AI Marketing for Real Estate Agents: A Four-Week Content Plan

Post one anchor piece a week and let AI turn it into the rest: a listing in week one, an area guide in week two, a market update built on your own figures in week three, and a process-and-proof piece in week four. Each anchor becomes a written post, a 45-second video script and a short email.

AI writes the words; you supply every fact. Prices, measurements, distances and dates come from your listing sheet and public records, because a model fills gaps with plausible details and a buyer who relies on a wrong one becomes your problem. That single rule shapes every prompt in this plan, and most of the checking time goes on it.

Follow me on Instagram@sagnikteaches

The four weeks on one page

Illustrative numbers for a solo agent with two active listings and a database of about 600 contacts. The time column is people-time once the AI drafts exist, with filming included at the length shown.

Connect on LinkedInSagnik Bhattacharya

That is 9.5 hours a month, or two to two and a half hours a week. If you have no listing in week one, swap in a buyer-side anchor such as "what a first showing is for". The rest of the plan does not change.

Subscribe on YouTube@sagnikteaches

Set up one project with your facts and your limits

Create a single project in ChatGPT or Claude. Both offer projects on their free plans, so every chat inside starts from the same instructions instead of a blank page. The block below runs unchanged in either. Fill the brackets once, paste it into the project's instructions, and paste each week's facts into a chat inside the project.

You help [agent name] of [brokerage] market homes and build a local profile. You write first drafts. [Agent name] checks every fact before anything is published.

VOICE: Warm, specific, plain. No hype. Short paragraphs. Write as "I".
Never use: stunning, nestled, boasts, dream home, won't last, must-see, hidden gem.

FACTS: Use only facts in the LISTING SHEET, AREA NOTES or MARKET TABLE I paste into a chat. If a fact you need is missing, write [ASK: what you need]. Never estimate, round up or infer. Never describe a material, finish, view or condition unless I have stated it.

PEOPLE AND PLACES: Describe the home and the area's facilities and distances. Never describe who lives there, who would fit in, or who a home is "perfect for". Never comment on schools' quality, safety, crime or demographics.

NEVER: predict prices; promise a sale price, a sale time or a return; give legal, tax or lending advice; name a client or quote a testimonial.

ALWAYS: end with [brokerage and licence line]. After every draft, add a FACT CHECK list: each number, measurement, distance, date and claim in the draft, with the source I gave you.

FORMAT: When I say WRITTEN, give a post of 600 to 900 characters. When I say SCRIPT, give a 45-second spoken script of 100 to 120 words with [SHOT] cues. When I say EMAIL, give three subject lines and a plain-text email under 200 words.

The FACT CHECK line is the most useful one. It does not make the draft correct, but it makes your check quick, because you hold each listed item against the sheet instead of reading the whole draft hunting for errors. The "people and places" paragraph exists because listing copy that describes who a home suits can breach fair-housing and equal-opportunity advertising rules in many markets. Ask your broker for the wider list of words your brokerage forbids, and paste it into the NEVER line.

Week one: launch a listing from the sheet, not the photos

Start from a listing sheet you wrote or checked, never from photos alone. An illustrative sheet for the week-one anchor:

LISTING SHEET
Price: $389,000
Beds: 3, plus a study with no built-in storage (not counted as a bedroom)
Baths: 2
Interior: 1,640 sq ft (from the survey)
Lot: 6,200 sq ft
Built: 1994. Kitchen redone 2021 (seller's receipt). Roof 2022 (seller's invoice).
Heating: gas furnace, 2019
Parking: single garage, driveway for two
Association fees: none
Property tax last year: $4,120 (public record)
Confirmed by me on site: south-facing back garden, laminate worktops, original hardwood in the living room
Measured walking times: park 6 min, bus stop 3 min, supermarket 11 min
Open: Saturday and Sunday, 12 to 3

Then one instruction in the project chat: "WRITTEN, for the listing sheet below, then five caption angles in one line each." An illustrative first draft:

Just listed: a light-filled three-bedroom on a 6,200 sq ft lot, priced at $389,000. The kitchen was redone in 2021, the roof is from 2022, and the south-facing back garden gets sun most of the day. The park is a six-minute walk and the bus stop is three. Open Saturday and Sunday, 12 to 3. Message me for the full details sheet.

FACT CHECK: $389,000 (sheet); 6,200 sq ft lot (sheet); kitchen 2021 (seller's receipt); roof 2022 (invoice); south-facing (my note); park 6 min, bus 3 min (measured); open Saturday and Sunday 12 to 3 (sheet).

