The Realtor Hive Evergreen ยท AI Prompt Library
the whole shelf, whenever you want it

The Hive AI Prompt Library

Everything we'll teach across the year, in one place. This is a reference, not an assignment โ€” each month we take one piece of it and work it properly. Browse when you're curious, and ignore it when you're not.

Consistency Follow-up Draft, don't publish Fair housing safe 13 prompts Reference, not homework
The rule, always

AI does the part that breaks your consistency. Never the part that earns the client.

Every source worth reading agrees on where AI is weak, and it is a uncomfortably specific list: negotiation, local judgment, and relationship work. Which is more or less the entire job.

What it's genuinely good at is the repetitive, first-draft, data-heavy work sitting between you and the relationship โ€” the twenty-five messages, the CRM notes, the recap email, the caption you've rewritten four times. That work is why follow-up dies in week three. Hand over that, and only that.

And the warning worth taping to your monitor: don't let building flyers eat into your prospecting time. AI makes marketing so satisfying to produce that it becomes a very sophisticated way to avoid picking up the phone. This month is about conversations. AI is here to buy you time for them โ€” not to replace them.

The line

What to hand over, and what to never

Give it this

  • First drafts. The follow-up text, the recap email, the listing description, the caption. You edit; it types.
  • Personalizing at volume. One message, twenty-five versions, each with the detail only you know.
  • Turning mess into structure. Voice-note your showing notes on the drive home, have AI turn them into CRM entries.
  • Summarizing. Inspection reports, HOA docs, market data โ€” into plain English you can send a client.
  • Reformatting. One long post into a carousel, a caption into a reel script, an email into a text.
  • Getting you unstuck. Ten hook options at 6am when you have none.

Never this

  • The relationship. Do not automate a reply to someone you know by name. People can tell, and it costs more than the time it saved.
  • Negotiation. Strategy, reading the other agent, knowing when to hold โ€” none of that is in the training data of your market.
  • Local judgment. AI does not know which street floods, which builder cut corners, or what the neighbors are actually like.
  • Numbers you'll be held to. Prices, rates, days on market, closing costs. Verify every figure before it leaves your hands.
  • Anything you haven't read. If you didn't read it, you didn't write it, and you're still the one on the license.
  • Fair housing language. See below. This is the one that can actually cost you.
Do this before anything else

Build your brand system first

This is the step everyone skips, and skipping it is why people try AI design tools, get generic output, and go back to Canva. Set it up once and every single thing you make afterward comes out on-brand.

The Starbucks question

What if everything you ever made was designed by the same brain? That's the whole idea. Most agents' marketing is a hodgepodge โ€” a Canva flyer here, an Etsy template there, a postcard from three brokerages ago. Nothing matches, because a different brain made each piece.

A brand system is you handing over the brain once. Your fonts, your logo, your headshot, your colors, your compliance block โ€” loaded in a single time, then applied to everything without you re-uploading anything ever again.

What to load in

  • Your logo and any variations you have.
  • Your headshot โ€” the current one, not the 2019 one.
  • Your fonts, if you have brand fonts.
  • Your compliance graphics โ€” brokerage logo, Equal Housing, license line.
  • Only the templates you're proud of. If you don't love it, don't teach it.
  • A written description of your business โ€” market, niche, who you serve, how you sound.

Skip the developer fields โ€” GitHub, Figma, code. They're not for you.

What comes back

  • A full color and type system pulled from what you gave it.
  • A brand motif derived from your logo, which then shows up across everything.
  • Logo variations, badges, buttons, quote cards.
  • Sample pages and templates so you can check it understood you.
  • Compliance blocks already formatted.

Give it about twenty minutes to generate. Then review it properly โ€” colors, spacing, type โ€” and give feedback until it's actually you. You can keep adding to it forever.

If the written part is where you stall

The brand description field is where most people stall. Don't write it from scratch โ€” ask your AI for the most thorough description of your business it can produce based on your website and your posts, read it, fix what's wrong, and paste that in.

And if you don't have a brand at all โ€” most agents don't. You don't need to go hire someone first. Start with your brokerage's requirements, since many mandate the logo, colors, and compliance language anyway. Then pull the rest from whatever you've already made that you actually liked: a flyer that worked, the colors in your headshot backdrop, the way you already write. Load that in and let it propose a system from there. You can refine it forever โ€” you cannot refine nothing.

