Pull the month's figures from your MLS or association report into a spreadsheet, do the sums there, then give AI only those verified numbers and ask it to write around them. Keep a source line for every figure, and send only after a second person or a fresh read confirms each number against the source.
The prose is rarely the risk in an AI-written market update. The risk is a plausible number that nobody can trace: "average days on market fell to 23" when the report said 32, or a median price described as an average. Readers who are about to list a home will quote your newsletter to a spouse or a lender. The method below is built around one rule: the assistant never supplies a figure, it only receives and phrases them.
Decide the five figures before you open any data
A monthly update needs fewer numbers than most agents think. Pick five and keep them identical every month, so readers learn to compare. A sensible default set for a residential market looks like this.
| Figure | What it tells a reader | Watch the definition |
|---|---|---|
| Median sale price | What a typical home sold for | Median, not average; say which property types are included |
| Closed sales | How active buyers were | Closed in the month, not pending |
| Active listings at month end | How much choice buyers have | Count date matters; use the last day of the month |
| Median days on market | How fast homes sell | Some systems reset the count on relisting |
| Months of supply | Whether the market favours buyers or sellers | Often active listings divided by the month's closed sales; some reports use pending sales or a 12-month average |
Add a sixth, such as the sale-price-to-list-price ratio, only if sellers among your readers ask about it. Each extra number is another thing to verify. Write down the geography too: a neighbourhood, a school district, a set of neighbourhoods. "The market" is not a place. A newsletter for a suburb with 40 sales a month will show wild swings that a wider regional report smooths out, so state the area at the top of every issue.
Pull the data and build a figure register
Export or copy the statistics from your MLS's market-statistics tool, your association's monthly report or your brokerage's reporting dashboard. Take the summary figures, not individual listing records. Many MLS licences limit how listing-level data can be shared or republished, so read yours once; if it restricts reuse of its statistics, say so in the notes and find out what attribution it requires.
Open a sheet called "Figure register" with these columns: figure, this month, last month, same month last year, source (the report or screen name), date pulled, definition. Fill it by hand from the export. This is the step that makes the newsletter verifiable, because every number the AI will see is now in a row with a source next to it.
Then compute the changes in the sheet with formulas, not in the assistant. A change from $412,000 to $425,000 is 3.2 per cent: =(425000-412000)/412000. On the simple convention used here, months of supply is active listings divided by that month's closed sales, so 318 active listings and 106 closed sales gives 3.0 months. Reports differ on this one, so copy whatever method your MLS or association report uses and say which in the definition column.
A filled-in register for an illustrative suburb
These figures are made up for illustration and describe no real market.
| Figure | This month | Last month | Same month last year | Source |
|---|---|---|---|---|
| Median sale price | $425,000 | $412,000 | $401,500 | Association monthly report, table 2 |
| Closed sales | 106 | 94 | 118 | MLS statistics tool, closed, single-family |
| Active listings (month end) | 318 | 296 | 241 | MLS statistics tool, active |
| Median days on market | 27 | 31 | 22 | Association monthly report, table 4 |
| Months of supply | 3.0 | 3.1 | 2.0 | Calculated: active divided by closed |
Look at the register before any drafting. Here a careful reader would notice that closed sales fell about 10 per cent against last year while active listings rose by about a third, and that supply moved from roughly two months to three. That is the story, and you found it from the sheet, not from the assistant.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowPrompt the draft from the register only
Paste the register as plain rows and give the assistant tight boundaries. The wording below keeps it from inventing context.
Write a monthly market update email of 250 to 350 words for homeowners and prospective buyers in [neighbourhood name].
USE ONLY THE FIGURES IN THE TABLE BELOW. Do not calculate new figures, round differently, or add any statistic, trend, forecast or reason that is not in the table. If you want to mention something that is not in the table, put it in a list titled "Questions for the agent" instead.
For each figure you mention, put its source name in square brackets straight after it, exactly as written in the Source column.
Structure: a one-sentence headline of what changed; three short paragraphs (prices, how many homes sold and how fast, what buyers can choose from); then two sentences each for "If you are thinking of selling" and "If you are thinking of buying". No predictions. Plain, warm wording.
TABLE:
[paste the register]An illustrative draft, trimmed:
Headline: More homes to choose from, and they are taking a little less time to sell than last month. The median sale price was $425,000 [Association monthly report, table 2], up from $412,000 in the previous month and from $401,500 a year ago. 106 homes closed [MLS statistics tool, closed, single-family], compared with 118 in the same month last year. Buyers had 318 active homes to choose from at month end [MLS statistics tool, active], roughly three months of supply [Calculated: active divided by closed]...
Check three things in the draft. Every number appears in the register with the same rounding. No word like "surged" or "plummeted" has been added where the data shows a few per cent. And the assistant has not blended two geographies, for instance by describing a wider region because it "knows" about it. If any of that fails, correct the draft by hand, not with another prompt, since a second prompt can introduce a new error elsewhere.
The verification step, done properly
This is the part most newsletters skip. Do it in this order, and ideally have a different person do it.
- Print or open the draft beside the register. Tick each number in the draft against the register cell it came from.
