You already pay for the source document. Every quarter, some combination of CoStar, Yardi Matrix, ALN, RealPage or your own market survey lands in your inbox, you skim the occupancy number, you screenshot one chart for the asset management deck, and then it dies in a folder called "Q3 stuff (final) v2."
That report is a month of content. Not because AI is magic, but because the expensive part of content is having something true to say, and you just bought that. The writing part is the cheap part now.
Here's the repurposing pass: one source doc in, one newsletter, six social posts, and a leasing talking point out. Ninety minutes, start to finish, and most of that is you editing.
Step 1: Pick a document that actually has numbers in it
Not a blog post. Not a trend piece. You want something with rent comps, absorption, supply pipeline, concession data, or your own leasing numbers. Good candidates:
- Your submarket's quarterly report from whoever you subscribe to
- Your own market survey spreadsheet, the one your leasing teams update on Mondays
- A city planning commission agenda listing new permits near your community
- Your last 90 days of CRM data: source mix, tour-to-lease, average days to lease
The last one is underrated. Nobody else in your market can publish your data, which makes it the only thing you have that's genuinely yours.
Step 2: Extract the claims before you write anything
This is the step people skip, and it's why the output reads like a horoscope. Before you ask for a single social post, upload the document and ask the model to pull out facts, not themes.
Ask it: "Extract every quantitative claim in this document that relates to [SUBMARKET]. For each one, give me the number, the time period, and the page or section it came from. Do not interpret. Do not summarize. Just list."
You'll get 15 to 40 claims. Read them. Delete the ones that are boring, wrong, or would make your owner nervous. What's left is your content calendar, and it took you eleven minutes.
Now here's the important part: check the numbers against the source. Language models are enthusiastic about arithmetic and bad at it. If it says rents grew 4.2% and the chart says 2.4%, it's not lying to you on purpose, it just transposed a digit and would happily do it again with your name on the post.
Step 3: The newsletter first, everything else after
Write the long thing first and cut it down. The other direction produces six posts that don't agree with each other.
Your newsletter should be 400 to 600 words with one argument. Not "here's the Q3 market update." Something like "new supply in the [X] corridor peaks this spring, which means the best renewal offer you'll see all year is the one on your desk right now." That's a real point of view, it's supportable, and it makes a resident do something.
Give the model your three or four best claims, your audience, and the argument. Do not give it the whole report and hope. It'll write the beige version, the one that says "the market remains dynamic," and you'll rewrite it from scratch anyway.
Step 4: Slice into six posts, with different jobs
Six posts from one newsletter, and each one has a different job. Tell the model that explicitly, because left alone it will write the same post six times in slightly different fonts.
1. The number. One stat, one sentence of context. Works on LinkedIn. 2. The counterintuitive read. "Everyone's watching new deliveries. The thing that actually moves your rent is [X]." 3. The local proof. Something visible from the property. A new grocery, a road project, a permit. 4. The resident-facing version. Same fact, translated into "what this means for your lease." 5. The team version. For your own leasing associates. Give them one line they can say on a tour. 6. The question. Ask your audience something you actually want to know, then read the replies.
Ask for platform-specific formatting: LinkedIn gets a hook line and white space, Instagram captions get shorter with the point up top, and a resident email gets a subject line under 45 characters.
Step 5: The leasing talking point
This is the piece that pays for the whole exercise, and almost nobody does it.
Take your best claim and turn it into two sentences a leasing associate can say out loud without sounding like they're reading a bond prospectus. Something like: "About 1,400 new units delivered in this submarket this year, so pricing is soft right now. That's why we can do what we're doing on this unit, and it's not going to look like this in March."
Then pressure-test it. Ask the model: "A prospect responds with [OBJECTION]. Give me three responses under 20 words each that don't overpromise." Print it. Put it in the leasing binder. Your team will use it because it's short and it's true.
The one thing that will get you in trouble
When you get to the resident-facing and ad-facing posts, models write about the person, not the place. Ask for a post about your renovated units and you'll get "perfect for young professionals" or "great for families" without being asked. That's a fair housing problem: describing who should live somewhere, rather than what the property offers, is exactly the kind of language HUD guidance treats as a preference statement.
The fix is a standing instruction in every prompt: describe the unit and the community, never the resident. Say "in-unit washer and dryer, 12 minutes to downtown" instead of "ideal for busy commuters." Then read the output yourself before it ships. Your compliance review still applies to anything AI wrote, and the model does not know your fair housing policy.
What breaks
Two failure modes worth knowing.
The model will invent a comp, a submarket name, or a percentage that sounds plausible. Always check against the source doc.
And if you run this pass every quarter with the same prompt, your content starts to sound identical quarter over quarter. Change the argument each time. Nobody unsubscribes from a bad post, they unsubscribe from the fourth one that sounds like the third.
Do this once and you'll have next month's calendar built before lunch. Do it four times and you're the only marketer in your company who never opens a blank content calendar in a panic on the 28th.