It is close, and three phrases still need work. "Light-filled" has no source on the sheet, since only the garden's orientation is recorded. Either confirm it at the property or cut it. "Gets sun most of the day" turns "south-facing" into a claim about hours of light, so change it to "faces south". And "three-bedroom" should become "three-bedroom with a study", because the study is the thing a buyer will ask about.

The five caption angles it returned were price and headline, the 2021 kitchen and what the receipts cover, the south-facing garden, the walk to the park and bus, and the open-house reminder. Each takes about five minutes to expand and check. For more on turning one listing into a month of social posts, see AI real estate social media: a month of posts from your listings. Add a Business Profile update for the same listing: the update is a short description, a photo and a button, so use the south-facing garden photo and a button to the listing page.

Week two: an area guide that makes you the local source

A guide is where AI helps most and misleads most, because it will happily write about a neighbourhood it has never seen. So the facts come from a walk. Spend 45 minutes in the area and jot ten notes: the coffee place and its opening days, where parking is hard, the weekend market, the quiet street, the park gate that is locked at dusk. Add three or four facts you can verify from public sources, such as a bus operator's timetable page. Paste both into a chat as AREA NOTES and ask:

Write a 700-word area guide from the AREA NOTES below. Use five headings that a resident would recognise. Use only the notes. Where a heading has no note behind it, write [ASK: what I need]. Add a line at the end: "Details last checked [date]."

The draft comes back organised under headings like getting around, morning coffee and groceries, green space and weekends, what the homes are like, and questions I hear most. Your job is the other half: add one personal recommendation per heading (the thing only someone who works there knows), and put real numbers under "what the homes are like" from your own listing-service data, such as how many of last month's sales were detached houses and how many were flats.

A small sample shows the mechanism. Three illustrative lines of AREA NOTES: "Corner café opens 7am, Tuesday to Sunday, closed Monday (checked at the door)"; "Weekend market on the main street, first and third Saturday, 9 to 1 (organiser's page, checked this week)"; "Bus stop on the main street, route 12 every 12 minutes at peak (operator's timetable)". An illustrative draft paragraph: "Mornings start at the corner café, which opens at 7am from Tuesday to Sunday. On the first and third Saturday of the month the market takes over the main street from 9 to 1. The route 12 bus runs every 12 minutes at peak and gets you into the centre in about 20 minutes." The last clause is invented, because the notes hold no journey time. Replace it with "[ASK: journey time]", ride the bus once at the right time of day, and write down what you find.

Then read it for steering and for rot. Steering language describes who should live somewhere. Rot is a fact that will stop being true. Two illustrative AI lines and their repairs:

• "A safe, family-friendly neighbourhood with great schools." This describes people and makes claims you cannot substantiate. Replace it with something checkable: "Two parks within a ten-minute walk and a library open six days a week (hours checked on [date])."
• "The area attracts young professionals." Replace it with what is physically there: "Four cafés and a weekend market on the main street."

Cut the school lines entirely from AI drafts, as the project instructions say. Post the guide's best single fact as the week's Business Profile update, with a button to the guide, and date it: posts older than six months are archived unless they carry dates. To turn the same guide into a search-friendly page without steering, see AI real estate SEO: neighbourhood pages that avoid steering.

Week three: a market update from your own figures

This is the week where an assistant is most dangerous, because it will produce market statistics from nowhere. So the assistant never fetches or calculates anything. You pull four numbers from your listing service for the area, compare them with last month and the same month last year, and paste a table. An illustration:

Do the sums yourself before the AI sees them. $412,000 against $405,000 is up 1.7%, and against $398,000 is up 3.5%. Median days on market fell by three days on last month and six on last year. Active listings rose by three on last month and fell by three on last year. Now the instruction: "WRITTEN, describing only what the table shows. No predictions, no advice, no adjectives stronger than 'edged up' or 'eased'." An illustrative draft, run without that last clause:

Prices surged this month as demand soared, with the median sale price jumping to $412,000. Homes are flying off the shelves in under three weeks!

Three problems. "Surged", "soared" and "jumping" describe a 1.7% move. "Under three weeks" quietly turns a median into a promise about every home: half of last month's sales took 18 days or fewer, which is not the same thing. And the sample is 14 sales, so one or two deals can move a median, which is why a market update from a small area should say how many sales it rests on. The repaired line reads: "The median sale price edged up 1.7% on last month and 3.5% on a year ago, based on 14 sales. Half of the homes sold in 18 days or fewer."

Add a sentence for sellers that stays on the right side of advice: "If you are thinking about selling, what your home could fetch depends on its condition and the sales most like it, and I am happy to walk you through those." Check your listing service's rules on publishing statistics, since some require a source line or a disclaimer. For the email version of this update, with a method for keeping every figure checkable, see real estate newsletter with AI: a market update you can verify.

