Build log
Most rank-and-rent advice starts with a list of "best niches". I don't trust those lists, because a niche is only good in a particular city. So I had an AI agent score trade and city combinations instead, and it wrote down why it kept or dropped each one. This is the method behind my two live sites, medfordseptic.com and roanoketreeremoval.com. Neither has a renter yet, revenue is $0, and Search Console numbers aren't recorded yet, so this is a method, not proof that it works.
Step 1: shortlist trades, not keywords
A trade makes the shortlist when it has at least two of these:
- Recurring demand, like pumping or cleaning
- Emergency demand, like storms, backups and leaks
- High-ticket jobs, like repairs and replacements
- Lots of small operators, any of whom could rent the site
The last one matters most for rank and rent. If there are no small operators, nobody will rent the site.
Step 2: shortlist cities you can picture
Mid-size metros and rural-edge areas work best. In big metros, national brands usually own the results already. I also stick to places where the agent can find enough local detail to write area pages that aren't just the same page with the town name swapped.
Step 3: let the agent tear down the top five
For each combo the agent searches "[service] [city] [state]", opens the top five organic results and scores each one 0 or 1 in a CSV:
- Thin one-page or template site
- Gmail or Hotmail contact address
- No dedicated service pages or city pages
- Typos or leftover template text
- A directory (BBB, Yelp, Angi) ranking among the businesses
- Strongest competitor based in another town
There's also one red flag that rejects a combo outright: national franchises or big brands holding most of the top results. Every row gets a URL and a checked date, so I can open the result myself before agreeing with the score.
What it decided on 10 Oct 2026
- Foundation repair, Fort Wayne IN: rejected. The results were held by Groundworks, Basement Systems of Indiana, EverDry and Acculevel. It looked like a good trade, but that city was the wrong one for it.
- Septic pumping, Medford OR: picked. The top five were small local operators. Weak spots included a leftover "send us bug reports" footer, Gmail addresses, a typo in the city name and no city pages. The strongest competitor was an excavator based in Grants Pass, for whom septic is a side line.
- Tree removal, Roanoke VA: picked. The area is heavily wooded and gets ice and wind storms. The sites were thin and used Gmail addresses, a BBB listing ranked among them, and the strongest competitor was based in Moneta.
Step 4: check demand without a paid tool
I didn't pay for a keyword tool. The agent collects Google Autocomplete suggestions for a few seed phrases, saves the raw output and rates each keyword High, Medium, Low or Very low by how often it comes up. These are relative signals, not search volumes, and the file says so at the top. If you're about to spend real money, check the volumes in a proper tool first.
Look out for city names that exist in more than one state. If the suggestions drift to another state, put the state in every title.
Step 5: list possible renters before you build
The agent lists at least five local operators with public URLs and phone numbers. I don't contact them yet; I won't pitch anyone until I have impressions or logged calls to show. The list just proves there's someone who could rent the site one day.
Step 6: check the domain, then stop and approve
The agent checks domain availability with an RDAP lookup and records the date. I prefer [city][service].com or the other way round. Buying is gated, so I approve the spend myself. When I priced it, a domain at cost was about $10.46 a year.
Where this leaves me
- Combos picked: 2. Rejected on the franchise flag: 1 (Fort Wayne foundation repair).
- Sites built: 2, at 32 pages each
- Search Console impressions and clicks: none recorded yet
- Renters: 0. Revenue: $0.
The scorecard narrows down where to build. It can't tell you whether a site will rank. Only months of waiting will tell me that.