There are three basic types of property description generators
When we first launched Nila June’s purpose-built, agent-centered AI system for listing descriptions, there were only a handful of property description generators on the market. Now there are too many to mention and rank by name, but they are logically sortable into three categories. Here they are, from least to most sophisticated:
- Low end: Most property description generators are simple overlays to OpenAI’s GPT or other large language model (LLM). With a little bit of information, they produce unremarkable, cookie cutter results that require significant editing to ensure uniqueness, accuracy, and compliance with fair housing regulations.
- Mid-tier: A tech-heavy hybrid system that attempts to read photographs uploaded by the agent, and then feeds the information gleaned from those photographs into a large language model such as OpenAI’s GPT. While the technology might be interesting to some agents, the output is hampered by the same problems presented by low-end overlays to GPT.
- High end (Nila June): The best descriptions come from a purpose-built, easy-to-use AI system, created by professional writers in consultation with real estate agents, without relying upon ChatGPT or other Large Language Models (LLMs). As far as we can tell, this is a category of one: Nila June.
Let’s look at each of these three categories.
Low-end, a simple overlay to OpenAI’s GPT
Low-end services offer an interface that asks users to select only the most basic facts about the home from a brief checklist. The simplest ones don’t even offer the checkboxes, but instead ask users to freetype details about the property into an open text box. When regarding these description generators, listing agents should ask themselves: With this paucity of information, could even a professional human writer come up with a great description?
These low-end services send bits of information to a large language model, usually OpenAI’s GPT. The results can be disappointing. The output is an assembly of the most common property description phrases. Every view is “stunning.” The home “boasts” a tire swing. (Free advice: Don’t use “boasts.”) The home is “perfect for” this or that purpose or demographic. “Perfect for entertaining loved ones” shows up a lot. And “Perfect for small families”–ouch. (More free advice: Don’t use “perfect for”). Sometimes the descriptions are simply factually incorrect, with the addition or omission of whole rooms, floors, and exterior features!
In addition to these problems, other tell-tale signs that a service is in the low-end category are: a) no writers involved in the creation of the writing product, b) no agents involved in the creation of the listing product, and c) the service is subscription based, charging a monthly fee in conformance with OpenAI’s monthly pricing strategy.,
Mid-tier, photo upload and then a simple overlay to OpenAI’s GPT
Some companies replace the checkbox or open-text input forms with a photo upload requirement. This makes for a sharp product video, but in practice it imposes a mechanical and QA burden on the customer. Those who are interested in the technology and willing to play a QA role in photo recognition may embrace the process–but in our experience, few busy real estate agents are in this category.
The information gleaned from the photos and QAed by the agent is sent to an LLM. That’s logical, but now we’re back to all of the textual disadvantages described above. (And of course, there is also the monthly subscription fee!)
High end, Nila June’s purpose-built AI system
There’s a reason why the tech-forward MIAMI Association of REALTORS® and others have selected Nila June as their preferred vendor in this category.
With no dependence on ChatGPT, no photo upload requirement, and no monthly subscription fee, Nila June stands out for ease of use, and for delivering the best property descriptions.
Nila June’s data input process is based on a curated online survey developed in consultation with our client base. The survey helps organize the listing agent’s perspective on the property, and helps ensure that no details are left out of the posted description.
Nila June does not use a commercially available large language model, such as GPT. Instead, professional human writers created Nila June’s proprietary language system, with an eye toward freshness, creativity, and engagement.
When we pair the listing agent’s survey responses with Nila June’s purpose-built language system, the result is a great property description.
Unlike the others, we don’t charge a monthly fee. Why should you pay Nila June during months when you don’t have a listing? On the other hand, when you have a commission in sight, it makes sense to spend $19.99 for a great description—even if you have several listings in one month. Many of our customers have this “problem.” They are happy to buy as needed—maybe multiple times in a single month, maybe none for a month or two.