Week four: process and proof

The anchor is a process piece. The question every buyer and seller asks, in some form, is "what happens next?" Write the sequence you actually follow between an accepted offer and closing, in plain steps with who does what, and let the AI format it. Timelines vary, and you are not giving legal advice, so end with "your exact dates will be in your contract; I will walk you through them." The assistant must not invent steps, so paste your own list (for example: inspection, appraisal, financing, title, final walk-through) and ask it to explain each in two sentences.

The proof half has three parts. First, ask recent clients for a review with a plain message, and do not offer anything in return:

Hi [first name], it was a pleasure working with you on [the sale / the purchase]. If you have a few minutes, would you share how it went on my Google profile? Here is the link: [link]. Honest feedback of any kind is welcome. Thanks again, [your name]

Second, a "just sold" post, only with the seller's permission and inside your brokerage's and listing service's rules on showing prices. Third, any testimonial you publish must be the client's own words, used with permission. Do not ask the AI to polish one, because a smoothed quote can say something the client never did. Finish the month with 30 minutes of review: for each anchor, count the replies and enquiries it produced, choose the format to repeat, and add any new banned phrases to the project instructions.

Video scripts that fit 45 seconds

A script of 100 to 120 words usually runs somewhere between 40 and 55 seconds, depending on your pace. Read yours aloud with a timer, because your pace is the only one that matters. The structure that works for property is a one-line opening that names the thing and the price, three facts each tied to a shot, and a single action. An illustrative script for the week-one listing, in the format the project produces:

[SHOT: front of the house, you walking to the door]
This is a three-bedroom with a study, listed at $389,000, and I want to show you what buyers ask about first.

[SHOT: kitchen]
The kitchen was redone in 2021. Laminate worktops, and the seller has the receipts.

[SHOT: living room]
The original hardwood is still in the living room, and the roof was replaced in 2022.

[SHOT: back garden]
The garden faces south, and the lot is 6,200 square feet in total.

[SHOT: street]
There is a single garage and a driveway for two. The park is a six-minute walk, and the bus stop is three.

[SHOT: you, to camera]
Open Saturday and Sunday, twelve to three. Message me for the full details sheet.

[ON SCREEN: brokerage and licence line]

Notice what the script leaves out: every adjective with no source. At about 100 words of speech it should land near 45 seconds at an unhurried pace; time it aloud and cut the roof line if it runs over. If you plan to make video from photos with AI tools, AI real estate video from photos: what it costs and what to disclose covers the costs and the disclosure side.

The email: one anchor, one link, one reply

Each week's email is the anchor in plain text, under 200 words, with one link and one thing to reply to. Plain text looks like it came from you, because it did. The EMAIL instruction in the project returns three subject lines and a body; pick the subject that says something specific. A week-by-week set of subject lines and reply prompts, as illustrations:

Keep your newsletter list tidy and lawful: use people who asked to hear from you, include an unsubscribe link, and check the marketing-email rules that apply to your list before the first send. The reply prompts matter more than the open rate, because a reply is a conversation you can act on.

Facts the AI must never fill in

This is the verification table the whole plan hangs on. Print it, or paste it into the project as a reminder of where each fact comes from.

A line the AI invented from a photo

An illustrative failure from week one. The agent uploaded the kitchen photo and asked for a description without the listing sheet. The draft said "granite worktops and stainless-steel appliances." The sheet says laminate worktops and records nothing about the appliances. Nothing in the photo settled it, because lighting and a clean surface can look like stone to a model, and the model was happy to guess. Had it been published, the first sign would have been a buyer's raised eyebrow at the viewing, and a first impression of the agent as someone whose descriptions run ahead of the facts. In the worst case, a buyer who relied on the description could raise a misrepresentation complaint.

The repair is procedural. The facts sheet goes in first, and photos only support it. The project instructions say "never describe a material, finish, view or condition unless I have stated it." And the FACT CHECK list exposes the gap, because "granite (photo)" is not a source the agent gave. For a fuller catalogue of how listing copy goes wrong and how to check each type, see AI listing mistakes that can mislead buyers, and how to check. If you edit or stage photos with AI, label every altered image and keep the originals; a draft guideline from one listing service in July 2026 asks for an unaltered image of each staged room and allows staging only to add or remove personal property, and your own listing service's policy is the one that counts.

Counting the month

Judge the plan by conversations, not reach. At the end of each month, count the messages and calls that mention a post, a guide or an email, the showing requests, the replies to the email prompts, and any valuation conversations that began from the market update. The Business Profile's performance report shows how people found you, and Google aggregates its search terms by month, so read it once a month and let it choose next month's area. Three months is the smallest fair test.

