How realtors can use ChatGPT for market research without fabricating data

ChatGPT is useful for market research, but not in the way most realtors try to use it. It cannot pull MLS data directly. It will hallucinate statistics if you let it. Used correctly, with browsing mode on and source-anchored prompts, it becomes a solid research assistant for summarizing, structuring, and drafting. The workflow below is how I approach it at Monstera, and it's what I teach in my coaching work.

The fabrication problem and why it's worse for realtors

When a content marketer uses ChatGPT to write a blog post and it invents a statistic, the risk is embarrassing. When a realtor sends a client a market summary with a fabricated average sale price or days-on-market figure, the risk is a lot higher. You're advising someone making a decision worth hundreds of thousands of dollars.

The pattern I see most often: a realtor opens ChatGPT, types "what's the average condo price in [neighbourhood] right now?", gets a confident-sounding answer, and copies it into an email without checking. The number might be from training data that's 18 months old. It might be a statistically plausible figure the model generated because it fits the pattern of things it's seen. Either way, it's not sourced.

The fix isn't avoiding ChatGPT for research. The fix is a workflow that treats ChatGPT as a synthesis and drafting layer, not a data source.

Setting up browsing mode correctly

ChatGPT's browsing mode is available on paid plans (ChatGPT Plus at $20/month, or Team/Enterprise tiers). You can verify what's current on OpenAI's ChatGPT pricing page.

To enable it, open a new conversation, click the model selector at the top, and choose a model that supports web browsing. As of 2026, GPT-4o has browsing built in by default on paid plans. You don't always need to toggle it manually, but it's worth confirming the model is actually citing live sources and not drawing from training data alone.

One way to check: ask it to tell you today's date and what the most recent data point it found was. If it hesitates or gives you a training cutoff date instead of a live search result, browsing isn't active in that session.

Even with browsing on, ChatGPT can't access paywalled databases, your board's MLS portal, or data behind a login. What it can access: public-facing reports from TRREB, CREA, municipal planning documents, news coverage of specific developments, and publicly posted brokerage market reports.

How to write source-anchored prompts

A source-anchored prompt tells ChatGPT where to look, not just what to find. This is the most practical change most realtors can make right now.

Generic prompt (avoid this):

"What's the average sale price for condos in the Toronto waterfront market?"

Source-anchored prompt (use this instead):

"Search TRREB.ca and CREA.ca for the most recent published data on average condo sale prices in the City of Toronto. Cite the specific report name, publication date, and the exact figure. If you cannot find a recent source, tell me that rather than estimating."

The last sentence matters. ChatGPT tends to fill gaps rather than admit uncertainty unless you explicitly give it permission to say it doesn't know. "Tell me that rather than estimating" is the instruction that catches a lot of would-be hallucinations.

Other anchors worth naming in your prompts: Statistics Canada housing reports, local planning department documents (for zoning or development context), and the websites of specific developers or community associations when you're researching a building or neighbourhood.

The verification step (non-negotiable)

Whatever ChatGPT returns, your job before using it is to open the source it cites and find the number yourself.

The verification workflow is three steps:

1. ChatGPT gives you a figure and cites a source.
2. You open that source (the actual report, not a summary of it).
3. You find the number in the source and confirm it matches what ChatGPT told you.

If the source doesn't contain the number, or the source doesn't exist, don't use the figure. This sounds obvious but it takes about 90 seconds and most realtors skip it because the ChatGPT output looked authoritative.

The places I go most often to verify Toronto-area data are the TRREB Market Watch page (monthly, free, downloadable PDF), the CREA stats page, and for local development context, the City of Toronto's development applications portal.

When to use Perplexity instead of ChatGPT

Perplexity is purpose-built for sourced research. Every answer surfaces citations inline, and you can click through to the source immediately. For tasks where you need to show your work, or where the verification step needs to be fast, Perplexity tends to be more reliable than ChatGPT's browsing mode.

The pattern I use: Perplexity for the fact-gathering phase, ChatGPT for the drafting phase.

Concretely, that looks like this:

• Open Perplexity, ask it for the most recent TRREB data on a specific neighbourhood and sub-type. Read the citations it surfaces. Copy the verified figures.
• Open ChatGPT, paste in the verified figures, and ask it to draft a two-paragraph market summary for a buyer presentation or a newsletter section.

