Using AI for Real Estate Marketing: Local Posts

How agents use AI for local marketing posts

Every record follows the same pattern:

1. Give it something local. The agent hands the AI local input.
2. Get content back. The AI turns it into a post, script or newsletter.
3. Send it out. The content goes out.

What changes is the input, and how much of the work the AI does after the words.

4 records from 3 people cover this task. The AI tools for real estate marketing they name are ChatGPT and Claude, with Canva for the design. 2 name ChatGPT and 2 name Claude.

Tools named, per record (social media and local market content)

1. ChatGPT2 of 4
2. Claude2 of 4
3. Canva2 of 4

There are three variants:

• Your own MLS data into a neighbourhood post. @sandyisyouragent pulls every active, under-contract and sold home in one subdivision from the last six months. She exports it from the MLS as a CSV file and uploads it to the paid version of ChatGPT. She asks for key insights and a Facebook post, an Instagram post or a video script with stats. Then she adds a photo of the subdivision sign, her name and her firm, and posts it in that community’s Facebook groups every few weeks (@sandyisyouragent).
• Neighbourhood names into a video script. @kristamashore_ asked ChatGPT to compare two Brentwood, California neighbourhoods, asked follow-up questions, then asked for a one-minute TikTok script (@kristamashore_).
• A Canva template that Claude fills in. Carrie Dahle connects Canva to Claude Cowork. She sends Claude the link to a local-guide template and asks it to customize the post and swap in photos from her Canva folders. Claude can also write the caption. She saves the steps as a Claude skill to run again (Carrie Dahle). A second video uses the same setup for her monthly email newsletter (Carrie Dahle).

• 

  TikTok2025-04-22

  @sandyisyouragent — social media and local market content

  The agent goes to the MLS hot sheet or stats for one subdivision and pulls every active, under-contract and sold home from the last six months.

  ChatGPT (paid) + MLS export · Says "so much community engagement" (no number)

The inputs differ. @sandyisyouragent’s record is the only one that starts from the agent’s own market data. That gives the post real local numbers, not general facts about a place (@sandyisyouragent).

What they report

1 of 4 records states a number, and it is self-reported.

Reported results (social media and local market content, self-reported)

1. @kristamashore_ after: ChatGPT (2024) · before: not stated~3 min

• @kristamashore_: About 3 minutes for comparison + script

@kristamashore_ says the neighbourhood comparison and the script took about three minutes (@kristamashore_). She runs AI events for agents and promotes her next event in the caption, so read the number with that in mind.

The other records give no number:

• @sandyisyouragent says she has “gotten so much community engagement” (@sandyisyouragent).
• Carrie Dahle says the Canva posts save her “so much time” and calls the newsletter “a huge time-saver” (Carrie Dahle, 2 cases).

No record reports leads, listings or sales from this content.

Where it goes wrong

The cases name few errors. The bigger risk is the steps they do not show.

It does not sound like you

@sandyisyouragent says: “don’t just copy paste everything ChatGPT says. Put your own spin on it.” You may have to tell it to be shorter or more informative (@sandyisyouragent). @kristamashore_ does not read the script word for word; she makes it more personal (@kristamashore_).

The design needs a check too

After Claude swaps the images into the Canva post, Carrie Dahle tells viewers to “go check its work” (Carrie Dahle). She does not describe a specific error.

Automatic does not mean reviewed

Her newsletter runs as a scheduled task on the 1st of each month. Claude gathers content from websites she does not name and fills the Canva template. The video shows no review before the newsletter goes out (Carrie Dahle).

• 

  YouTube2026-09-03

  Carrie Dahle — social media and local market content

  Carrie set up a scheduled task in Claude Cowork that runs on the 1st of each month and gathers the content for her email newsletter from the websites it is set to visit (she does not name them).

  Claude Cowork + Chrome extension + Canva · Calls it "a huge time-saver" (no number)

Nobody shows how they check the facts

• @kristamashore_: her comparison of schools, amenities and housing came from ChatGPT with no local data uploaded. The video does not show how she checks it (@kristamashore_).
• @sandyisyouragent: her stats come from her own MLS export, but her record does not mention a check of the numbers ChatGPT gives back either (@sandyisyouragent).

From ChatGPT scripts to Claude and Canva

The 2024 record shows the earlier pattern:

1. Ask by name. Ask ChatGPT about two neighbourhoods.
2. Follow up. Ask more questions.
3. Turn it into a script. Use the answer for a video.

The agent uploads no data and connects no design tool. The AI writes the words and the agent does the rest (@kristamashore_).

The current records change two things:

• Local data goes in. An MLS export for one subdivision replaces asking the AI what it knows about a place (@sandyisyouragent).
• The AI does the design work too. Claude fills a Canva template with photos and captions, and one newsletter runs on a monthly schedule (Carrie Dahle, 2 cases).

The second change also moves the agent further from the output. In 2024 the agent filmed the script herself. In 2026 a scheduled task can build the newsletter with no step shown between the AI and the reader (@kristamashore_; Carrie Dahle).

The sample is small: 1 record from 2023–24 and 3 from 2025–26. Both Claude records come from the same agent, one day apart (Carrie Dahle, 2 cases). Treat this as a direction, not a trend.

What to take from these cases

These points come only from the records above:

1. Start from your own data. An MLS export for one neighbourhood gives the AI real numbers to work with (@sandyisyouragent).
2. Make it yours before you post. Add your spin, your photo and your name. Do not paste the raw output (@sandyisyouragent; @kristamashore_).
3. Check the facts and the design. None of these records shows how the agent checks the numbers, so that step is yours to design (@sandyisyouragent; @kristamashore_). Look at every image the AI swapped in (Carrie Dahle).
4. Add a review step to scheduled jobs. If a task runs by itself, look at the result before it reaches your list (Carrie Dahle).
5. Save the steps. A saved Claude skill lets you run the same post again from a new template (Carrie Dahle).

Know the limits:

• Only three people appear in these records.
• The only number comes from an agent who runs and promotes AI events (@kristamashore_), and 1 record on this page has a business tie.
• The other results are opinions with no number.

See every record in the case register and the method in How we collect cases.

Questions agents ask

How do real estate agents use AI for social media?

In these records, they turn local input into posts: an MLS export into Facebook and Instagram posts (@sandyisyouragent), two neighbourhood names into a TikTok script (@kristamashore_), and a Canva template into a local-guide post (Carrie Dahle).

Can AI write a real estate market update?

One agent uploads six months of her own MLS data for one subdivision to ChatGPT. She asks for key insights and a post with stats (@sandyisyouragent). Another got a one-minute script that compares two neighbourhoods, with no local data uploaded, and the video does not show how she checks it (@kristamashore_).

Can Claude make Canva posts for realtors?

In one agent’s setup, yes. Carrie Dahle connects Canva to Claude Cowork. Claude customizes a local-guide template, swaps in her photos and can write the caption, and she tells viewers to check its work (Carrie Dahle). The same setup builds her monthly newsletter on a schedule (Carrie Dahle).