TRD Tests AI Listing Generators

Brokerages, agents and other industry players are jumping on the ChatGPT buzz to launch their own artificial intelligence tools.

Since its debut last November, OpenAI’s chatbot has powered a wave of real estate tech tools, including AI assistants, content generation services and enhanced search functions. Just last month, the Eklund-Gomes team launched its new website with a built-in AI chatbot called Maya.

Now agents can lean on these tools to answer questions about the market, craft listing descriptions and create social media posts. But with all the new products on the market, which products make the most sense for their business needs?

ChatGPT

Back in January, we asked ChatGPT to come up with a listing description for a fake luxury mansion in Malibu and the site of an infamous murder.

To generate the text, we fed ChatGPT some specs about the house, including the usual metrics like price, square footage and bedroom-bathroom count and some facts about the home’s history. We also ran the prompt through ChatGPT multiple times to refine its response.

Our takeaway at the time was that the bot crafted a thorough response, though the language was stale, repetitive and robotic. The description repeated the phrase “perfect for” at least four times.

While ChatGPT can certainly be a starting point, it requires some heavy lifting from users.

Saleswise

The process is in line with that of another tool, called Saleswise, that markets itself as an AI-powered listing generator. The interface for the product is similarly basic, though it prompts users specifically to input home features.

We ran the same prompt from the ChatGPT trial through the generator, and the output was not as stiff or mechanical as our last attempt, though it was heavy on the adjectives.

Instead of “the spacious 5 bedrooms, 8 bathrooms, and state-of-the-art home theater, provide you with the ultimate comfort and relaxation,” Saleswise wrote, “indulge in the luxuries of this home, starting with the 5 opulent bedrooms and 8 lavish bathrooms.”

But other tools on the market don’t require as much comprehensive input from agents. Instead, the platforms prompt agents to fill in categories with drop-down menus, multiple-choice sections and key phrase generators.

Marketproof

Marketproof launched its AI assistant in May, drawing on the data analytics firm’s trove of proprietary data to help agents come up with listing descriptions.

The tool allows users to select a specific unit in a building and generate a listing description based on Marketproof’s existing information about the property. Users can designate the word length using a sliding scale and select whether they want the listing to have a professional, neutral or casual tone.

The next section includes an already-generated list of features to highlight in the description based on Marketproof’s data on the property. These can include phrases like “terrace,” “double-height ceilings” and “walls of windows.”

On Marketproof’s site, we generated a 400-word, casual description for Unit 107 at Extell’s Central Park Tower.

While the output did achieve a casual tenor, agents will have to give descriptions generated with this tool a healthy few read-throughs. Several sentences were missing key words or phrases , including starting a paragraph with “not only a family room but also a media room, conservatory, and library/office, offering ample space for relaxation and entertainment.”

But regenerate the description a few times, and the tool will reword sentences to correct those errors.

Listingcopy.AI

Dennis Saglam, a Douglas Elliman agent in Sag Harbor, released his own platform called Listingcopy.AI, which allows users to create listing descriptions, social media posts and emails.

Saglam worked with a developer to launch the site, which he said has about 2,000 users and pulls data from Zillow and other third-party websites to help users compose their descriptions.

To get started, agents are prompted to plug in the address of the property, select the number of bedrooms and bathrooms from a drop-down menu and fill in some other key features in a text box.

The tool will then generate a couple of example listings, but users can customize the output by choosing a framework for the text, length, audience and tone.

We used Listingcopy.ai to create another listing for the Central Park Tower unit, selecting the framework as “SEEDS: Storytelling & Evocative Emotion-Driven Sales Text,” the length as “Medium — Maximum 400 words,” audience as “The Luxury Buyer,” and the tone as “Inspirational.”

While the drop-down menus and prompts make this tool easy to use, unlike Marketproof, agents will need to manually add property features to ensure they’re mentioned in the description. And the output, while true to the tone and length selected, needs a human touch.

The tool also made promises that it can’t keep. One of the sentences read, “For those seeking a secondary home or investment property, this duplex promises a significant return, both in terms of financial gain and unparalleled luxury living.”

After testing out a few of these tools, it’s clear most of the listing description generators are more alike than different, though some are more explicit in walking new users through the process.

But ultimately, each requires a careful editor and a detail-oriented prompter to ensure the output reads smoothly.