directory creates a natural pipeline for a paid product serving the same businesses listed in the directory. That is a structural advantage: the directory builds the audience, and the SaaS monetizes the relationship. For anyone building in a niche with ongoing business relationships, this upgrade path is worth planning from day one.
Key Points
- Data is the moat. Successful directories are built on proprietary, enriched datasets that competitors cannot easily replicate. The quality and depth of your data determines whether a directory gets traffic and generates leads.
- Every durable directory helps users save time, save money, or make money. Price transparency in opaque service industries is one of the highest-value things a directory can provide because it is genuinely difficult to obtain at scale.
- Frey built a fully enriched luxury restroom trailer directory in four days for under $250, replacing a process that would previously have taken over 2,000 hours of manual work. The cost breakdown was $100 for Claude Code Max, $100 for data, and $50 in Claude API credits.
- The seven-step data pipeline goes: scrape raw data (Outscraper), clean obvious junk (Claude Code), verify listings by website content (Crawl4AI), enrich with product inventory, enrich with images (Claude Vision), enrich with amenities and features, enrich with service areas.
- Run enrichment passes one type of data at a time. Trying to get Claude Code to extract everything in a single pass produces low-quality results. Sequential passes with edge-case corrections between each pass produce clean, reliable data.
- Crawl4AI is a free, open-source LLM-friendly web crawler that runs locally and can crawl multiple websites concurrently. Combined with Claude Code, it automates the manual verification work that was previously the biggest bottleneck in directory building.
- Niche specificity is the SEO strategy. Instead of competing in broad categories like senior living or bathroom contractors, target specific sub-niches such as senior living for people with dementia or ADA accessible bathroom contractors -- where you can realistically rank and where the niche-specific criteria are deal breakers for users.
- Directories serve users in the decision-making phase, not the discovery phase. This makes them structurally different from content sites and means they remain relevant even as LLMs take over initial search and discovery for many queries.
- The directory-to-SaaS upgrade path is one of the strongest monetization progressions available. Building a directory establishes trust, audience, and relationships with businesses in the niche, which creates a natural pipeline for a paid vertical software product.
- Lead generation from directories works even with a low-quality product. Frey's original porta potty directory still had lorem ipsum on the homepage and generated leads including a $20,000-plus order from the New Mexico State Fair. Strong niche demand compensates for weak execution in the early stages.
- Directories are not a fast-money strategy. SEO takes time, especially without an established domain and backlinks. Anyone who needs revenue in under six months should look elsewhere. The payoff is in the medium term once organic rankings mature.
- In an AI search world, niche directories that return a handful of relevant results are better positioned than broad horizontal directories, because LLMs delivering three answers will reference the most specific and trusted source they can find for a given query.
Quotable
AI-generated from source material. Verify important details against the original source.
Frey Chu
"I've seen probably 1,000 or 2,000 people build directories in the last year. And 100% this is where people either quit or give up. And I totally empathize. There is a point in my life where I was waking up and being like, 'All right, I'm going to sit down in this chair for 6 hours and click on this website and just manually verify if this is a luxury restroom trailer.' And it was hell."
Grounds the technical solution in lived experience. Anyone who has done manual data work immediately understands the pain being solved, which makes the Crawl4AI reveal land harder.
Frey Chu
"If your timeline is to make money in less than 6 months, I would not build a directory. With that said, I don't know if there's a better playground and structured approach to a project where you can learn how Claude Code works and learn SEO, and learn how to sell, and get into the business of lead generation."
Unusually honest for a creator in this space. Counterbalancing the excitement of the demo with a realistic timeline expectation increases trust and gives listeners a clearer decision framework.
Greg Isenberg
"In a world where people are finding products and services through LLMs like Perplexity and ChatGPT and Gemini, instead of getting a thousand little blue links on Google, you're getting three. So if I'm building a directory this year, I'm trying to build it super niche so I can play into that new AI search era."
Reframes AI search as an argument for niche directories rather than against them. The insight that LLM responses reduce the result set to a handful of sources makes ranking in a highly specific niche more valuable, not less.
Frey Chu
"Data can be a moat, it can be a differentiator, and price transparency is a big one because it's so hard to get. There's so many industries and niches where price transparency just doesn't exist. Tons of boring ones."
