Listly

The problem with using AI for property copy is that a language model will happily fill gaps with plausible detail. For a listing, that is a real risk. Inventing a nearby school or overstating a walk to the station is the kind of error that misleads a buyer and lands an agent in trouble.

Listly is built so that can't happen. The agent works through a wizard: sale or let, the address, the property facts like beds, price and parking, and then the area. For the area, they confirm what is actually nearby, and anything unconfirmed stays out. The AI writes a short summary, key features, a full description and a location paragraph, but only from what the agent has confirmed. It won't add a room, a school or a station that wasn't entered.

We also set the tone deliberately. It writes like a high-street agent, specific and warm, and avoids the "stunning", "nestled", "must-see" filler that makes listings blur together.

The AI Model is reached through Vercel's AI Gateway rather than wired in as a separate product. That keeps the app simple and lets us route the AI calls through the same platform that hosts it, and it means we can change the underlying model without rebuilding the app around it. The AI does two jobs: it can read optional photos to suggest features the agent might tick, such as a garden or a kitchen, and it writes the copy. The photos only ever suggest features for the agent to confirm or ignore. They never go into the listing themselves. Address lookup, the area map and nearby places all come from Google Maps and Places.

One deliberate choice: nothing is saved. There's no login and no database. Listly is a session tool, so when the agent closes the tab, it's gone. And if Maps or the photo analysis fails, the agent can still carry on and generate copy from whatever they've already confirmed.