- Open the original report or MLS screen for two of the five figures and confirm the register copy. Choose a different two each month, so over a quarter every figure is spot-checked.
- Check the date range in the first sentence matches the data (month and year, not "recently").
- Check every comparison word against the sign of the change: "up", "down", "fewer", "faster" are the usual casualties.
- Delete the bracketed source labels or convert them into a footnote line. A tidy version is "Sources: association monthly report (tables 2 and 4); MLS statistics for [neighbourhood], pulled on [date]".
- Save the register and the sent version together in a folder named by month. If a reader challenges a number in March, you can prove where it came from.
A realistic mistake shows why step 4 exists. Suppose the market improved in speed: median days on market dropped from 31 to 27. A draft that says "homes are taking longer to sell" has the direction reversed because the assistant confused days on market with "days of delay". It reads fluently, it is wrong, and only a tick against the register catches it.
Ideas for the part that is not numbers
A newsletter of five figures is a report, and reports get filed. Add one short section that is yours. Here are six ideas, each of which an assistant can draft from a few notes you supply.
- One street, one story. A sentence or two on a recent sale that shows the trend, with the owner's permission and no address. Give the assistant your notes; it can tighten them.
- The question of the month. Take a question a client asked, such as "does a new kitchen pay back before I sell?", and answer it in 100 words. Write the facts yourself and ask AI to simplify the wording.
- A seller checklist for the season. Ten items for autumn gutters and kerb appeal, trimmed from your listing-prep sheet.
- A buyer's cost reminder. A plain list of costs beyond the price. Verify with your closing partners before quoting any amounts.
- Local events. Three dates you have confirmed with the organisers. Never take dates from an assistant's memory.
- A "what I'm watching" line. One sentence with no prediction: "I'm watching how many listings go under contract in the first two weeks, because that is where the change shows first."
If you want a general routine for the non-numeric parts, writing a customer newsletter with AI in 30 minutes covers the drafting loop. This tutorial adds the verification layer that a property update needs.
Two more templates you can reuse
A one-paragraph version for a quiet month
Some months have too few sales to say anything. Do not stretch it. Ask the assistant for 100 words saying how many homes sold, that the sample is small, and that the figure may change when late reports arrive. A sentence such as "with only 14 closed sales, one or two high-priced homes can move the median noticeably" is honest and explains the swing.
A before-and-after of a weak line
Before: "The market is hot and prices are shooting up, so now is the perfect time to sell!" After: "The median sale price rose from $412,000 to $425,000 [Association monthly report, table 2]. With three months of supply, buyers have some choice, so a well-priced home is still selling in about four weeks." The second version holds up when a reader checks it, and it gives sellers a reason to talk to you without a claim you cannot support.
Segment the same figures for two readers
Once the register exists, one extra prompt turns it into two versions without new data. Sellers care about speed and price; buyers care about choice and how quickly a good home goes. Ask the assistant to rewrite the same draft for each reader, reusing the same bracketed sources, and to keep the figures unchanged. The seller version opens with the median price and days on market; the buyer version opens with active listings and months of supply.
Say your list has 340 contacts, of which 90 are people you know are thinking of buying and 60 are thinking of selling. Send the buyer version to the 90, the seller version to the 60 and the general version to the remaining 190. The numbers are identical in all three, so the verification step is done once. What changes is the first paragraph and the closing two sentences. If your CRM cannot tag by intent, do not guess: send the general version to everyone and add a one-line link, "Thinking of buying or selling? Reply and I will send the version for you."
Use a second assistant pass to audit, not to write
After the draft is edited, you can use the assistant as a checker, as long as you treat its answer as a prompt for your own check and not as proof. Start a fresh chat, paste the register and the final text, and ask:
Here is a table of figures and a draft email. List every number in the email and, for each, say which table cell it matches, or "no match". Flag any comparison word (up, down, more, fewer, faster, slower) that does not match the sign of the change in the table. Flag any claim that is not supported by the table. Do not rewrite the email.An illustrative reply would be a list such as "$425,000: matches 'Median sale price, this month'. 'Fewer buyers': no match, the table shows closed sales, which is not the same as buyers." That second line is the useful kind of catch: the draft had turned "closed sales" into "buyers", which sounds the same and means something different. The audit pass adds about two minutes and tends to find one or two wording slips a month. It does not replace ticking against the register, because the assistant can miss a transposed digit that you would see.
A second month, with a less comfortable result
To see how the method handles bad news, take the next month for the same illustrative neighbourhood. The median price slips from $425,000 to $419,000, closed sales fall to 88, active listings rise to 341 and median days on market climb to 34. The computed change in price is about minus 1.4 per cent, and months of supply is 341 divided by 88, which is 3.9.
A weak draft would hide this behind "a more balanced market". The draft that holds up says it directly: "The median sale price slipped to $419,000, down from $425,000. Fewer homes closed (88), more were on the market at month end (341), and the typical home took 34 days to sell, up from 27. That is about 3.9 months of supply, so buyers have more choice and negotiating room than last month." Then one sentence for each reader: sellers should price to the market, not to last spring's sale next door, and buyers should still expect competition for the best homes. That is accurate, it is useful and it does not predict where prices go next.