Prompt ยท Write your brand description

Get the words for the brand field

Act as a brand strategist. I'm a real estate agent and I need a thorough written description of my business to load into a design tool. Here's what I've got: I serve [City] and [neighborhoods/areas]. I've been doing this [X] years. My clients are mostly [first-time buyers / move-up / luxury / relocation / sellers]. My brokerage is [brokerage]. Here are a few things I've written so you can hear how I talk: [paste 2โ€“3 captions or emails]. Write me: who I serve, what makes working with me different, the tone I write in, words I use and words I'd never use, and three adjectives for how my marketing should feel. Keep it factual. Don't invent awards, numbers, or specialties I didn't give you.
That last line matters. Left alone, AI will happily describe you as an award-winning luxury specialist. Read every word before you paste it in.
Prompt ยท Refine what it built

Tell it what it got wrong

Here's the brand system you generated. A few things to change: The [color] is too [bright/dull/cool] โ€” my brand is warmer than that. Closer to [reference]. The headline font feels [too corporate / too casual] for how I actually talk. [Anything else that's off.] Regenerate with those changes. Then show me a sample listing flyer and a sample social post so I can check it before I start using it.
Ask for samples every time you change something. It's faster to spot a problem on a mock flyer than to discover it three assets later.
The big win

One prompt, a whole listing launch

Once your system exists, this is what it buys you. Not one image at a time โ€” the entire package, from one plain-English request.

Upload ten photos and ask for the pack

Add the basics from your MLS sheet or tax record โ€” address, beds, baths, square footage, lot size. The exact prompt is directly below this card โ€” fill in the brackets and paste it. What comes back, all on-brand:

Listing flyer Open house flyer Single property page QR code Social square Carousel Instagram stories Just-listed email Sign-in sheet Property tour deck Door hanger Yard sign rider Postcard

Uploading the photos beats connecting your cloud drive โ€” it's faster and it gets the right images. Ten is plenty.

Then make it repeatable. Save that request as a saved instruction โ€” "every listing launch, make these seven assets" โ€” and from then on you type launch my new listing and go do something else.

Prompt ยท The full listing pack

One request, the whole launch

Using my brand system, create a full listing launch package for this property. [Upload up to 10 photos] Address: [address] Beds/baths: [X/X] ยท Square footage: [X] ยท Lot: [X] Price: [$X] The three things that actually sell this house: [what you'd tell a friend] Open house: [day and time, or "none yet"] Make me: a listing flyer, an open house flyer, a single property page with a QR code, an Instagram carousel, three story frames, a social square, and a just-listed email. Keep all copy fair housing safe โ€” describe the property, never who should live there. Include my compliance block on every printed piece.
The fair housing line in the prompt is doing real work. It's much easier to prevent bad copy than to catch it later across nine assets.
Prompt ยท Make it repeatable

Save the workflow so you never type it again

Save this as a standing instruction called "Listing Launch." Whenever I say "launch my new listing" and give you photos and property details, produce every asset from the request above, in my brand system, with fair housing safe copy and my compliance block. Ask me for anything you're missing before you start, rather than guessing.
The last line prevents the most annoying failure โ€” it inventing a price or an open house time because you forgot to include one.

Finishing touches

If you mostly like it but want to nudge it, send it to Canva โ€” you get an editable file with fully separated layers. Best of both: AI does the layout, you do the last 10%.

For quick-cut video editing, Descript or similar tools are better and cheaper than making Claude do it.

Other things worth trying

  • CMAs from an MLS export. Pull a few hundred comps over several years into a CSV, upload it, and get trends, average prices, and days on market for your listing appointment.
  • Your inbox into content. Have it read two weeks of builder emails, find the best incentive, and turn it into a carousel.
  • A real website. Agents with zero design skill are building their own. What most need is a digital business card, a blog, and a home search link.
Prompt ยท CMA from your MLS export

Turn a CSV into a listing presentation

Attached is a CSV of comparable sales in [area], covering [X] years. Analyze it and tell me: average and median sale price by year, average days on market by year, list-to-sale price ratio, what's selling fastest, and any trend a seller should know about. Then build me a listing presentation section that explains this to a homeowner in plain English โ€” no jargon, no charts they'd need me to interpret. Three or four takeaways, not twenty. Flag anything in the data that looks wrong or incomplete before you analyze it.
Don't get too analytical, even for an analytical seller. Three clear takeaways beat a twenty-page data dump every time.
Prompt ยท Your inbox into content

Mine the emails you already get

Here are the last two weeks of builder and lender emails from my inbox. [paste, with personal details removed] Find the two or three things a buyer in [City] would actually care about โ€” incentives, rate buydowns, new phases, price improvements. Turn the best one into an Instagram carousel: a hook slide, three or four detail slides, and a final slide with a clear CTA. Keep my voice. Don't overstate any offer โ€” quote the terms exactly as written and flag anything I need to verify before posting.
Strip names and addresses before pasting. And verify every incentive yourself โ€” you're the one advertising it.
Read this part twice

AI writes fair housing violations by default

Not occasionally. By default โ€” because the phrases that create liability are the same ones that sound like good marketing copy, and that's what it was trained on.

