Real Estate Stack · Chapter 06 of 10
Listing launch and marketing
From signed listing agreement to the first open house: the checklist, the photo rules that changed in 2026, and a listing description workflow that runs through Claude before it runs through the MLS.
A listing launch has two failure modes. The slow one: the photographer, the sign, the description, and the syndication all happen on different days because nobody owns the sequence. The expensive one: a virtually staged photo or a line of copy gets the file, the agent, or the brokerage a fair housing complaint or an MLS fine. Both are avoidable with a checklist and a habit of running every word of listing copy through the same check before it goes live. This chapter builds both.
The pre-list checklist
Run this the day the listing agreement is signed, not the day before photos. Assign an owner to every line so nothing waits on "someone."
| Item | Owner | Tool | Done when |
|---|---|---|---|
| Listing agreement and disclosures signed | Agent | SkySlope / DigiSign | Signed docs land in the file |
| Facts sheet built (beds, baths, sqft, lot, year, systems, upgrades, HOA, taxes) | Agent, from county record and seller | County assessor site plus seller input | Every field filled or marked unknown, no guesses |
| Declutter and prep walkthrough | Seller, agent advises | Punch list, phone photos | Rooms photo-ready |
| Photos and video scheduled | Agent | Photographer, or Restb.ai-equipped MLS tool company claim | Shoot date confirmed in writing |
| Virtual staging decision made, per room | Agent and seller | Virtual Staging AI, Collov, or BoxBrownie report | Staged rooms list is final before ordering |
| Listing description drafted and fair-housing-checked | Claude drafts, agent verifies | Claude, facts sheet, county record | Every sentence traces to a fact, wording check passed |
| MLS entry | Agent, human hands only | Flexmls / Matrix | Listing active, remarks match the verified draft |
| Sign, lockbox, flyer box ordered | Agent | Sign vendor, Canva Pro report | Installed at the property |
| Social post drafts and sphere email drafted | Claude drafts, agent sends | Claude, Canva, email tool | Human approved and sent |
| Open house scheduled and sign-in tool set up | Agent | Curb Hero or Spacio report | Event live on MLS and calendar |
Photos and media
The rule that matters most in 2026: if software or AI added, removed, or significantly changed what is in a photo, most MLSs now require the original unaltered image to sit next to the altered one, with a label. Brightness, contrast, and cropping are exempt everywhere researched. Adding furniture that is not there, removing a tenant's belongings, changing paint or flooring, or generating landscaping are not.
NAR's Code of Ethics Standard of Practice 12-5 requires disclosure whenever a photo is digitally altered in a way that materially changes the property's appearance, shown "clearly and conspicuously" in or next to the photo report. That is the floor. Your MLS's own rule is usually stricter and carries a dollar fine.
| MLS / rule | What it requires | Effective | Penalty |
|---|---|---|---|
| ARMLS Rule 8.23, Digitally Altered Media | "[Digitally Altered]" MLS watermark, paired with the original unaltered image, whenever software or AI adds, removes, or significantly changes image content verified | Effective May 28, 2026; education phase to Nov 2026, no fines during education | $200 per violation once enforcement starts, Dec 2026 verified |
| CRMLS Rule 11.5.2 (implements California AB 723) | Original, unaltered version must appear immediately before or after the altered image; caption must say "digitally enhanced," "digitally altered," or "virtually staged"; AI-generated landscaping is prohibited outright verified | Effective January 1, 2026 | $250 after an uncorrected warning verified |
| Bright MLS | "VIRTUALLY STAGED" in all caps, on the image or in the caption; edits that change permanent or structural elements are prohibited report | Standing rule | Media violations are $250 under rule 1.7; Bright's schedule lists no separate staging-label fine verified [14] |
| Texas MLS AI tagging (statewide) | AI-generated images tagged in the MLS and labeled in third-party syndication, including Zillow, Realtor.com, Homes.com report | Adopted February 2026 | Not found |
The pattern across every rule researched: original-plus-label. Order virtual staging with that in mind. Keep the unedited photo, order the staged version, and plan to publish both, captioned. If your MLS is not one of the four above, do not assume you are exempt: search "
None of the researched vendor pricing pages for virtual staging (BoxBrownie, Virtual Staging AI, Collov) document the disclosure requirement on their own ordering flow report. The disclosure is the agent's job, not the vendor's.
The listing description workflow with Claude
Claude drafts from a facts sheet you control, checks its own wording against a fair housing list, and then a human verifies every fact against the MLS input sheet and the county record before anything is pasted into the MLS. Claude never enters the MLS itself; a human types or pastes the final remarks.
