Redfin & Zillow launch plugins for ChatGPT

AI - Artificial Intelligence
#1

OpenAI is partnering with companies to feature plugins in its paid version of ChatGPT. These plugins integrate with ChatGPT to provide added functionality, including access to the internet.

Entering this arena are Zillow and Redfin. Both seem to be offering the option to search listings by conversing with ChatGPT. And, of course, the information that gets presented (in ChatGPT) comes from either Zillow or Redfin, depending on which plugin you use.

The value-add seems to be that you can perform searches in the conversational way that ChatGPT is suited to. However, this also means you have to deal with ChatGPT’s little quirks as well.

For example, I found this demo video of someone using the Redfin ChatGPT plugin. They performed a search for homes with “2 bedrooms and 1 bathroom”. The issue (as seen in the demo), is the very logical and literal way that ChatGPT interprets prompts.

You see, for any number of bedrooms OVER 2, it is ALSO true that there are 2 bedrooms. So if you ask “show me homes with 2 bedrooms”, ChatGPT will show homes with anywhere from 2 to infinity bedrooms.

But that’s not how humans understand the question, and not how we want to perform real estate searches. So this kind of search may have some issues to work out before it becomes a viable alternative to the search functionality you get on any real estate website or portal.

But, it is interesting that some big players are getting into this space. What do you think? Will people got to ChatGPT to look for homes? Where could this idea go?

References:
Zillow’s announcement
Redfin’s announcement
Microsoft’s announcement
Article on The Verge

2 Likes
#2

Interesting, thank you for these articles.
I wonder if the AI will help us to get around grey areas in the MLS data, specifically things like this:
Let’s say a buyer wants a 2- or 3-storey home with at least 3 bathrooms: one must be an ensuite off the primary bedroom, and every storey must have at least one bathroom. Most feeds don’t have this kind of data, but maybe the bot could deduce it from looking at floorplans?

Or, a buyer might say they want an ocean view. In my board, an agent might check the “ocean view” box, or they might not. And an agent with a property with a TINY ocean view, might check the box. I wonder if the bot could review photos, or reference some external geographical set, to determine the likelihood of a proper ocean view.