Condenses the core opportunity thesis into two sentences. Boring niches with opaque pricing are underserved by existing media and directories, which makes them fertile ground for a new entrant willing to do the enrichment work.
Frey Chu
"I would say the top three ways to get distribution nowadays are ads, organic social, and SEO. Like those are the big three. I made the mistake of not focusing on distribution first, and directories are a distribution first model."
Positions directory building not as a niche business strategy but as a distribution acquisition strategy. This reframe is useful for anyone who has struggled to get eyes on their products and is looking for a structured approach to building organic traffic.
Concepts
Core Frameworks
The Data Moat
The single most important differentiator for a durable directory is the quality and exclusivity of its underlying data. Traffic and backlinks matter, but a competitor can eventually match your SEO. Proprietary enriched data is harder to replicate, especially when it was acquired through a significant investment of time, automation, or specialized access.
Price transparency is the highest-value form of data enrichment because it is genuinely difficult to obtain at scale and is a primary decision driver for buyers. Any niche where pricing is opaque, variable, or scattered across dozens of websites is a candidate for a directory that aggregates and normalizes that information.
The Three-Value Framework
Every successful directory examined in this episode helps users do one of three things: save time, save money, or make money. This framework is simple but useful as a filter when evaluating niche ideas. A directory that does not clearly serve at least one of these functions will struggle to generate return visits and referral traffic.
GasBuddy saves money on gas. A Place for Mom saves time finding senior living options. Parting.com saves time and potentially money when arranging funeral services. The luxury restroom trailer directory helps event planners save time comparing vendors and ultimately save money by identifying the right product for their event size.
The Directory-to-SaaS Upgrade Path
Parting.com illustrates a recurring pattern where a directory builds audience and trust in a niche, then uses that position to launch a vertical SaaS product sold to the businesses listed in the directory. The directory attracts users through organic SEO; the SaaS monetizes the relationship with businesses who want to reach those users more effectively or need operational software.
This combination is described by Frey as one of the most potent monetization structures available. The agency-directory combo (where the directory owner also offers marketing and lead generation services to listed businesses) follows the same logic. Traffic and trust are the assets; the specific product built on top of them can evolve.
Strategies
Niche Specificity as SEO Strategy
Competing against established directories like A Place for Mom or Angie's List in broad categories is impractical without major budget and time investment. The viable alternative is to take a sub-slice of the same category -- senior living for people with dementia, ADA accessible bathroom contractors, luxury restroom trailers rather than porta potties broadly -- and build deep topical authority in that micro-niche.
The SEO advantage is real: a directory with hundreds of pages all focused on a single specific niche builds topical relevance quickly, which helps it rank for location-specific queries (luxury restroom trailer Bakersfield, California) before gradually expanding to broader terms. The niche focus also means every user who arrives already has highly specific intent, which improves conversion rates for lead generation.
Sequential Enrichment Passes
Attempting to extract all data points from business websites in a single crawl pass produces poor results. The correct approach is to run separate passes for each data type -- first verifying whether a business belongs in the niche, then extracting product inventory, then images, then amenities, then service areas -- and to review the output after each pass before continuing.
Edge cases appear at every stage. After a pass, you will find unusual data patterns that need to be fed back to Claude Code as corrections before the next run. This iterative quality control is what separates a clean dataset from a messy one, and it cannot be shortcut by trying to get everything in one prompt.
Crowdsourcing as a Data Strategy
GasBuddy illustrates a different path to data moat: user-generated content incentivized by gamification. Their 2004-era public leaderboard and e-bike giveaways created a self-sustaining data collection engine that became extremely difficult to replicate. No competitor could just launch a gas price aggregator and instantly have 5,800 consecutive days of contributions from their top reporter.
For directory builders, crowdsourcing is worth considering in niches where data changes frequently (prices, hours, availability) or where community members have natural motivation to contribute. The initial incentive structure -- points, leaderboards, giveaways -- does not need to be expensive, but it needs to be intrinsically aligned with why users are on the site.
Mental Models
Discovery Phase vs. Decision-Making Phase
LLMs like ChatGPT and Perplexity are excellent at the discovery phase -- helping someone understand what options exist, what questions to ask, what factors matter. But by the time a user arrives on a directory, they are typically in the decision-making phase: they know what they want and are comparing specific options to make a choice.