Perplexity is also better for news-style research: "what's been reported about [development project] in the last 6 months?" gives you a more reliably sourced answer in Perplexity than in ChatGPT. Perplexity's site has a free tier that's workable for occasional use; the Pro plan at around $20/month gives you higher daily limits and access to better models.

What ChatGPT is genuinely good at in this workflow

Once you have verified data, ChatGPT is strong at a few specific tasks:

Summarizing long documents. Paste in a status certificate, a strata report, a developer's disclosure statement, or a planning document. Ask it to pull out the key risks, the dates that matter, the fees, and the red flags. It's faster than reading 80 pages yourself, and it surfaces things you might skim past.

Drafting market commentary. Once you have verified numbers, ChatGPT can turn them into a readable client-facing summary faster than writing from scratch. Give it the data, your tone preference, and a word count.

Structuring comparables narratives. "Here are five recent sales in this building. Help me write a one-paragraph narrative explaining what they suggest about current value." It won't pull the comps for you, but it will help you write about them clearly once you have them.

Preparing questions for listing appointments. "Based on these neighbourhood market conditions, what are the three or four questions a buyer's agent would likely raise about pricing strategy?" It's good at anticipating objections from data you supply.

The common thread: ChatGPT works well as a synthesis and drafting layer. It's not a data source. Treat it that way and it's genuinely useful. Ask it to be the data source and you'll eventually send something wrong to a client.

What I'd actually do

My starting point for any market research session: Perplexity first to gather cited facts, then ChatGPT to structure them into something usable. I don't skip the verification step even when I'm in a hurry, because the reputational risk of a wrong number in a client email isn't worth the two minutes I saved.

The other thing I'd do is build a saved prompt template in ChatGPT for the research tasks you repeat most often. A status certificate summary prompt, a neighbourhood market commentary prompt, a competitive analysis prompt. Writing a good source-anchored prompt once and saving it is better than improvising a new one every time.

If you want to pressure-test your workflow, run the same research question through both ChatGPT and Perplexity and compare what gets cited. The differences in sourcing behaviour will tell you a lot about where each tool is reliable and where it needs a tighter leash.

---

FAQ

Can realtors use ChatGPT for market research? Yes, but with guardrails. ChatGPT with browsing enabled can pull recent sales data summaries, news, and market commentary. It cannot access MLS directly. Every number it produces needs to be verified against a primary source — TRREB, CREA, or your board's own published data — before you share it with a client.

What is browsing mode in ChatGPT and why does it matter for market research? Browsing mode lets ChatGPT search the web in real time rather than drawing only from its training data. For market research, this matters because real estate data changes monthly. Without browsing enabled, ChatGPT may cite statistics that are 12-24 months out of date with no indication they're stale.

What is a source-anchored prompt? A source-anchored prompt instructs ChatGPT to pull from specific, named sources rather than generating a plausible-sounding answer from training data. For example: "Search TRREB.ca and CREA.ca for the most recent average days-on-market for condos in Toronto. Cite the exact report and date." This reduces fabrication risk significantly.

When should a realtor use Perplexity instead of ChatGPT for research? Perplexity is built for sourced, cited research. Every answer surfaces its sources inline, which makes verification faster. For market research tasks where you need citations you can show a client or a colleague, Perplexity tends to be more reliable than ChatGPT's browsing mode. ChatGPT is better when you need to synthesize research into a formatted report, email, or presentation after the facts are gathered.

How do I verify what ChatGPT tells me about the market? Always go back to the primary source the model cites. If it cites a TRREB market report, open that report and find the specific number. If the number isn't there, the model may have hallucinated the source or the figure. For MLS-level data, your board's portal or a direct pull from your MLS system is the only reliable source.

What market research tasks is ChatGPT actually good at for realtors? Summarizing large documents like status certificates or strata reports, drafting market commentary once you've supplied the verified data, synthesizing news about a neighbourhood or development, and structuring raw data into client-facing formats. It's a drafting and synthesis tool, not a data source.

---

Emma Pace — strategic marketing consultant, AI coach for realtors, keynote speaker. Realtor at Monstera Real Estate. Builds AI-operated marketing systems at emmapace.ca.