I Built an AI That Scouts Competitor Listings While I Sleep: The Paranoia Tool
# I Built an AI That Scouts Competitor Listings While I Sleep: The Paranoia Tool
A competitor’s listing dropped $25,000 on a Thursday night at 11:47 PM.
I got the alert while I was half-watching a Marvel rerun. Most agents in this market found out Monday morning — if they were paying attention. Some found out when a seller asked them why their listing was priced the same as one that just dropped.
That’s the whole pitch. Right there.
—
## Manual Monitoring Is Just Anxiety With Extra Steps
Here’s the thing nobody says out loud: “I keep an eye on the market” is not a system. It’s a coping mechanism.
The way most agents actually track competitors looks like this: a few MLS bookmarks, some Zillow stalking, maybe a Google Alert that sounds useful but was never designed to watch listing databases or catch status changes in structured real estate data. It’s the kind of setup that feels productive until it isn’t.
And here’s the math that makes it worse. The more successful you get, the less time you have to maintain the habit. The more properties you need to watch, the faster it breaks down. Manual monitoring scales with anxiety, not with insight. You’re not getting better at it by trying harder. You’re just burning attention you needed for something else.
That’s the thing that stings: it’s not time. It’s attention. You can recover time. The 40 minutes you spend scanning MLS at 9 PM the night before a listing appointment — that’s not just 40 minutes. That’s your sharpest thinking, spent on a task that shouldn’t require any thinking at all.
The problem isn’t information scarcity. There’s plenty of data. Zillow, Redfin, MLS feeds, brokerage pages — it’s all there. **The problem is signal starvation.** Too much noise. Not enough signal. The agent who wins isn’t the one with the most data. It’s the one who spots the delta fastest.
—
## What I Actually Built (And Why I Called It the Paranoia Tool)
The name came from a conversation with myself at about 1 AM on a night when I couldn’t sleep and kept wondering what was happening in my farm areas.
Not paranoia about losing business. Paranoia about not knowing. There’s a difference.
So I outsourced the paranoia.
Here’s what the system actually watches:
– A defined watchlist of specific competitor agents in my market — not everyone, not every zip code. Tight. Five or six agents whose moves would change how I price, position, or pitch.
– New listings from those agents that hit my core price segments
– **Status changes** — active to pending is not just interesting, it’s demand data. In North Idaho where you might have eight properties in a sub-area instead of eighty, one pending status tells a story
– Price reductions, with thresholds calibrated by segment. A $25K drop means something different on a $450K property than on a $1.6M one
– Repeated relists — when a listing fails, gets pulled, and comes back with adjustments, that’s a pattern. That’s an agent recalibrating their playbook in real time
– Language changes in public remarks — concession terms buried in the text, showing instruction shifts, anything that hints the seller’s strategy moved without a formal price change
– Photo order edits. Sounds minor. Isn’t. If the hero shot changes, someone made a call about what’s not working
The watch → score → escalate logic matters here. Not every alert is the same urgency. The system has to know the difference, or it becomes the firehose problem all over again.
What it ignores: everything else. The filtering is where the system earns its keep. If I’m getting more than a handful of meaningful alerts in a day, the filters need tightening — not the watchlist expanding.
The output is short. A quick summary I can act on in under 60 seconds. If reading an alert requires effort, the system gets ignored within a week. I’ve seen that happen. It’s not the tool failing — it’s the design failing.
Building the watchlist is where discipline lives. Too many agents, too many areas — the whole thing collapses into noise by day three. Tight input equals useful output. Loose input equals another inbox you avoid.
—
## What It’s Caught That I Would Have Missed
Two examples. These are real.
The Thursday night price drop I mentioned at the top. The next morning I had a seller call me about pricing their property. I already knew the competitive landscape had shifted overnight. They didn’t. That conversation went differently than it would have if I’d found out Tuesday.
Second one’s quieter but honestly more interesting. A competing agent had a listing go active → pending → back to active over about twelve days. No announcement. No obvious price change on first look. The system caught the relist. When I dug into the remarks, the concession language had changed — seller was now offering to cover closing costs, which wasn’t in the original listing. That’s not a headline change. That’s a small edit that signals a strategy shift. A human doing a spot check wouldn’t catch that unless they were specifically looking for it and comparing versions. I wasn’t looking for it. The system was.
In a thin-inventory market, these aren’t trivia. One listing resetting expectations for a pocket area changes what sellers in that pocket think their home is worth. A same-day alert is categorically different from a weekly recap when inventory is that tight. The gap between Thursday night and Monday morning is the whole advantage.
—
## The AI Isn’t the Oracle. It’s the Filter.
Let me be straight about what the tool actually does, because there’s a version of this conversation that goes sideways into hype fast.
The AI is not reading the market for me. It’s not making judgment calls. It cannot tell me that a particular street in Bonner County sells differently than the one behind it. It doesn’t know that one seller’s agent has a pattern of overpricing the initial list and then cutting twice. That’s local knowledge. That lives in my head, not in any model.
What the system does is compress the distance between *something changed* and *I know about it*. That’s it. The judgment call is still mine.
The cognitive load reduction is real, though. There’s a version of my mornings where I’m scanning, checking, refreshing, making sure I haven’t missed anything. That version is exhausting and it’s not even effective because human checking frequency doesn’t match market-change frequency.
The current version is: alerts come to me. I read them in 60 seconds. I make a call. Then I use the brain I would have spent on scanning to prepare for the listing appointment instead.
That’s not a small thing. Attention compounds. Where you spend it decides what you’re capable of.
—
## How to Build One Without Losing Your Mind in the Process
No tool list here. Tools age. Decision logic doesn’t.
Start with the watchlist, not the tooling. Who specifically matters in your market? Two to five competitor agents whose moves would change your strategy. Not twenty. Not “everyone active in my primary zip.” You need names.
Define your thresholds before you build anything. What percentage price drop actually matters in each segment you work? What’s the DOM number that signals a real problem versus normal absorption? What keywords in remarks indicate a strategy shift? If you can’t answer these questions before you start building, your alerts will be useless.
Build for the delta, not the snapshot. The tool should tell you what changed, not just what exists. That distinction drives every decision downstream.
Calibrate by segment. This is where most people get it wrong. A flat alert for any $10K price change will cry wolf constantly in some segments and miss meaningful moves in others. The system has to know the difference.
Keep the output short enough to act on in 60 seconds. Long alert, ignored alert. Every time.
A few things worth monitoring that human spot-checking consistently misses:
– Remarks edits that add or change concession language
– Showing instruction changes (seller getting more flexible = motivated seller)
– Photo updates after DOM starts climbing
– Relists under new MLS numbers — a favorite move when a listing fails quietly
– Status flips on weekends and after 9 PM — exactly when human checking drops to zero
One honest caveat before you go build this: automation doesn’t remove discipline. It demands more of it. If you’re not willing to maintain the watchlist and update the thresholds as the market shifts, the tool decays into noise fast. The system is only as good as the judgment you put into building it. That’s not a warning against building it — it’s a warning against building it sloppily.
—
## The Market Moved Last Night. I Know. Do You?
The agents who feel like they’re always one step behind aren’t less talented. They’re not less experienced. They’re just relying on human checking frequency in a market that moves in human absence.
The Paranoia Tool doesn’t make me smarter. It keeps me from giving away a time advantage I earned — for free — every night I’d otherwise be asleep and not checking.
If you want to talk through what I’m actually watching in the North Idaho market, or you’re curious about the structure I built and whether something similar makes sense in your market, reach out. I’m genuinely happy to talk through it.
I already know what happened while you were sleeping.