This distinction matters for evaluating the AI search threat to directories. For any complex, high-stakes decision -- choosing a senior living facility, hiring a lawyer, renting specialized equipment for a large event -- users will do their own comparative research and will not simply accept the first LLM-generated recommendation. Directories that serve users in this phase remain structurally important.
Directories as Distribution-First Businesses
Most businesses build a product and then try to figure out distribution. Directories invert this by building distribution first and then figuring out what to sell to the audience they have assembled. The organic SEO traffic that a well-built niche directory generates is itself the asset, and it can be monetized in multiple ways: ads, lead gen fees, SaaS, agency services, affiliate commissions, marketplace transactions.
This makes directories a lower-risk way to enter a niche than launching a product directly. You validate demand through traffic and inbound inquiries before committing to a specific product or offer.
Claude Code as a Non-Technical Operator Tool
Frey emphasizes repeatedly that he is not technical and that this is intentional, not a disclaimer. Claude Code allows non-technical operators to direct complex data processing tasks through natural language prompts written in conversational style. He writes prompts by talking through the problem with ChatGPT first, then feeds the resulting prompt to Claude Code to execute.
The practical implication is that the bottleneck in directory building is no longer coding ability. It is clarity of thinking about what data you need and what questions you need answered at each stage of the enrichment process. Domain knowledge about the niche matters more than technical ability.
Warnings
Image Scraping Legal Risk
Scraping images from business websites without explicit permission carries legal risk. Frey acknowledges this directly and notes his plan to contact businesses and ask them to claim their listings, which provides retroactive permission. He also notes that directories can rank and generate leads without images, as his original lorem-ipsum porta potty site demonstrated.
For anyone building in this space, the safer paths are using stock images, reaching out to businesses proactively for media, or using publicly licensed images. The image enrichment step Frey demonstrates is valuable as a proof of concept but should be approached with awareness of the risk before publishing.
SEO Timeline Realism
Directories are a medium-term investment. Building the data asset and the website is the easy part now that Claude Code has collapsed the cost and time. Building domain authority and ranking for competitive-enough keywords to generate meaningful traffic takes months, sometimes longer, especially on a new domain without backlinks.
Anyone who needs revenue quickly should not start a directory. The model rewards patience and is best suited to operators who can treat the first directory as a learning project and build systems that compound over time.
Implementation
AI-generated from source material. Verify important details against the original source.
Choose a niche using the three-value filter
Identify a niche where your directory will help users save time, save money, or make money. Prioritize niches with opaque pricing, where comparison is genuinely difficult today, and where the category is specific enough that you can realistically build topical SEO authority without a massive backlink budget. Use Reddit threads, TikTok comments, and Facebook groups to understand what decision-making friction actually looks like for people in the niche. The ideal niche is boring, local-service-oriented, and has no obvious dominant directory already.
Scrape raw data from Google Maps using Outscraper
Use Outscraper to pull a large CSV of businesses in your niche across your target geography. For national directories, run the scrape across all states or provinces. Expect a large and messy dataset -- Frey started with 71,000 rows for the restroom trailer niche. This is expected. The volume is the raw material. Keep all the data at this stage and do not attempt manual filtering. The goal is completeness, not cleanliness.
Remove obvious junk with Claude Code
Write a Claude Code prompt specifying clear removal criteria: listings with no business name, no address, permanently closed listings, and obvious category mismatches (big box retailers, unrelated businesses). Give Claude Code all your CSV files and ask it to apply the criteria across all of them. This pass should reduce your dataset by 60-75% without any manual review. Frey went from 71,000 rows to 20,000 with a single prompt pass.
Install Crawl4AI and verify listings by website content
Install Crawl4AI locally (it is free and open-source). Give Claude Code the Crawl4AI documentation link and ask it to help you set up the tool -- Frey reports this took about 15 minutes. Then write a prompt instructing Claude Code to visit every business website in your remaining CSV and identify which ones actually serve your niche. Include a list of all synonyms and related keywords for your niche so the crawler knows what to look for. Enable the async web crawler module for concurrent crawling to reduce run time. This pass will run for several hours but replaces what would have been weeks of manual verification.
