A listing description is four lines, not a paragraph. Most ChatGPT drafts fail because they treat it as a paragraph: too many adjectives, no real fact on the line, and a closer that promises the kitchen has "character" without telling you what that means. The frame that ships is hook, fact, softener, closer — one sentence per line, four total, forty-five words.
The hook is the single line a buyer skims on a phone. The fact is the one number a saver sees (square footage, lot size, year built). The softener is the human element: the south-facing kitchen, the three-block walk to school, the quiet street after 9pm. The closer is the action line: "Schedule a private tour this week."
The prompt template, ready to paste: "You are a listing agent writing the just-listed MLS blurb for a [PROPERTY TYPE] at [ADDRESS]. Write four sentences — under 45 words total. Sentence 1 is a hook the buyer sees first; sentence 2 is one numeric fact; sentence 3 is one human element; sentence 4 is the call to action. Strip every adjective from your draft before returning the result ('beautiful,' 'stunning,' 'gorgeous,' 'charming' — gone). Plain text only."
The adjective rule is the bug fix. Without it, the model writes "beautiful three-bedroom with a stunning kitchen" — exactly the prose buyers skim past. Telling it to strip adjectives before returning forces it to lean on the facts and the softener, which is where the listing becomes memorable.
Use the same template twice in a row with different facts. The variation feels like two different agents wrote them rather than the same draft re-pasted — that is the whole reason listings need distinct blurbs rather than a single boilerplate. Twenty seconds per draft. Spend the rest of the hour on the photos, which are the part of the listing that actually sells.