QWE AI Academy Here’s the unpopular take: Mamdani’s rule isn’t a ban on AI in listings – it’s a ban on hiding it. The landlords worried about compliance costs are missing the obvious win. A properly disclosed AI-staged photo converts better than a bare, empty room. The people in trouble aren’t the ones using AI; they’re the ones pretending they aren’t.
If you rent apartments, list them, or hunt for them in NYC, this is a step-by-step checklist of what to actually do this week. No political recap.
What just dropped (the 90-second version)
NYC Mayor Zohran Mamdani released the Rental Ripoff Report – a 68-page document laying out proposals aimed at deceptive landlord practices (source: Emekce, July 2026). One clause targets landlords and realtors who alter listings with AI without disclosing it, and the city says it will coordinate with StreetEasy and Zillow. Disclosure only – no ban.
It’s the same direction California already went with AB 723. The honest state as of July 2026: enforcement mechanics aren’t defined yet. The mayor’s office hasn’t published fine amounts or an enforcement agency assignment. That gap matters – landlords who wait for the final rule risk a policy shift with no grace period.
The disclosure workflow that actually complies
Skip the debate about whether the rule is fair. Here’s the sequence that will keep you clear under Mamdani’s proposal, California’s AB 723, and existing NAR ethics – all at once.
- Shoot the room empty and unedited first. This is your baseline. Save the original file untouched.
- Stage or edit in a tool that embeds provenance metadata. Anything generated in Adobe Firefly writes a C2PA credential automatically. Same for OpenAI’s image tools.
- Add a visible caption on the image itself. “Virtually Staged” in a corner. Not tiny grey text. Actually visible.
- Post the original photo next to the altered one. Right beside it in the listing gallery – not buried at position 18.
- State it in the listing description. One sentence: “Photos 3 and 5 are AI-staged; furniture is not included.”
CRMLS spells this out under Rules 11.5(c) and 12.10 (as of July 2026): label the altered image, post the original right before or after it, don’t remove walls, fixtures, or unsightly features. If your workflow already handles that, you’re 90% of the way to any reasonable interpretation of the NYC rule. The NAR Code of Ethics Article 12, Standard of Practice 12-7, backs the same principle: photos must accurately depict the property.
Here’s the question nobody has a clean answer to yet: where exactly does the line sit between “photo editing” and “AI alteration” under a disclosure rule? Brightening a hallway is fine. Adding a virtual couch is clearly AI staging. Swapping a grey sky for golden hour – more on that below. The definitions aren’t locked in, which is why erring toward disclosure costs almost nothing.
Which tools carry a paper trail (and which don’t)
Turns out most major AI image generators now embed C2PA “Content Credentials” – a cryptographically signed manifest that travels inside the image file. Think of it as a receipt for what was done to a picture. C2PA is an open technical standard that lets publishers verify a file’s origin and edit history. OpenAI uses it alongside SynthID watermarks on every image from ChatGPT and DALL-E (source: OpenAI Help Center, as of July 2026). Adobe Firefly has embedded content credentials in every generated image since launch.
| Tool | Embeds C2PA? | Pricing note |
|---|---|---|
| Adobe Firefly | Yes, since launch | See adobe.com for current plans |
| ChatGPT / DALL-E | Yes, plus SynthID watermark | See openai.com for current plans |
| Virtual Staging AI / Collov / Apply Design | Varies – check per plan | $16-$79/month for volume plans (as of July 2026, per neuhausre.com) |
| Midjourney | No native C2PA at time of writing | Check midjourney.com for current pricing |
Verify what your tool actually writes by uploading the output to contentcredentials.org/verify. Nothing shows up? Your tool isn’t giving you a paper trail – compensate with heavier visible labeling.
Pro tip: Adobe’s Firefly manifest records “AI generated” as an action inside the file. If a tenant later claims deception, that manifest is your evidence you disclosed the AI origin at creation – at no extra cost. But only if the platform you upload to preserves it. Most don’t.