Run sequential enrichment passes for each data type
Do not try to extract all data in one pass. Run separate Crawl4AI passes for each data type: product inventory first, then images, then amenities and features, then service areas. After each pass, review the output for edge cases and unusual results. Feed corrections back to Claude Code before running the next pass. Expect to rerun each pass two or three times before the data quality is acceptable. This iterative correction process is what separates a polished directory from a low-quality one.
Use Claude Vision to filter and rank images
After scraping candidate images from business websites, pass the top three candidates per listing through Claude Vision via the Claude API to identify the best image for the listing. This step filters out logos, favicons, and low-quality images that a basic scraper will collect alongside real product photos. Connect your Claude API key, write a prompt describing what a good listing image looks like for your niche, and let it run overnight. Budget roughly $30 for 700 listings. Only proceed with images where you have permission or intend to contact businesses to obtain it.
Build the directory site from the enriched dataset
Strip the enriched dataset down to only the columns that matter for the directory. Give Claude Code the cleaned CSV and ask it to create a Supabase database using the exact column structure. Then use Claude Code to build the front-end directory site -- listing pages, filter functionality, search, lead forms, and service area displays. At this stage, the hard work is done. The quality of your data determines the quality of your directory. The build itself is the easier half of the project.
Choose your monetization model based on the niche
Match your monetization to your niche rather than forcing a generic model. Lead generation (charging businesses per referral or per first month's revenue) works in high-ticket service categories. Display ads work when traffic volume is high enough to qualify for Mediavine or similar networks. Affiliate commissions work in product niches. A vertical SaaS built on top of the directory works when listed businesses have a shared operational need you can solve. The directory can evolve its monetization over time as you learn which businesses respond to outreach and what problems they have beyond just the directory listing itself.
Tools & Resources
Mentioned Resources
| Resource | Description |
|---|---|
| Outscraper | Google Maps scraping tool. Used to pull large CSVs of business listings by category and geography. Frey recommends it as the cheapest and most straightforward option for initial data acquisition. |
| Crawl4AI | Free open-source LLM-friendly web crawler and scraper. Runs locally, supports concurrent crawling via the Async Web Crawler module, and integrates directly with Claude Code for automated website verification and data enrichment at scale. |
| Claude Code | Anthropic's agentic coding tool. Used throughout the workflow for data cleaning prompts, directing Crawl4AI, image filtering via Claude Vision API, and building the directory site itself. Frey uses a Claude Code Max subscription ($100/month). |
| Claude API | Used specifically for the Claude Vision image classification pass. Frey spent approximately $30 in API credits to classify images for 700 listings. |
| Supabase | Open-source database backend used to store the enriched directory dataset. Claude Code generates the database schema directly from the enriched CSV column structure. |
| ShipYourDirectory | Frey Chu's directory-building resource. A reference for those wanting to go deeper into the directory niche business model. |
| Data.gov | US federal government public data portal. Frey highlights it as a source of underutilized public datasets that can form the basis of directories in niches like water quality and air quality. |
| Ahrefs | SEO tool referenced for validating monthly search volumes on niche keyword targets. Used in the episode to show that a term like "senior living homes for people with dementia" generates meaningful organic search volume. |
| Mediavine | Premium display advertising network. Referenced in the context of a tap water quality directory reaching 40,000 monthly visitors and qualifying for Mediavine ad monetization. |
Suggested Resources
| Resource | Description |
|---|---|
| Frey Chu on YouTube | Frey publishes one directory-focused video per week. The most directly relevant follow-on resource for anyone implementing this workflow. |
| Apify | Alternative to Outscraper for web scraping. Mentioned briefly by Frey as a viable option, though he continues to prefer Outscraper for cost and simplicity in Google Maps data extraction. |
| SimilarWeb | Traffic and competitive intelligence tool useful for validating directory niche ideas by estimating existing traffic levels for comparable directories before committing to a build. |
| Idea Browser | Greg Isenberg's tool for identifying startup trends and ideas. Relevant context for sourcing niche directory ideas based on emerging search and market trends. |
