This week's signal is the handoff — AI moved deeper into the listing-marketing stack: editing the photos, plugging into the CRM as an authorized agent, prospecting sellers in plain English. But everything the machines can't sign for got sharper: more than nine in ten AI-edited listing photos carry no disclosure as California enforces labels (and even regulators can't always tell real from fake); Compass is prepared to sue any MLS that punishes publicly marketed office exclusives, with an October 6 deadline; and with one in five sellers cutting prices, the relationships that close deals are worth more, not less. The workflow is getting automated. The accountability isn't.
01
Make it shine
Tools that make every listing look irresistible.
2 stories
UpdateVideo & Visual
Nine in ten AI-edited listing photos go undisclosed
Inman's September 30 Tech Roundup spotlighted research from Coraly.ai: of nearly 40,000 primary listing photos from Zillow, Redfin, Realtor.com and Homes.com in the first quarter of 2026, about 11% of listings used AI-edited primary photos — and more than 90% of the altered images carried no disclosure that AI was used. Rates ranged from 8.7% on Realtor.com to 12.4% on Homes.com (Zillow 11%, Redfin 11.2%), and sky replacement was the most common edit, appearing in 69% of altered images. The findings land as California's AI-photo disclosure law — in effect since January — meets live enforcement questions: New York issued a consumer alert last November, and New York City proposed rental-listing disclosure rules in July. Note: Coraly.ai is itself an AI proptech vendor, so its methodology is company-sourced, not peer-reviewed.
Why it matters
If your stack includes AI editing or staging, disclosure is now a compliance line item, not a nicety. Inman's takeaway: a clear “digitally altered” label is cheap insurance against both regulatory trouble and buyer backlash over “housefished” listings — and agents who can explain to sellers what was altered, and what wasn't, turn a risk into a trust advantage.
Australia's AI-photo crackdown misfires on a real photo
New South Wales, Australia recently introduced rules requiring landlords and agents to disclose digitally altered or AI-generated rental-listing photos — with fines up to $15,321 for non-disclosure. But the crackdown misfired immediately: in a media release last week, the state government cited an “AI-doctored rental listing in Sydney's inner west” that turned out to be genuine — the seemingly impossible Bondi Beach view through a Leichhardt apartment's windows was a $5,000 courtyard mural commissioned by the landlord, not a digital edit. The listing agency, IB Property, confirmed the image “has not been altered or digitally generated using AI, Photoshop or any other image-editing software.” The minister's office did not directly address the mistake.
Why it matters
The cautionary half of the disclosure story: enforcement is going to be messy, because even regulators can't reliably tell AI edits from real photos. For US agents, the practical read is the same as California's — label what you alter, and keep the receipts (before/after files) for anything a buyer might question.
Tools that put every listing in front of the right eyes.
1 story
NewMarketing Stack
CRMLS defies Compass's litigation threat over office exclusives
On September 30, California Regional MLS publicly refused a September 8 demand from Compass International Holdings to stop fining agents who publicly market office-exclusive listings. Compass's letter — reviewed by Real Estate News — set an October 6 deadline for written confirmation or face an antitrust lawsuit, calling a case against CRMLS “straightforward and compelling.” CRMLS countered that its Rule 7.9 already lets sellers market fully off-MLS, accused Compass of wanting “the ability to benefit from the cooperative sharing of information while publicly marketing for-sale properties to benefit themselves,” and said it will establish an “MLS Cooperation Legal Defense Fund.” At a CMLS event the same day, CEO Robert Reffkin said Compass is “prepared to sue any MLS that fines or punishes agents with publicly marketed office exclusives.” Compass previously sued Northwest MLS over listing policies in 2025; that fight ended this year with NWMLS changing its pre-marketing rules. (CRMLS's letter also alleges Reffkin offered $1,000 incentives to offices that leave CRMLS — an allegation from one side of the dispute, not an established fact.)
Why it matters
The distribution fight is moving from portal-vs-portal into the MLS rulebook itself — and the October 6 deadline lands next week, so agents in California should watch whether their MLS's marketing rules change. Whatever side you're on, the practical question is the same one from the Zillow–MRED arbitration: who gets to set the rules for where a listing can appear.
Turn attention into relationships, and relationships into commissions.
