September 25, 2026

AI can draft a Central Coast comp sheet. Your street still sets the price.

Josh McEwen, Invision Real Estate, DRE 02178333

This is not a market stamp. No Freddie print. No Altos medians. Just how I use AI when buyers and sellers ask what a house is "worth" on the Central Coast — and where I throw the chatbot answer out.

What a good comp sheet actually needs

Recent solds that match the subject: beds, baths, square footage band, lot feel, condition, and the same town. On this coast the town matters more than the county headline. Arroyo Grande is not Atascadero. Paso is not Santa Maria. Shell Beach rides with Pismo / 93449. A tool that averages those into one "Central Coast comp" is already wrong.

List price and sold price are different columns. A sitting ask is not a closed sale. AI will mash them if you let it. I do not.

Where AI helps

Fast first pile. I can ask a tool to pull a short list of recent public solds in a zip or radius and sort by date. That is homework prep, not a CMA.

Condition flags. Feed it the subject's photo set and ask which rooms look dated next to the sold comps you already trust. Useful for a seller deciding what to fix before we list — not for inventing square footage.

Question list. Before we walk a house, AI can turn your notes into a checklist: roof age, HVAC, septic vs sewer, HOA docs, flood or fire overlay. You still verify on the ground and in the disclosures.

Where AI breaks on this coast

Thin samples. Some towns only have a handful of closed sales in a normal month. A chatbot will pad the pile with wrong towns, wrong years, or list-price "comps" that never closed. Small sample means walk the street, not trust the average.

Condition lies. Two 1,600 sq ft ranch homes can be $100k apart because one has a 1990s kitchen and the other was remodeled last year. AI reading Zillow cards does not smell smoke damage or hear Highway 101 at 5pm.

Fake precision. "Fair value is $1,247,500" from a model is not an appraisal and not a CMA. On the Central Coast I price to what just sold and what just listed on your block — after I have checked the solds myself.

How buyers can use it this week

1. Ask an AI tool for a short list of recent solds in the two or three zips you care about. Bring that list to me. We throw out anything that is the wrong town, the wrong property type, or still only a list ask.
2. Have it compare your must-haves against each address (garage, yard, single-story, school preference). We still map the keepers onto live CRMLS.
3. Do not accept a generated "offer price" as gospel. We set strategy from real closed sales plus current competition on the street.

How sellers can use it this week

1. Comp homework: let AI gather recent solds as a starting pile. I verify each one against the street before we pick a list price.
2. Gap list: ask what buyers will notice first in your photos versus the sold comps you like. Fix or stage those items. Do not publish AI-invented rooms.
3. Pricing: we land on a number from verified solds and active competition — not from a chatbot's "estimated value" badge.

The short version

AI is a fast filing clerk for solds and checklists. I am still the agent who throws out the wrong towns, keeps list and sold in separate columns, and prices to your street. If you want that mix on Arroyo Grande, SLO County, and North Santa Barbara County — DM me, call 805-668-7238, or email [email protected].

Josh McEwen, Invision Real Estate, DRE 02178333, (805) 668-7238, [email protected]. Broker: Aaron David Peck, DRE 01387499.

Sources: This note is process, not a market stamp. No Freddie print and no Altos medians. The dollar lines in the body are examples of where a chatbot fails, not a CMA and not a listing.