| WordPress | A Place for Mom, one of the largest directories in the world, runs on WordPress. For operators who are not ready to use Claude Code to build a custom stack, WordPress with a directory plugin remains a valid starting point validated by large-scale examples. |
Source Material
[00:00]
So you want to start a startup that does anywhere between $2,000 and $10,000 a month passive revenue. Something that only you have to work on 10, 15, 20 minutes a week. But you don't have a lot of money to start. Maybe you have $200, $500, $1,000. Well, what kind of business would you start? I would recommend an online directory. A directory is basically a website with a bunch of different links and data. And in this episode, I talk about how you can use Claude code to actually build an online directory that is going to get traffic on autopilot. And I brought on Frey, and Frey is Mr. Directory. And he shows us step-by-step how to use Claude code to create a very big directory. I really enjoyed having him just take us through the entire process. What are the prompts he uses? How to think about data enrichment? How do you think about cleaning the data? So that you can create a online directory that generates traffic every single day. Now I know what some people are thinking. They're thinking, "That's really boring. An online directory? Come on. I want to build a mobile app. I want to build a SaaS product." But what they miss is that building an online directory is going to get you thousands of visitors on autopilot. And once you have that, then go and buy Code a SaaS product. Then go and buy Code an app. In the beginning of this episode, we share multiple examples of people making millions of dollars a year using online directories. And some of them have apps that go with it, too. So, this is an incredibly important, yet boring, episode of the podcast. But those who stick to the end, I think, are going to have an unfair advantage.
[02:15]
By the end I really want people to have a solid understanding of how to use Claude Code to build AI coded directories with an emphasis on what I think is the hardest part, which is getting the valuable data. And we're going to go deep on that. But first, I think it's only fair we run it back. The last time I was on the pod, we played a game where I showed you a few cool directories, and you guessed the traffic. So, if you're down, I'll share three cool directories. And if you want, you can try to guess the way that these monetize as well.
[03:00]
So the first one is parting.com, and we can see it's a funeral home directory. How much traffic do you think that this website makes? Greg guesses at least 100,000 visits a month. It's actually 61,000 organic monthly visitors. And this is reported to do over a million dollars a year in revenue. There are some quotes that go up to $5 million, but they even raised money -- $1.5 million in 2021. The bulk of their money actually comes from a software called Parting Pro, which is cremation arrangement software for funeral homes. The agency directory combo is a potent one, and I see it all the time. Some of the biggest directories that I know monetize this way.
[05:42]
The second one is aplaceformom.com for senior living. It is getting 824,000 organic monthly visitors. The estimates on revenue range from $100 million to $500 million a year; a gut estimate of $50 million a year seems more likely. The way they monetize is basically purely from lead generation. They have over 18,000 senior living homes listed, and they collect a portion of the first month's rent or a fixed fee for every referral. Just a beastly, but very simple and straightforward directory. It's also built on WordPress.
[08:00]
The last one is gasbuddy.com -- a crowdsourced directory to give you gas prices near you. It is getting 1.1 million organic monthly visitors and they monetize in two main ways. They make a ton of money on ads -- probably $30,000 a month on the lower end just in ads. But their big play is actually a debit card partnered with MasterCard that allows you to save money on gas. It's $10 a month or $99 a year. Back in 2004, all they did was create a public leaderboard and did giveaways -- e-bikes in exchange for gas price reports. That got a self-sustaining contributor base. Their top reporter, Prince33 from Gary, Indiana, hasn't missed a day in over 5,800 days.
[12:32]
All three of these directories execute a framework that I think all successful directories do, which is they help people save time, save money, or help people make money. And price transparency is a massive opportunity for directories. It's incredibly unscalable and manual, but there's so many industries and niches where price transparency just doesn't exist. Tons of boring ones. Data can be a moat, it can be a differentiator, and price transparency is a big one because it's so hard to get.
[14:02]
I built this directory in 4 days. And it literally started from a raw, massive, unorganized CSV of data of over 70,000 rows that I scraped for the entire country for porta potty suppliers. My Claude Code directory covers luxury restroom trailers. Boring, but awesome niche. People are spending $1,000, $2,000 a day renting these out for their weddings -- corporate events, film sets, really all sorts of stuff.
