Tuesdays are 3 AI updates, one hot take, one action. Weekends are 1 practical real estate agent workflow.
Two years ago the hard part was getting agents to try any of this. I spent most of 2024 in rooms asking questions where the hands stayed down. New surveys of more than 100 brokerage leaders this month found the share of brokerages not implementing AI daily has fallen to less that 4%. Everyone in your office has the same subscriptions open now, so the interesting question moved from whether you use it to what it does on your behalf. Here are this week’s insights…
1. Ask AI who the best agent in your city is, and there's a 91.5% chance the actual top producer never comes up. The SEO agency Local Falcon, which tracks how brands surface inside AI answer engines, published the The Realtor AI Visibility Index which ran 37,500 searches across the 100 largest US cities on five platforms. It checked every agent named against RealTrends Verified's 2026 production rankings. Of established agents who publish a website, 91.5% were never named once, and only 21.2% of the highest-volume agents in those cities got cited anywhere. What the models read is not production data: Google Business Profile alone drives 36.6% of citations, Zillow another 15.9%, and eighteen of the twenty-five most-cited agent websites carry some version of a "best agents in [city]" post. The platforms barely agree with each other either, with 82.6% of named agents cited by exactly one platform and no others. It’s worth a read to continually refine your personal AEO strategy.
2. Claude started marking what it writes, and for now nobody can read the mark. Anthropic announced on August 14 that Claude embeds an invisible statistical watermark in the text it generates, worldwide, ahead of a European transparency law that took effect August 2. Google does the same on Gemini output through SynthID, and OpenAI signed the same European code without shipping text watermarking yet. The practical impact today is close to zero, because there is no public detector. Anthropic says a detection API is coming and has not shipped it, so any site claiming to check your text for Claude watermarks right now is guessing. Worth filing away rather than worrying about it but wanted you to be aware.
3. ChatGPT & Claude can read and answer your text messages On August 20 OpenAI released an Apple Messages plugin for the ChatGPT desktop app that searches your threads, summarizes a conversation and drafts or sends replies through Messages. MacRumors reports it lives inside ChatGPT Work and Codex rather than an ordinary chat, so plenty of people will go looking for it and never find it. Claude has had a version of this for a while through Anthropic's own iMessage plugin, which reads your message database locally, sends through AppleScript, keeps everything off external servers and lets you allowlist exactly which contacts can reach it. That's the better privacy posture of the two. This is one of the most requested things agents ask me for so if you didn’t know it existed before, this is one to pay attention to.
HOT TAKE
The most valuable thing you could connect AI to is your text messages, and almost nobody has done it. Your sphere lives in Messages. Every referral you ever got started as a text you answered fast, and every one you lost went cold in a thread you meant to return to. An assistant that can read two years of that and tell you who you dropped is worth more than every listing-copy tool you have ever bought put together. Both companies let you do this now, so spend the hour.
1 THING TO TRY
Install one of the two message plugins and ask it a single read-only question before you let it write a word. On the Claude side that's Anthropic's iMessage plugin at claude.com/plugins/imessage, on the ChatGPT side it's the Apple Messages plugin in the Plugins interface, and either way you'll grant Full Disk Access and it will only run on an Apple silicon Mac. Then ask exactly this:
"Search my messages from the last 90 days. List every conversation where the other person sent the last message, or asked me something I never answered. Give me the name, the date and the question. Don't draft anything."
Give this a try and Friday I'll go deeper into how you might want to use it.
BOSTON! I’M COMING FOR ONE DAY IN-PERSON.
Thursday September 24 I'm running a one-day Claude workshop in Boston. We build the system on your own listings and your own database, and we do the message setup from story three in the room together. Small room on purpose, and it's $499. Grab your seat here.
Reply with one way AI has changed your day-to-day and I'll send you a code for 10% off.
-Matt
A final note from Wispr Flow b/c its a tool I use every day…
PRDs by voice. Bug reports by voice. Ship faster.
Dictate acceptance criteria and reproductions inside Cursor or Warp. Wispr Flow auto-tags file names, preserves syntax, and gives you paste-ready text in seconds. 4x faster than typing.
