AI Can Speed Up Work But Still Needs Your Review

When attorney Timothy Shields asked AI to help redecorate a room in his house, he supplied it with a photo, color swatches and furniture ideas. Within moments, he had a beautiful image.

Then he noticed the window. The imaginary window that AI created. The real room didn't have one.

For Shields, a data privacy and cybersecurity attorney and Florida Realtors®' outside counsel, this home decorating experiment was indicative of what can go wrong with AI and a warning for Realtors. If this was a listing photo, that same invented window could give buyers the wrong impression of a property before they ever walk through the door. It's a legal liability that should give every agent pause.

Shields recounted the experience during Florida Realtors’ Oct. 20 webinar, “Is That Dream Home a Hallucination? AI Risks in Real Estate,” explaining how everyday uses of AI can create problems when the output goes unchecked.

“Treat every AI output as a first draft,” he said.

Shields uses AI in his own practice to move documents along faster, freeing more time for substantive review. But a confident answer isn’t necessarily a correct one. He cautioned against using AI to venture into subjects users don’t know well enough to check independently.

For real estate professionals, that means reviewing listing descriptions and emails for made-up facts or changes. Contract summaries carry particular risk because the tool may overlook important information or change legal terms users don’t fully understand, he said.

“You can never blame the AI tool for a bad output,” Shields said.

Keeping that property picture accurate

Photo editing deserves the same scrutiny, Shields said.

In listing photos, removing water stains from a ceiling or concealing wall damage could misrepresent a property’s condition, he warned. Even an attempt to tidy up a messy garage can produce a misleading image.

Shields recommended retaining both original and edited photos. Labeling an image as AI-edited is the most conservative approach, he said.

He also emphasized that existing rules governing licensure, advertising, fair housing, deceptive practices and the Code of Ethics still apply to AI-generated material.

“You are the one who is responsible. You can never blame the AI tool for a bad output,” he said.

Set boundaries before using the tools

Choosing an AI product is only part of the decision. Brokerages also need to define what people may do with it and what information belongs in it, Shields said.

An AI policy can be just one page. It should identify whom it covers, approved and prohibited tools, how new tools are reviewed and permitted uses. It should also address data protection, incident reporting and human review before publishing.

Shields offered these additional safeguards:

• Verify AI output before sharing, especially facts and fair housing language.
• Check edited photos for misleading changes. Keep originals and consider AI-edited labels.
• Check providers’ data practices and keep customer information out of unapproved tools.
• Review email tools’ storage practices and meeting transcripts’ accuracy and privacy risks.
• Limit what customer-facing chatbots can say and share.
• Use unique passwords, watch for suspicious links and scrutinize wire instructions.
• Review insurance coverage with your broker and practice responding to an incident.

“Never put client identifying information into a tool that’s not approved,” Shields said. “Never publish facts that you do not know to be 100% true, and then never alter an image in a way that would be deceptive.”

© 2026 Florida Realtors®