[15:52]
A year and a half back when I didn't know Claude Code I created portapotty match.com in WordPress. It still has lorem ipsum on the front page. All these images are the same for every listing. It's a bad directory, ugly and bad. But to my surprise, I just built this and left it up and got some leads coming in. One guy named Eric wanted a restroom trailer for a film shoot. And the big one was a lead from the New Mexico State Fair -- an order that would have been $20,000 plus of porta potty and luxury restroom trailers. So despite really poor data curation, it showed me there is a massive need in this space.
[19:04]
I built this in under 250 bucks. $100 of that was my Claude Code Max subscription. Another $100 was for the data. And then $50 for some Claude API credits to do some deep cleaning. I probably saved over 2,000 hours in what would have been manual data cleaning, manual data enrichment. Also, record low costs. I would have had to hire a developer to write custom Python scripts before and it is not necessary nowadays.
[21:23]
The first step is just scraping the data. I went straight to Outscraper. Outscraper is still the cheapest option for me. I got 71,000 rows and potential listings, and covered the entire state. I build nationwide directories, and the whole game plan is to sell leads in this high-ticket niche. The problem is 71,000 rows is massive, and we definitely have to do a lot of cleaning.
[22:25]
This is when I introduce Claude Code into the mix. I'm just telling Claude Code, "Here are my five CSVs. Go ahead and just look at every single one, and use this criteria to clean the data." Remove things like listings with no business name, address, city, state, permanently closed ones, and any obvious ones that don't relate to my niche, like big box retailers. This simple prompt got me from 71k down to 20,000 listings.
[23:27]
The next round is where it gets more interesting. I installed Crawl for AI. This is an open-source LLM-friendly web crawler and scraper, totally free, which is insane. This is something that you just install locally to your computer. I'm brand new to AI coding, only 6 months in. I just took the documentation link, gave it to Claude Code, and told it to help me install it. It took about 15 minutes. Now Crawl4AI is the engine that allows us to at scale look at every single website. And then Claude Code is the brain, so we can prompt Claude Code and say, "Look at every single one of these 20,000 websites, and just go and identify the luxury restroom trailers."
[28:01]
There's a ton of edge cases that you want to check after every single pass. One of the first mistakes that I made was I would just give it a massive laundry list of things to get and a massive CSV -- get the trailer inventory, get the images, get any amenities, features, pricing, all of that -- and it just didn't work. It was super low quality. So that's why I'm going slowly one at a time. After each pass, I examine the results, find edge cases, let Claude Code know it done messed up so it can fix it, and rerun. I rerun this I think two or three times until pretty much all of my data was good.
[31:33]
The result was 20,000 websites down to 725 luxury restroom trailers. It even gives us like verification confidence. That is nice. For the next step, trailer inventory, I used Crawl4AI in the same way but this time instead of looking for keywords like restroom trailer, I told it to look for the full fleet of products that a business offers. For me that's stall configurations -- two-stall and three-stall trailers are good enough for a 50 to 100-person wedding, but some people are looking for more.
[35:02]
For trailer images: I just told Claude Code to look for any alt text, file names, examine the page that it's on and scrape the highest quality images. Then I told it to take the top three candidates for trailer images and send them over to Claude Vision, so it was able to look at these images and identify the best ones. Because the first time I did this I got logos and crappy images and favicons. Claude Vision cost about $30. All I had to do was go to Claude and connect my API key. This took probably about an hour or so for 700 listings.
[36:33]
I did this with the amenities and features. Basically just mentioned that the first time I ran this, it didn't do a good job -- there were a bunch of weird words like "it" and "and" and "the" showing up as features. But once I started understanding the edge cases and said, I want you to go to the homepage, look for any page with restroom trailers, and go deep -- it was able to create these cool amenities. And this is what it came up with -- tons of different features that would ultimately become the filter on my directory. So people can choose by stall, or click "running water" and see all the luxury restroom trailers with that specific amenity.
[38:31]
The last one was the service areas. A lot of people want to know if a place can drive out 100 miles to a remote location for an outdoor wedding. The only challenge was that some businesses were listed in different states. If a business was based in Florida, sometimes the first run would show me Florida, Texas, Arizona. So I had to make some adjustments. But these are the new columns it created -- I broke it down by city, region, and radius.