2 stories
NewMarketing Stack
Lofty opens its CRM to outside AI agents via MCP
On September 28, Lofty — the Phoenix-based brokerage platform formerly known as Chime — launched the Lofty MCP Server, a connector built on Anthropic's open Model Context Protocol that lets brokerages plug their own AI agents and internal systems into the Lofty platform with no custom integration code. MCP-compatible AI clients can read Lofty data, analyze activity, and, where authorized, take action inside the platform; it works alongside Lofty AOS, the company's own agentic AI system. “AI is only as valuable as the business context it can access,” CEO Joe Chen said in a statement. Lofty is not alone: Rechat and Inside Real Estate have also rolled out MCP servers in 2026, following data providers ATTOM (January), PriceHubble, and Local Logic. (Disclosure: Inman named Lofty its official AI partner in August 2026.)
Why it matters
The CRM is becoming an AI workbench: instead of toggling between tools, agents may soon query and update their CRM from whatever AI assistant they already use. The honest caveats: how much access reaches individual agents depends on what their brokerage chooses to connect and authorize — and “securely accessible” is the company's claim, not an independent audit.
PropertyRadar's Radar AI finds sellers in plain English
On September 29, PropertyRadar expanded its platform with predictive seller signals: 0–100 “Likely to List for Sale” scores for each homeowner, filterable alongside 250+ criteria like equity and ownership length; Radar AI, a natural-language list builder (“homeowners with more than 50% equity, owned 10+ years, showing signs of distress”); expanded FSBO lifecycle data; 30+ distress and life-event signals including probate, divorce, and pre-foreclosure; and “Plays,” ready-to-run direct-mail and email campaigns. The pitch is ending reliance on “renting your pipeline from a portal” — CEO Mark Hockridge: “Find the seller before they become a lead, and you're not one of four calls — you're the one they already trust.” The expansion is included in every plan at no extra cost; entry is $99/month prepaid annually, with a free trial.
Why it matters
In a market with fewer deals changing hands, the highest-leverage AI use may not be writing listing descriptions — it's finding the seller before the listing exists. The caveats: predictive-accuracy claims are vendor statements, and this is a company press release with no independent coverage found.
Redfin: one in five sellers cut prices — strongest buyer's market on record
Redfin's September 30 data release: 21.1% of U.S. home sellers with active listings cut their asking price in the four weeks ending September 20 — the highest share for this time of year in Redfin's records (back to 2022), up from 19.8% a year earlier. Denver led at 30.9%, followed by Indianapolis (29.9%), San Antonio (26.8%), Dallas (26.6%), and Austin (26.1%). Nearly half of U.S. homebuyers got concessions from sellers — repairs, closing costs, or rate buydowns. San Francisco sat at the other extreme: under 10% of sellers cut prices, in what Redfin calls an AI-wealth-driven seller's market. Caveat: “strongest buyer's market on record” is Redfin's framing of its own MLS analysis, and the release plugs Redfin Early Access.
21.1%
of sellers cut prices in the four weeks ending Sept 20 — the highest for this time of year in Redfin's records
30.9%
of Denver sellers cut prices, the nation's highest; three Texas metros followed at 26.1–26.8%
~1 in 2
U.S. homebuyers got concessions from sellers — repairs, closing costs, or rate buydowns
<10%
of San Francisco sellers cut prices — the smallest share nationwide, in an AI-wealth-driven seller's market
“The marketing is getting automated. The accountability isn't.”
Count the handoffs: AI edits the listing photos, an MCP server lets an outside AI agent work your CRM, and plain-English queries find tomorrow's sellers. None of that is the agent's job anymore — but disclosure is (nine in ten AI-edited photos carry no label, and California's law is already in force), distribution rights are (Compass's October 6 deadline puts the MLS rulebook on trial), and the relationship is (in a market where one in five sellers cuts prices, the listing goes to whoever earned the trust first). Every tool this week made the marketing faster; none of them signed the paperwork.
This week, audit your own handoffs: label anything AI-touched in your listings and keep the before/after files; ask your brokerage what an external AI agent would be allowed to do inside your CRM before you connect one; and if you're farming sellers, price the conversation around reality — in this market, the agent who prices honestly from day one beats the one who chases the market down.
Editorial disclosure: NanaStack is independent. Nothing in this issue is sponsored. Vendor claims (Coraly.ai's research methodology, Lofty's security and capability claims, PropertyRadar's predictive-accuracy claims, Redfin's market framing) are reported from the cited sources and marked where not independently tested. Note: Inman disclosed naming Lofty its official AI partner in August 2026. This issue's research window was September 28–October 1, 2026; the Coraly.ai findings fall just outside it and are dated in the text.