The pitfalls that’ll actually bite you
Zillow and StreetEasy strip your metadata. The catch: when an image with content credentials is shared on a platform that reprocesses files through compression, resizing, or format conversion, C2PA metadata is typically stripped away (TrueScreen, citing RAND Corporation, June 2025). RAND’s analysis, titled “Overpromising on Digital Provenance and Security,” concluded that end-to-end compliance across an open ecosystem is unrealistic. Your signed manifest dies the moment Zillow re-encodes the JPEG. Visible captions burned into the pixels are the only signal that survives the upload.
Twilight sky conversions count. Most agents don’t think of “changing the sky from grey to golden hour” as AI editing. Under a strict reading, it is – neuhausre.com notes (as of 2026) that AI twilight conversions fall under the same disclosure rules as virtual staging. Disclose it or reshoot at the real time.
Removing existing furniture is a trap. Virtual staging in an empty room is safe. Digitally erasing the current tenant’s furniture to make the space look bigger? That crosses into misrepresenting actual property condition – the exact harm both NAR guidelines and Mamdani’s proposal target.
How tenants can actually verify a listing
Three tools. None of them perfect.
- Right-click, save, upload to a detector. Hive Moderation hits 94% accuracy across Midjourney v6, DALL-E 3, and Stable Diffusion XL (ddiy.co independent benchmark, as of 2026). Illuminarty offers 5 free scans per day with heatmaps showing which regions triggered detection. That heatmap is the most useful part – it points at the suspicious pixels rather than just returning a score.
- Check for C2PA credentials. Drop the image into contentcredentials.org/verify. If it was made in Firefly or OpenAI and hasn’t been re-uploaded three times, the manifest may still be there.
- Reverse image search. Same “apartment” photo on five listings? Stock image or generated asset.
94% accuracy sounds high. But that 6% false-positive rate is real – professionally retouched real estate photos can trigger AI detectors without any AI involvement. What these tools give you isn’t a verdict. It’s a reason to ask harder questions before you sign anything.
When to skip AI staging entirely
The unit is already furnished and photogenic. AI staging exists to give prospective tenants a sense of scale in empty rooms. If the room isn’t empty, a wide-angle lens and good lighting do the job without any disclosure overhead.
You’re listing a rent-stabilized apartment with a known waiting list. The audience isn’t browsing for a mood board – they want proof the apartment exists and is livable. Staged photos add friction without adding conversions.
Your images will bounce through five platforms and a WhatsApp group. By the time your listing ends up on a broker’s Instagram story, every provenance signal is gone and the visible caption is probably cropped. If you can’t control the distribution chain, invest in real photos.
FAQ
Is this a law yet, or just a proposal?
A proposal. Specifically, a mayoral report – not a passed ordinance. No enforcement agency assigned, no fine amounts published as of July 2026.
Do I need to disclose if I only used AI to brighten a photo or fix white balance?
Probably not – traditional exposure and white balance corrections have been standard in real estate photography for two decades and don’t change what the property is. The line, based on how California’s law and CRMLS Rule 11.5 frame it: does the edit change the property itself? Adding a virtual couch to an empty room → disclose. Brightening a dim hallway → fine. Removing a carpet stain → that’s where it gets murky. You’re now hiding a defect, which is the exact harm both Mamdani’s proposal and California’s AB 723 target. When in doubt, label it. A caption costs nothing; a tenant who discovers a hidden stain on move-in day and posts the before/after on TikTok costs a lot more.
What’s the fastest way to comply without redesigning my whole listing process?
Two steps: add “Virtually Staged” as a text overlay in the corner of every altered photo, and add one line at the top of your listing description saying which photos are staged. Five minutes per listing. That covers almost every existing rule and the proposed NYC one.
Your move today: pull up your three most recent listings. Open every image. Ask one question: did I decorate this room, or did I change this apartment? Anything in the second bucket needs a label before Monday.