[39:15]
This kind of opened my mind to the types of directories I can build. Unless you're an expert SEO with unlimited budget for backlinks, you're not going to rank for senior living homes. But you might rank for senior living homes for people with dementia. Bathroom contractor -- super competitive, Angie's List and HomeGuide are everywhere. But what about an ADA accessible bathroom contractor? Way more specific. Public databases like data.gov have an unreal amount of data. A guy named Andy in our community built a tap water quality directory with no backlinks, started it in November, and is getting over 40,000 monthly visitors. He got accepted into Mediavine and sells water filters through Amazon affiliate.
[43:38]
For the people that are naysayers about directories -- if your timeline is to make money in less than 6 months, I would not build a directory. I've been building them for a few years and I just want to save people the time. With that said, I don't know if there's a better playground and structured approach to a project where you can learn how Claude Code works and learn SEO, and learn how to sell, and get into the business of lead generation. The way that I think about directories is that it is still to this day a playground for me to learn how to get people's eyes on my projects. I made the mistake of not focusing on distribution first, and directories are a distribution first model.
[47:51]
In a world where people are finding products and services through LLMs, I would say I have two points. First -- if someone's browsing a directory in 2026, they're sort of past the discovery phase. LLMs help you discover things, but by the time someone's on a directory, they're in the decision-making phase. For senior living homes, you're not just going to go on ChatGPT and choose the first one. You're going to compare all the options near you. Pricing is another thing -- anytime someone can save money, that drives people to really go out of their way and look at directories. Second -- local SEO hasn't really changed that much. If we go to "haircut Los Angeles" in Google, you still get the local pack and organic listings. So from an SEO perspective, any kind of local SEO is still pretty low-hanging fruit. And in an AI search world where you're getting three results instead of a thousand blue links, being the niche directory that gets referenced is more valuable, not less.
AI Prompt
AI-generated from source material. Verify important details against the original source.
AI Implementation Prompt
CONTEXT This prompt is based on a YouTube interview between Greg Isenberg and Frey Chu (published February 16, 2026) titled "Claude Code built me a $273/Day online directory." Frey Chu is a niche directory builder who has been building and studying online directories for several years. The episode demonstrates a complete seven-step system for using Claude Code and Crawl4AI to build a data-rich niche directory in four days for under $250 -- a process that previously would have required over 2,000 hours of manual work. The core thesis is that data is the moat for any durable directory business. The most successful directories help users save time, save money, or make money -- and price transparency in opaque service niches is the highest-value opportunity because it is genuinely difficult to obtain at scale. The episode uses three real-world examples (Parting.com at $1M+ revenue, A Place for Mom at $50M+ estimated revenue, GasBuddy at 1.1M monthly visitors with a MasterCard debit product) to illustrate what is possible, then walks through Frey's full build of a luxury restroom trailer directory using Outscraper for raw data, Claude Code for cleaning and orchestration, Crawl4AI for automated website verification, Claude Vision API for image classification, and Supabase for the database backend. KEY PRINCIPLES 1. Data quality is the primary competitive advantage. Enriched, proprietary data is harder to replicate than SEO rankings or ad spend. 2. Price transparency in boring, local-service niches is a high-value and underserved opportunity that most directory builders ignore. 3. Every durable directory helps users save time, save money, or make money -- if a directory does not clearly serve one of these functions, it will struggle to build return traffic. 4. Sequential enrichment passes produce better data quality than trying to extract everything in a single crawl. Run one pass per data type, review edge cases, correct the prompt, and rerun before moving to the next data type. 5. Claude Code combined with Crawl4AI eliminates the manual bottleneck that previously made data enrichment at scale impractical for solo operators. 6. Niche specificity is both an SEO strategy and a product strategy. Narrow enough to rank, specific enough to be a deal-breaker filter for your target user. 7. Directories serve users in the decision-making phase, not the discovery phase. This structural position remains valuable even as LLMs change how people discover products and services. 8. The directory-to-vertical-SaaS upgrade path is one of the strongest monetization progressions available: traffic and trust from the directory create a natural pipeline for a paid product. 