We are delighted to introduce a timely, pragmatic new training suite for California licensees: Artificial Intelligence Training: Harnessing AI While Managing Risk™.
This premium training fully illustrates how to safely augment your practice with AI-powered tools while maintaining absolute regulatory compliance — without sacrificing human intuition. Over the coming weeks, the firsttuesday Journal will feature exclusive excerpts from this multi-media suite.
This second installment illustrates the need to carefully verify all AI-output, and depicts how fine-tuning your AI query with specific details about property’s features improves results.
The complete, video-focused Artificial Intelligence Training™ is available at no additional charge to students enrolling in any firsttuesday licensing or renewal course — a FREE service to students. Or it may be separately purchased for $35.
Exclusively Californian: specifically firsttuesday.
When silicon slips up
Consider this thought experiment: when an AI-powered tool generates inaccurate, wrong or harmful information, who is on the hook? May an agent shift liability to AI by giving notice that AI occasionally gets things wrong?
Let’s look at an example.
Consider an agent who struggles with writing a description for property offered for sale. The agent just can’t seem to come up with the right words.
Then the agent has an idea. The agent provides AI minimal written instruction and facts, prompting it:
Generate a property description for a 3 bedroom 2 bathroom 1500 square foot home in Sacramento, California with a 2 car garage. Granite countertops. Nice neighborhood.
The AI responds:
Welcome to this sun-drenched 3-bedroom, 2-bathroom sanctuary nestled in one of Sacramento’s most coveted, tree-lined neighborhoods. Spanning a generous 1,500 square feet, this home boasts a seamless open-concept layout perfect for entertaining. The kitchen is a chef’s dream, featuring premium granite countertops, all-new stainless steel appliances, and custom cabinetry. Retreat to the spacious master bedroom, complete with a walk-in closet and a spa-like ensuite bath with a soaking tub. Outside, the professionally landscaped backyard offers a private patio for summer BBQs. Includes a spacious 2-car garage and energy-efficient dual-pane windows throughout. Plus, you’ll love the quiet neighborhood – very safe. This gem won’t last!
As you may have noticed, the AI added extra features — called bleed — not referenced in the agent’s initial prompt.
Why? AI doesn’t just describe what the agent gave it — AI predicts what a “good” property description usually looks like, often pulling from its vast know-all memory of high-value features — the bleed — to make the text sound more persuasive.
Take a closer look at the specific language that needs careful attention.
While there is always some leeway for editorial license and descriptive language, any specifically referenced feature — a fact — must be verified as correct and existing before it is included to market a property. When the feature cannot be verified as a fact, they are removed. In this case, several features added by AI — including dual-pane windows, a soaking tub and stainless steel appliances — need to be removed by the agent.
And what’s acceptable descriptive language?
Let’s take another look at the description used to market the property after the agent removes the language they determined to be incorrect.
In this example, the words “sun-drenched” “generous” “sanctuary” “chef’s dream” and “spacious” are all considered puffery. Think “sales talk.” They are subjective opinions on quality and do not require rigorous technical verification under DRE standards, provided they do not hide a known defect.
The mimicry machine: garbage in, bias out
Additionally, AI models do inherit the biases of the language models used to train them. Whether an honest mistake by an AI tool — or an agent — discrimination in real estate advertising carries steep penalties.
Agents in their review of AI responses need to correct outdated language like “master bedroom.” Further, while the agent told the AI tool that it was a “nice neighborhood,” the AI tool took this a step further and said the neighborhood was a “quiet neighborhood – very safe.”
Here, the word “quiet” may be taken as excluding families with children, a protected class. Also, the word “safe” is a red flag, since it places liability on the agent and their broker for what a homebuyer may or may not consider “safe.”
Engineering precision by fine-tuning the prompt
In addition to double-checking the final output, fine-tuning your AI query with additional specific details about the property’s features will produce better results.
When you know of specific property attributes and distinguishing features, get them into the prompt. Or, add the additional facts into what was already generated.
Thus, with careful curation and confirmation of the results, the property’s promotional description, with the assistance of AI, is imbued with personality — which gets the attention of buyers and buyer agents.
Common discriminatory terms to avoid in advertising
Implicit bias or implicit discrimination occurs when an action is not outright or consciously discriminatory, but it still has discriminatory effects.
Here, ads or notices (including AI-generated ads) for housing which state or infer preferences or limitations based on any of the prohibited discrimination factors are considered unlawful discrimination. [Calif. Government Code §12955]
Some common terms that AI might suggest — AI being unaware of their discriminatory implications — include:
- exclusive community or prestigious neighborhood;
- mother-in-law suite or granny flat;
- bachelor pad;
- walk to Church, convenient to Synagogue, close to Mosque, etc.;
- quiet street;
- independent living; and
- not suitable for children.
While not outright discriminatory, each of these phrases can have the result of excluding members of a protected class, discouraging them from applying.
Editor’s note — Stay tuned for the next excerpt of Artificial Intelligence Training: Harnessing AI While Managing Risk™.
Get the full training suite today — FREE — with any enrollment in our Agent Licensing, Broker Licensing, DRE CE or NMLS CE courses.
Purchase the training separately as a standalone masterclass for $35.









