AI / generative AI
Software that can create or change text, images, or video from instructions (or from your photos). “Generative” means it can make new-looking content — not just crop or brighten.
Prompt
The written instructions you give an AI tool (“brighten this kitchen, keep the same layout”). Clearer prompts usually mean fewer surprises.
Photoreal / realistic AI
AI content that looks like a real photo, video, person, or place — not a cartoon. This is when disclosure and platform labels matter most.
Virtual staging
Adding digital furniture and decor to empty (or sparsely furnished) rooms in a photo. The walls, windows, and layout should stay true.
AI-enhanced
A real photo that AI cleaned up or improved (lighting, sky, declutter) — or motion made from stills — without inventing a different home.
Watermark / on-image label
Small text on the photo itself (e.g. “Virtually Staged”). Optional extra — the listing and the app AI switch do the job. For video, use the app switch and a description, not a burn-in.
Platform AI label (native toggle)
The built-in switch in YouTube, TikTok, Instagram/Facebook (and ads tools) that marks content as AI-made or AI-altered — not the same as a hashtag.
Disclosure
Clearly telling buyers that media was staged or AI-changed, in plain words — in the listing and the description, plus the app’s AI switch when you post. A stamp on the photo is extra.
MLS remarks
The written description on the listing in your MLS. Important, but many buyers see photos on portals or social first — not this text.
Photo-to-video
Turning listing stills into a short moving clip. Keep the rooms locked to the real photos. Use the app’s AI label when it looks real.
Silent geometry changes
When a staging or cleanup tool quietly moves walls, windows, outlets, or views while “only adding furniture.” The home no longer matches reality.
Portals & syndication
When listing photos show up on public sites (like Realtor.ca) or get reshared. Those surfaces often show the image without your full MLS text.
Material defect / condition
A real issue that could matter to a buyer (damage, water stains, missing features). AI should not erase these to “pretty up” the listing.
Deepfake
AI that fakes a real person’s face, voice, or words. High risk for trust, privacy, and advertising rules — avoid without clear authority and disclosure.
In-app AI (CapCut, Canva, phone tools)
Editors that apply AI effects before you post. If the result looks real and AI changed it, label on the app where you publish (TikTok, IG, YouTube).
Human review
You (or your team) checking captions, facts, and every frame before publish. AI drafts; you are the publisher.
LLM (large language model)
The kind of AI behind ChatGPT-style tools: trained on lots of text so it can draft captions, emails, and listing copy. It does not automatically know your listing facts.
Model
The underlying AI “engine” (e.g. a specific image or text model). Different models behave differently; product names (ChatGPT, Midjourney) sit on top of models.
Training data / training
The material an AI learned from. Some tools may use your uploads to improve the model unless you turn that off in business settings.
Hallucination
When an AI confidently invents details that aren’t true (wrong bed count, fake schools, made-up square footage). Always fact-check listing copy.
Upscale / upscaling
Making a photo look higher-resolution or sharper with AI. Usually a light edit — still don’t let it invent new details that mislead.
Inpainting
AI fills or replaces a selected area of a photo (remove a cord, clean a spot). High risk if used to hide defects or change the property.
Outpainting
AI extends the edges of a photo beyond what was captured. Easy to invent yard, sky, or neighbouring context that isn’t real — use carefully or not at all.
Seed
A number some image tools use so you can try to recreate a similar result. Handy for consistency; not required for honest marketing.
Negative prompt
Instructions about what you don’t want (e.g. “no extra windows, no warped walls”). Helps keep listing photos honest.
Tokens / context window
Roughly how much text a chat AI can “see” at once. Long listing packages may get truncated — keep prompts focused on one job at a time.
Temperature
A setting on some text tools: higher = more creative/varied; lower = more predictable. For listing facts, keep creativity low and verify everything.
System prompt / custom instructions
Standing rules you give a chat AI (e.g. “never invent amenities”). Our safety lock prompt is this idea for photo tools.
Image-to-image
You upload a real photo and ask AI to change it (stage, relight, tidy). Safer for listings than pure text-to-image when you keep structure locked.
Text-to-image
Creating a picture only from words, with no real listing photo. High risk for inventing a home that doesn’t exist — avoid for specific listings.
LoRA / style pack
An add-on that steers an image model toward a look or style. Fine for brand polish; still must not invent false property features.
ControlNet / structure lock
Tools that try to hold layout, edges, or depth while AI restyles a photo. Useful when you want edits without moving walls — still verify the output.
Synthetic media
Umbrella term platforms use for AI-made or heavily AI-altered audio/video/images. Same idea as “Made with AI” labeling.
Chatbot / AI assistant
A text chat interface (ChatGPT, Gemini, Claude, Copilot, etc.) for drafts and Q&A. Great for first drafts — not a source of verified listing facts.
API
A way software talks to an AI service automatically (e.g. a CRM drafting tool). You still own the final content that goes public.
RAG (retrieval)
When a tool pulls from your documents before answering. Can reduce made-up answers if the docs are accurate — still check outputs for listings.
AI voice / text-to-speech
Computer-generated speech for Reels or tours. Use only authorized voices. Don’t clone another agent or seller.
AI avatar / persona
A digital character that presents content. Must not fake personal experiences of living in the home or claim results it can’t have.
Image generators (Midjourney, Firefly, etc.)
Tools that create or edit images from prompts. For listings, prefer editing real authorized photos over inventing rooms from scratch.
Open vs closed model
Open models can be run in more places; closed ones stay inside a vendor’s app. Either way, honesty and disclosure rules for marketing stay the same.
Bias
AI can reflect unfair patterns from its training (e.g. stereotypes in images or copy). Review outputs for tone and fairness before publishing.
Metadata / content credentials
Hidden or attached info about how a file was made (sometimes including AI provenance). Helpful for tracking; not a substitute for the listing, the description, or the app AI label.
AI agent (agentic tool)
Software that tries to complete multi-step tasks for you (research, draft, schedule). You still approve anything that becomes public advertising.
Fine-tuning
Further training a model on your examples (brand voice, past listings). Can help tone — never use private client files in tools that aren’t approved.
Diffusion model
A common type of image AI (builds pictures step by step from noise). You’ll see this name in technical tools; behaviour still needs human checks.
Plugin / integration
An add-on connecting AI to another app (email, drive, CRM). Check what data it can access before pasting listing or client information.