9. Directories are a distribution-first model -- build audience and organic traffic first, then figure out what to sell to that audience. 10. SEO takes time. If you need revenue in under six months, a directory is not the right business model. KEY LEVERS - Data acquisition cost and quality (determines whether the directory can compete) - Niche specificity (determines whether you can rank and whether users with high intent arrive) - Enrichment depth (determines listing quality, user trust, and filter functionality) - Monetization model alignment with niche (lead gen, SaaS, ads, affiliate, marketplace) - Domain authority and backlink strategy (determines how long it takes to rank) - Distribution channel beyond SEO (social, community, personal brand) WHAT THIS IS NOT - This is not a quick-money system. SEO takes months to compound, especially on a new domain. - This is not a product scraping strategy. Frey explicitly avoids product-based niches because they compete with retail giants and social content for SERP real estate. The approach works best for local service businesses. - This is not a strategy that requires coding ability. Claude Code allows non-technical operators to direct complex data pipelines through natural language. The bottleneck is clarity of thinking, not technical skill. - This is not a replacement for business fundamentals. Traffic without a clear monetization model that matches the niche is just a hobby. - This is not primarily an AI content strategy. The directory earns SEO authority through data depth and topical relevance, not AI-generated articles. IMPLEMENTATION MODES 1. NICHE SELECTION -- Help me evaluate potential directory niches against the three-value filter (save time, save money, make money), assess price transparency opportunity, estimate search volume and competition, and identify whether the niche is local-service-based with realistic ranking potential. 2. DATA PIPELINE DESIGN -- Help me design a sequential enrichment pipeline for my specific niche: what data types to extract in what order, what criteria to use for verification passes, what edge cases to anticipate at each stage, and how to structure my Outscraper scrape. 3. PROMPT WRITING -- Help me write effective Claude Code prompts for each stage of the pipeline: junk data removal, Crawl4AI verification, inventory extraction, image scraping, amenity extraction, and service area mapping. 4. EDGE CASE TROUBLESHOOTING -- Help me diagnose and fix data quality problems that appear after an enrichment pass: unusual values, missed records, category mismatches, geographic errors in service area data. 5. MONETIZATION STRATEGY -- Help me identify and evaluate the monetization model best suited to my specific niche: lead generation fee structures, vertical SaaS opportunities, display ad viability, affiliate commissions, marketplace model, or agency services. 6. NICHE IDEA GENERATION -- Help me generate specific, viable directory niche ideas using the frameworks from this episode: boring local service categories with opaque pricing, sub-niches of competitive categories where I can rank, and public dataset opportunities (data.gov and similar). 7. SEO AND CONTENT STRATEGY -- Help me build an SEO strategy for a new directory: keyword targeting by location and specificity, topical authority building, internal linking structure, and realistic timeline expectations. 8. AI SEARCH POSITIONING -- Help me evaluate how AI search and LLM-driven discovery affects my specific directory niche, and how to position the directory to benefit from reduced result sets in LLM responses. 9. TECH STACK DECISIONS -- Help me evaluate options for the directory build: when to use Supabase vs. WordPress, how to structure the database from an enriched CSV, and what front-end features matter most for user trust and conversion. 10. COMPETITIVE ANALYSIS -- Help me analyze existing directories in my target niche to identify data gaps, monetization approaches I could replicate or improve on, and positioning opportunities. AI OPERATING INSTRUCTIONS Stay grounded in the frameworks and practical details from this episode. Avoid generic SEO or business advice that could apply to any online business. When helping with niche selection, push for specificity -- broad categories are not actionable. When helping with data pipeline design, insist on sequential passes rather than all-at-once extraction. When evaluating monetization, require that the model match the niche rather than defaulting to display ads. Challenge assumptions about timeline if the user seems to expect fast revenue. Draw connections between directory type and optimal monetization model whenever relevant. Ask clarifying questions before making recommendations that depend on knowing the specific niche. GUIDED DISCOVERY Ask me up to three questions, one at a time, to determine: (1) what I am trying to accomplish with a directory -- building a learning project, generating passive income, validating a niche before a SaaS build, or something else; (2) which part of the directory workflow I am at or most focused on right now -- niche selection, data acquisition, enrichment, site build, or monetization; (3) what niche or category I am considering, even if only rough. Once you understand my situation, help me build a practical implementation plan.