The rule of thumb behind the wording check, stated plainly: describe the property, never who should live there. "Family room" describes a room. "Perfect for families" describes a preferred buyer, and that is the line the Fair Housing Act draws report. There is no current official HUD list of banned words; the widely circulated list traces to a 1989 HUD memo that was withdrawn and never formally replaced, but HUD has said it still uses that list informally in enforcement, so treat it as risk-reduction guidance, not statute text report.
Prompt structure to run in a Claude Project set up for this listing (see chapter 08 for the office file and project setup):
Facts sheet (paste from the MLS input sheet, no summarizing, exact numbers): [beds] [baths] [sqft] [lot size] [year built] [HOA $/mo] [taxes] [systems: roof year, HVAC year, water heater] [upgrades, with dates if known] [school district, factual name only] Task 1: Draft a 150 to 200 word MLS listing description using only the facts above. No claims not in this list. No adjectives about who would like living here. No comparisons to other listings. Task 2: Re-read your own draft and flag any phrase that describes a type of buyer or resident rather than the property (examples to catch: "perfect for," "ideal for," "walking distance to" any house of worship or school type, "quiet neighborhood," "safe area," any reference to a protected class, explicit or implied). Rewrite each flagged phrase as a property fact or cut it. Task 3: List every factual claim in the final draft as a short bullet, so a human can check each one against the county record and the facts sheet.
Before / after, the two phrases the research flagged by name:
| Before | After | Why |
|---|---|---|
| "Perfect for families, walking distance to churches" | "Fenced back yard, quarter-mile to the community park; multiple houses of worship within a mile" | Describes the yard and a factual distance; drops the preference language and the specific-faith reference report |
| "Quiet, safe neighborhood, great for kids" | "Cul-de-sac lot, sidewalk-lined street" | Cul-de-sac and sidewalks are property/street facts; "safe" and "great for kids" describe a preferred resident report |
Note on the "walking distance" phrase specifically: HUD guidance and industry practice treat factual proximity phrasing like "walk to the bus stop" as fine on its own; the risk is pairing distance with a protected-class landmark (a specific denomination's church, a specific ethnic business district). Swap "walking distance to churches" for a neutral phrase like "close to community amenities," or name the park, trail, or transit stop instead inference.
Claude workflow: listing description with a fair housing check
Trigger: the facts sheet is complete and photos are scheduled or back.
What Claude prepares: a 150 to 200 word MLS description built only from the facts sheet, a self-run wording check against the list above with flagged phrases rewritten, and a bullet list of every factual claim for verification.
What the human checks and does: verifies every bullet against the county record and the MLS input sheet, confirms no fact in the draft is missing from the facts sheet, and pastes the final remarks into the MLS by hand.
What is never automated: MLS entry itself, and any claim about square footage, lot size, or year built that has not been checked against the county record.
The launch sequence
- Human enters the listing into the MLS (Flexmls or Matrix), remarks pasted from the verified Claude draft.
- Syndication runs automatically from the MLS to IDX partners, Zillow, and Realtor.com; confirm it appears correctly within 24 hours.
- Claude drafts two or three social post variants (a photo carousel caption, a short video caption) built from the same verified facts, sized for Canva Pro templates report; human reviews, builds the graphic, and posts.
- Claude drafts the "just listed" email to the sphere, using the same facts sheet; human reviews and sends from the agent's own email, never an automated send.
- Claude drafts a short agent-to-agent note for co-op agents and any waiting buyer-agent contacts, flagging the open house date; human sends.
Every "Claude drafts" step in this sequence uses the same verified facts sheet as the MLS description, so the flyer, the social post, and the email never say something the MLS listing does not.
Open house kit
Sign-in: Curb Hero is a free, subscription-free digital sign-in app with QR touchless check-in and CRM export claim. Spacio (Lone Wolf) is the paid alternative, $25/mo for a solo agent, syncing sign-in leads straight to a Lone Wolf-family CRM report. Either way, every visitor's contact info becomes a follow-up task, not a business card in a drawer.
Claude workflow: open house follow-up drafts
Trigger: the open house ends and the sign-in list exports.
What Claude prepares: a short personalized follow-up draft per visitor (interested buyer, browsing neighbor, agent preview), referencing what each visitor asked about if noted, and a one-line seller summary of turnout.
What the human checks and does: reviews each draft for accuracy against what was actually said, adds any personal detail Claude would not know, and sends from the agent's own account.
What is never automated: the send itself, and any promise about price, terms, or the seller's flexibility that was not actually made in person.
Showing feedback loop
Showing feedback trickles in by text, ShowingTime-style request, or a phone call. Collect it in one place per property.
Claude workflow: showing feedback into a seller update
Trigger: the agent has three or more pieces of feedback since the last seller update, or a week has passed.
