Fair Housing & Compliance When Using AI

Protected characteristics in general terms, steering as a behavior, discriminatory advertising as a message — refreshed for a world where tools draft and place.

In this lesson:

• Name the protected characteristics in general terms, including local additions
• Recognize steering as a behavior and discriminatory advertising as a message
• Explain why AI raises the stakes on both

Generated language, targeting, chatbots, and screening — the places where a person used to decide and a tool now decides by default.

In this lesson:

• Locate the four exposure points where AI touches fair housing
• Rewrite generated text to describe the property, not the person
• Recognize chatbot and screening outputs as the brokerage's own statements and decisions

How a neutral tool can discriminate: the concept, the proxies, and the four-question working test.

In this lesson:

• Explain disparate impact as a concept without naming a case or statute
• Identify neutral inputs that stand in for protected characteristics
• Run the four-question working test on a tool or rule

Words, images, and audience selection — the three places generated campaigns signal preference, and the safe pattern for each.

In this lesson:

• Spot people words, coded words, and proximity claims in generated copy
• Direct image generation and selection so no group is the implied resident
• Place ads with audience settings that don't exclude by characteristic or proxy

Principle-level: what enforcement has focused on, the principles regulators apply to automated tools, and what an inquiry asks for first.

In this lesson:

• Describe where fair-housing enforcement attention has gone with AI, at principle level
• Translate regulator principles into brokerage practice
• Know what an inquiry asks for first

The pre-run checklist, how to run it in four steps, and the prompt that makes the reviewer's job easier — never unnecessary.

In this lesson:

• Apply the six-item pre-run checklist to generated marketing
• Run the review in four steps with a named reviewer and a record
• Write a generation prompt that starts the draft clean

What a property chatbot may say, the paired consistency test, and routing the questions it must not answer.

In this lesson:

• Define the in-scope and out-of-scope questions for a property chatbot
• Run a paired consistency test before launch and after updates
• Configure routing for safety, schools, 'who lives here', accommodations, and requests for a person

Where brokerages touch screening, the rules that apply to written criteria and tool scores, and a placement case done right.

In this lesson:

• Identify the five places a brokerage touches screening decisions
• Apply written criteria equally and treat a tool score as an input, not a reason
• Process applications in a neutral, explainable order

Recognizing a request in any words on any channel, the four-step response, and where automated tools get accommodations wrong.

In this lesson:

• Recognize an accommodation or modification request without magic words
• Respond in four steps with a person, in writing, with a reason
• Fix the common ways automated tools mishandle requests and accessibility

Who needs fair-housing AI training, what it covers, and what makes it change behavior rather than tick a box.

In this lesson:

• Identify everyone in the office who needs fair-housing AI training
• Structure training around the office's own tools and checklist
• Distinguish training that changes behavior from training that ticks a box

Versions, reviews, and decisions — what goes in the file, how it's kept, and what one campaign's record looks like.

In this lesson:

• List what a fair-housing AI compliance file contains
• Keep the file so it's captured at the time, versioned, and producible in a day
• Assemble one campaign's complete record

Five simple fairness checks a brokerage can run itself, the paired test step by step, and how to read the results.

In this lesson:

• Run five simple fairness checks on the office's AI outputs
• Conduct a paired test across chat, email, and web forms
• Turn findings into actions and records

The first-day steps when a fair-housing complaint or inquiry arrives, what not to do, and the record that protects the brokerage.

In this lesson:

• Execute the first-day steps: preserve, notify, hold, pull the file
• Avoid the responses that turn a complaint into a second problem
• Assemble the record that answers the questions an inquiry asks

Three anonymized composite scenarios — a lookalike audience, a chatbot under test, a screening score — and the practice that protected the brokerage in each.

In this lesson:

• Recognize the three most common AI fair-housing failure patterns in realistic settings
• Identify the specific practice that protected the brokerage in each case
• Apply the same practices to your own tools

The one-page checklist, the before-during-after rhythm, and your next campaign; the finale.

In this lesson:

• Carry the one-page fair-housing AI checklist into daily work
• Apply the before-during-after rhythm to any tool
• Run your next campaign as the first entry in the compliance file