One more habit helps: write the register notes in plain words as you go. A note such as "days on market resets when a home is relisted, so this figure runs slightly low" costs ten seconds in the definition column and saves a confusing reply in the inbox. Over a year the notes become a short glossary you can link from the newsletter footer, which readers use to understand terms like months of supply without writing to you.
Subject lines and preview text, written from the facts
Ask for ten subject lines from the verified figures only, and then choose one that states the finding. "Three months of supply: what it means for your home" says something. "Your monthly market update!" does not. Check that the subject line, like the body, matches the register: a subject that says "prices fall" when the median dropped 1.4 per cent will provoke replies about a crash. Aim for subject lines under 50 characters and preview text that adds one new fact rather than repeating the subject.
Keep a note of which subject lines got more opens in your email tool and which led to replies. Opens are an unreliable measure on their own, because some mail apps load images automatically, so judge subject lines by replies and link clicks as well.
The send checklist
- Neighbourhood and month named in the first sentence.
- Every figure ticked against the register; two re-checked at source.
- Changes recomputed in the sheet, not by the assistant.
- Source line at the bottom, with the pull date.
- Your sender name, business name, a valid postal address and a working opt-out link, because commercial email generally needs them.
- Your brokerage's required disclosures in the footer, where your broker requires them.
- No client names, addresses or sale details you lack permission to share.
- Links tested; subject line under about 50 characters; preview text written.
- A test send read on a phone.
- Register and final copy saved in the month's folder.
The send-time checks overlap with the general list in the email newsletter QA checklist; use that for formatting and deliverability, and use the register for the numbers. If your list also receives a loan-related update, the rules that apply to lenders are covered in AI market updates for loan officers.
Who should receive it, and how it fits the year
The newsletter works best as the monthly thread through a longer follow-up plan. Past clients and neighbours are the natural audience, and the schedule for them, with anniversaries and tier-based frequency, is covered in the sphere-of-influence touch plan. Keep the list in your CRM rather than in an email tool alone, so that opt-outs flow back into every other message; the realtor CRM comparison shows what small teams pay for that.
Time budget for a steady month: about 25 minutes to pull and record figures, 10 to compute changes, 5 to prompt and edit, 15 for the verification step and 5 to send, so about an hour. In the first month, expect nearer two, because you are building the register. From the third month the register is mostly copying.
Review the newsletter every quarter. Look at opens, replies and, more usefully, the conversations it started: "I saw your update and wondered what my house is worth". If readers reply with questions about one figure, that figure deserves a sentence of explanation next time. If nobody ever replies, shorten the update and add a question at the end.
Questions about AI market-update newsletters
Can I paste MLS data into ChatGPT or Claude?
Check your MLS's data licence first, because some limit how listing data may be reused or shared. Aggregate statistics you export or copy from a published report are usually fine, but individual listing records may not be. Use a business-plan assistant, and paste summary figures rather than raw listing exports.
Should AI do the arithmetic for months of supply or percentage changes?
No. Do the sums in a spreadsheet and paste the results. Language models can slip on arithmetic and will state a wrong figure with total confidence. Let the assistant write the sentence, and let the spreadsheet produce the number.
How long should a monthly market update be?
Around 250 to 400 words with three to five figures is enough. A reader should finish it in two minutes and know whether prices, stock and speed moved. Longer reports belong behind a link, not in the email body.
What if the numbers show a falling market?
Say so plainly, with the figure and the period. Readers notice spin, and a newsletter that only reports good news stops being trusted. Explain what it means for a seller and for a buyer in a sentence each, and avoid predictions.
Further reads
- Bookstore Newsletter Template: AI-Polished Staff Picks and Events — A different newsletter template, for staff picks and events.
- Gym Newsletter Examples: A Monthly Member Email Written With AI — A monthly member email, written from a fixed template.
- How to Automate Buyer and Vendor Follow-Up With AI — Follow up the replies the newsletter brings in.
- ChatGPT for Realtors: Brokerage Rules for Emails, Disclosures, Follow-Ups — Brokerage rules for what AI may touch in emails.
- Follow Up Boss Pricing: What a Small Team Really Pays — What a CRM to hold the newsletter list costs a small team.
- AI Marketing for Real Estate Agents: A Four-Week Content Plan — Week by week: a listing launch, an area guide, a market update from your own figures and a proof week, each turned into posts, a script and an email.
- AI Real Estate Listing Presentation: Build the Seller Deck — How a realtor can build a listing presentation with AI: the slide spine, a worked comps page, a costed marketing plan and checks before the meeting.
- AI Real Estate SEO: Neighbourhood Pages That Avoid Steering — How agents can draft neighbourhood pages with AI without steering language: verified facts, a page template, wording rewrites and a review routine.
- AI Real Estate Social Media: A Month of Posts From Your Listings — Plan four weeks of listing and just-sold posts with AI from your own source sheet, with fair-housing and brokerage disclosure checks built in.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ChatGPT pricing page and Claude pricing page (no training on business data by default); NAR existing-home sales methodology; the FTC's commercial email compliance guide; the reader's own MLS data licence.