Before I explain why we're testing local directories, let me explain what a local directory actually is — because it's one of those things that sounds obvious until you start describing it and realize people picture very different things.

A local directory is a website that aggregates information about businesses or services in a specific niche or geography. Think Yelp, but narrower. Think Angie's List, but for a specific type of service in a specific area. The business model is usually some combination of free basic listings and paid enhanced listings — you get your name in the database for free, you pay to get the larger profile, the priority placement, the inbound lead generation.

It's a model that's been around for a long time. Long before AI. And that's part of why we're testing it.

Why established models matter when you're testing new tools.

One of our selection criteria for streams to test is what I'd call "proof of concept by history." We are not trying to invent a new category of business. We are trying to take business models that have already demonstrated they can generate revenue — that the market has already validated — and find the place where AI removes whatever previously made them difficult or inaccessible.

Local directories have proven they can generate revenue. The problem has always been execution: building out the listings, keeping them current, writing enough SEO-friendly content to get search traction, managing outreach to local businesses. All of that is repetitive, labor-intensive work. Exactly the kind of work that AI handles well.

That's the hypothesis. Let's talk about what we actually did.

The experiments: two directories, two different angles.

We've been running two directory experiments. One is geographically focused — a specific type of local service provider in a defined metro area. The other is more niche-focused — a category of business across a broader geography.

I'm calling them LOZ and PCB internally, which are shorthand designations that will make more sense when I write more specifically about each. For now, the relevant data is about what's been working and what hasn't.

What the early data shows:

Search traction comes faster than I expected. With AI-assisted content production and good technical SEO structure, we started seeing organic impressions within the first few weeks of going live. Not converting traffic — impressions. The funnel from impression to ranking to click to lead takes time, but the foundation was visible faster than I would have predicted.

The outreach piece is harder than the content piece. Building listings from public data is something AI handles well. Getting businesses to claim and upgrade those listings is a sales process, and sales processes require human judgment and persistence. This is the bottleneck we did not underestimate exactly, but we may have underestimated how long the cycle takes.

The niche-focused directory is outperforming the geo-focused one on search traction, but the geo-focused one may have more monetization upside once the local relationships develop. These are early-stage observations — I hold them loosely.

What we're trying to prove.

The directory stream thesis has three components we need to validate before I'd call it a confirmed stream worth maintaining long-term:

One, that AI-assisted content production can get a new directory to meaningful search traction within a reasonable timeframe without a dedicated content team.

Two, that the listing upgrade conversion rate is high enough to produce revenue that justifies the ongoing maintenance effort.

Three, that the model is repeatable — meaning you could build a second or third directory using the same playbook and get similar results, which is the difference between a one-time success and an actual system.

We're somewhere in the middle of validating component one. Components two and three are ahead of us.

I want to be careful not to get too definitive about this too early, which is a temptation when something is showing early positive signals. Early positive signals are not proof of anything. They're permission to keep testing.

The thing I keep coming back to with directories is the compounding quality of SEO-based traffic. If you build it right, the asset keeps earning as long as it's maintained. That near-passive quality — work front-loaded, returns back-loaded — is exactly the profile we're looking for in this portfolio. Whether these specific directories deliver that remains to be seen.

We'll report back as the data develops.

This Week in AI: AI-assisted SEO content has become a significant enough category that Google has updated its quality guidance several times in the past year. The short version of where things stand: the question isn't whether AI wrote the content — it's whether the content is useful and accurate. Good AI-assisted content with genuine subject matter expertise behind it continues to perform. Thin content produced at scale continues to get filtered out. The bar for quality hasn't lowered; the tools for meeting it have improved.

Grab the free toolkit at start.tenstreamslab.com — includes our directory setup checklist and the SEO framework we're using.