Then do the sum. At 9.5 hours a month, if you value your time at $60 an hour, the plan costs about $570 a month in time. Set that against what one extra listing or one referred buyer is worth to you in commission; if it is more than a few months of that cost, three months of testing is cheap. If three months produce no countable conversation, keep the week-three market update and the review requests, which tend to produce replies, and reduce the rest to the anchor and the email.

Questions agents ask before they publish AI-drafted content

Can I post AI-edited or virtually staged photos?

Many listing services have rules on altered and staged images, and some places have laws on them. Common practice is to label every altered image and include an unaltered one for each staged room. One listing service's draft guidelines, published in July 2026, say exactly that and allow staging only to add or remove personal property. Check your own listing service's current policy before you post.

How many posts does a solo agent actually need?

The plan produces about five pieces from each weekly anchor, spread across the week. If time is short, publish the anchor itself, one social post and the email, and skip the rest. A smaller plan you finish every week does more for your local profile than a bigger one you abandon in week three.

Should I tell people I used AI to write a post?

I can't give you one rule for every brokerage and listing service, so ask your broker what your brokerage's policy says. Label AI-altered photos where required. Either way you are responsible for every word, so treat your name on the post as a promise that you checked each fact in it.

Can AI write my market update without my data?

Not safely. Without your figures it produces confident numbers from nowhere. Paste a small table from your listing service and tell the assistant to describe only that table. Then recalculate the percentage changes yourself, because a small area with 14 sales swings on one or two deals.

Should I use ChatGPT or Claude for this plan?

Both can do the work, and the prompts here run unchanged in either. Run the first listing through both with the same project instructions, keep the one whose drafts need fewer edits, and stay with it for a month before you compare again.

Further reads

• AI Virtual Staging for Property Listings: Costs and Disclosure — Costs and disclosure for staged photos.
• AI Lead Generation for Real Estate Agents: No Bought Lists Needed — Turn the content into enquiries without bought lists.
• Claude or ChatGPT for Realtors: Which Assistant Fits Your Day? — Which assistant fits an agent's day.
• AI Real Estate Listing Presentation: Build the Seller Deck — Build the seller deck from the same facts sheet.
• How to Automate Buyer and Vendor Follow-Up With AI — Follow up the people your posts attract.
• How Estate Agents Use AI to Write Property Descriptions — A deeper look at listing descriptions.
• Home Stager AI: Adding Paid Virtual Staging to Your Service Menu — Four packages, a price built from tool cost and your own minutes, a hybrid physical-plus-virtual offer, and the disclosure wording to use.
• Rechat AI for Brokerages: What Lucy Does and What It Costs — Lucy's features from Rechat's own help pages, the pricing Rechat does not publish, and a 30-day pilot plan for a ten-agent office.
• Virtual Assistant or AI for a Real Estate Agent: What Goes Where — Which real estate tasks suit AI and which need a virtual assistant, with costs, client-data rules and a 30-day trial plan.
• Best AI for Real Estate Marketing: Ads, Email and Listing Video — Pick an AI tool by job for real estate marketing: ad copy, email nurture sequences and listing video, with fair housing and disclosure checks.
• ChatGPT for Estate Agents: Rules for Vendor Updates and Feedback — For the branch or brokerage owner: the rules, templates and fact checks that let agents use ChatGPT for vendor updates and viewing feedback safely.
• Open House Sign-In Sheet With AI: Follow-Up Within the Hour — A printable sign-in sheet, consent wording, buyer-type message templates and a 60-minute plan for personal follow-up after every open house.
• AI Policy for a Real Estate Brokerage: What Agents May and May Not Do — Turn a broker AI policy into rules an agent can follow on a Tuesday: do and don't lists, sample clauses and a sign-off sheet.
• Can ChatGPT Write Real Estate Listings That Pass Fair Housing? — A listing-copy workflow with an edit-pass prompt, before-and-after wording and a human fair-housing check.
• Can Estate Agents Be Found in AI Search? What to Fix Now — A fix list for estate agencies that want to appear when buyers and sellers ask AI for local agents, with a monthly check you can run yourself.
• Can Estate Agents Use AI Photo Edits Without Misleading Buyers? — A line-by-line guide to AI listing-photo edits: what is routine, what misleads, the disclosure wording and a sign-off routine for the branch.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT and Claude project help and plan pages, Google Business Profile help on posts, and one listing service's draft guidelines on staged and AI-enhanced photos (July 2026), as checked in October and November 2026.