What Claude prepares: a short seller update that groups feedback into themes (price, condition, layout, competition) without naming which specific showing said what, plus a one-line recommendation if a pattern is clear (for example, three notes on the same room).
What the human checks and does: confirms the summary matches what was actually said, decides whether to raise a price or condition recommendation with the seller, and sends the update.
What is never automated: any pricing recommendation presented as a conclusion rather than a pattern for the seller to weigh, and the send itself.
Post-close
- Review request: send within a few days of closing while the experience is fresh, using a free tool like RateMyAgent, which ties reviews to actual verified transactions claim.
- Home-value updates: enroll the closed buyer in an automated monthly value and equity tracker like Homebot, agent plans running roughly $50/mo solo or lower when lender-cosponsored report; this is the ongoing touch that keeps the agent's name in front of a past client for years.
- Referral ask cadence: a short, direct ask at the one-year mark and again around the two-year mark, drafted by Claude from the closing facts (property address, close date, one detail about the transaction) and sent by the human. No referral fee terms are ever included in an unlicensed party's ask; any referral compensation between agents goes through a separate signed referral agreement.
What stays human
- Every fact in the listing description, verified against the county record and the MLS input sheet before it is entered.
- Deciding which rooms get virtual staging, and ordering the original-plus-label pairing your MLS requires.
- Typing or pasting into the MLS itself. Claude drafts, a human enters.
- Every send: social posts, sphere emails, agent-to-agent notes, open house follow-up, seller updates, review and referral asks.
- Any pricing or condition recommendation drawn from showing feedback, presented to the seller as a pattern to weigh, not a Claude-issued conclusion.
Do this today
- Look up your own MLS's digitally-altered-media or virtual-staging rule by name before you order any AI staging on your next listing.
- Build the facts sheet template (beds, baths, sqft, lot, year, systems, upgrades, HOA, taxes) as a Claude Project file so every future listing description starts from the same clean input.
- Set up Curb Hero or your brokerage's open house sign-in tool now, before your next open house, so visitor follow-up never depends on a paper sheet.
Sources
- NAR Code of Ethics, Standard of Practice 12-5, nar.realtor/about-nar/governing-documents/the-code-of-ethics, retrieved 2026-09-03
- ARMLS, Rule 8.23 Digitally Altered Media, armls.com/digitally-altered-media, retrieved 2026-09-03
- CRMLS, Rule 11.5.2 digitally altered image guidance, kb.crmls.org/knowledgebase/digitally-altered-image-guidance-faqs, retrieved 2026-09-03
- California AB 723 (Bus. & Prof. Code section 10140.8), legiscan.com/CA/text/AB723/id/3272851, retrieved 2026-09-03
- Bright MLS Rules, effective August 14, 2024, PDF via assets.ctfassets.net, retrieved 2026-09-03
- HUD, Fair Housing Act advertising guidance, hud.gov/program_offices/fair_housing_equal_opp/adcampaign, retrieved 2026-09-03
- HUD 1989 word-list withdrawal, summarized via secondary legal/industry sources, retrieved 2026-09-03
- Canva Pro pricing, via usecarly.com/blog/canva-pricing, retrieved 2026-09-03
- BoxBrownie and Virtual Staging AI pricing, getstageflow.com/blog/boxbrownie-review and aitoolsbakery.com/blog/virtual-staging-ai-cost, retrieved 2026-09-03
- Curb Hero, curbhe.ro, retrieved 2026-09-03
- Spacio (Lone Wolf), homespotter.com/spacio, retrieved 2026-09-03
- Homebot pricing, aiproductivity.ai/pricing/homebot, retrieved 2026-09-03
- RateMyAgent, ratemyagent.com, retrieved 2026-09-03
- Bright MLS, Policy on Rules Enforcement, Schedule 1 sanctions, assets.ctfassets.net PDF, retrieved 2026-09-04
Starting from zero? Use the field manuals first.
02 · BEFORE YOU BUY SOFTWARE
If you're still choosing the buyer, offer, stack, agents, content, customers, and delivery, don't start with a tool: start with the manuals. These create local worksheets only: no checkout, outreach, or account setup.
Find the first manual
Not sure which guide to open first? Start from your stuck point and let the finder recommend the next manual and the first worksheet question. No personal data is collected: it just points you at the right starting move.
Work the starter pack
Move through buyer, offer, stack, agents, content, customers, and delivery as a sequence of short local worksheet sessions. One artifact at a time, in order, so you build the business instead of browsing for it.
See what "done" looks like
Read a fully worked field manual before writing your own version, so the output feels concrete: a real, numbered workflow with the risky steps gated, not a fake proof claim.