Phrases AI hands you unprompted

Every one of these has appeared in AI-generated listing copy, and every one describes who should live there rather than what the property is:

Perfect for young professionals Ideal for families Great for retirees Perfect for joggers Walking distance to church Quiet neighborhood Safe area Exclusive

"Quiet" and "safe" feel harmless. They are among the most commonly flagged terms in fair housing enforcement, because of what they've historically been used to signal. The test is simple: does the sentence describe the house, or the person you're picturing in it?

Connectors and skills: where the actual risk is

As you connect AI to your email, CRM, and social accounts, the exposure changes. Two rules worth taking seriously:

  • Only install connectors from the official library or a company you already trust. Third-party connectors and skills from random gurus can do things you didn't agree to. If you can't name the company, don't connect it.
  • Know the difference between read and write. A read tool looks at your data and reports back โ€” low risk. A write tool can send, post, delete, or change things. Set write tools to require your permission every time, so nothing happens without you saying yes.

And keep client information out of your prompts regardless โ€” names, addresses, financial details, contract terms. Anonymize first. You are the one holding the license and the fiduciary duty.

The seven-point check, before anything gets published

  • Feed it facts only. Square footage, features, location data. Not "write a listing for a family home."
  • Describe the property, not the buyer. Delete every sentence naming a type of person.
  • Strip coded terms. Quiet, safe, exclusive, walkable-to-church, up-and-coming.
  • Fact-check every claim. AI invents square footage, school ratings, and year built with total confidence.
  • Keep client information out of prompts. No names, addresses, finances, or anything you wouldn't email unencrypted.
  • Disclose altered images. Virtual staging and AI-enhanced photos get labeled โ€” and the rules now vary by state, so check yours.
  • A human approves it. That human is you, and you're the one on the license.

Replace buyer language with verifiable features: natural light, updated flooring, first-floor primary, transit access, lot size, storage, views. Describe what's there. Let people decide if it's for them.

Prompt ยท Run this before you publish anything

The fair housing audit

Review this copy for fair housing compliance. [paste your listing description, caption, or ad] Flag any language that: โ€” describes who should live there rather than the property โ€” references or implies race, religion, national origin, family status, disability, age, or sex โ€” uses coded terms like quiet, safe, exclusive, walkable to church, up-and-coming, or family-friendly โ€” makes a claim about schools, crime, or neighborhood demographics For each flag, tell me why it's a problem and give me a replacement that describes a verifiable feature instead. Then tell me plainly whether you'd publish this as written.
This is the single most valuable prompt on this page. It takes fifteen seconds and it catches the thing that can actually cost you. Run it on everything client-facing โ€” including copy you wrote yourself.
This month

Where AI fits in your five weeks

Mapped to the Bee-Line, because the point isn't to add AI to your plate โ€” it's to make the plate you already have survive to week five.

Week 1
Name the Names

Cleaning up the database

Paste a messy export โ€” business cards, event lists, phone contacts โ€” and have AI dedupe it, standardize the formatting, and sort into your Now / Later / Cheerleader buckets based on the notes you give it. The job that eats a Saturday becomes twenty minutes. Strip personal details before pasting anything into a public tool.

Week 2
Start the Conversation

Twenty-five personal messages without twenty-five blank pages

This is the week AI earns its keep. Give it your voice sample and the one specific detail you remember about each person, and have it draft twenty-five openers. You edit every one โ€” but you're editing, not staring. Never send one you haven't changed.

Week 3
Ask Better, Listen Harder

Turning conversations into records

Voice-note what you learned on the drive home. Have AI turn the rambling into structured CRM entries: timeline, motivation, next action, follow-up date. The reason nobody logs their conversations is that typing it up at 9pm is unbearable. Talking for ninety seconds isn't.

Week 4
Make the Ask

Prep, not persuasion

Use it to prepare โ€” summarize comps into plain English, draft the pre-listing walkthrough agenda, build the one-pager. Do not use it to write the ask itself. The ask is a relationship moment and it should sound like you at your kitchen table, not like a proposal.

Week 5
Score the Month

The evaluation nobody does

Paste your month โ€” touches sent, replies, appointments, what you tried โ€” and ask it to find the pattern. Which opener got answered. Which day of the week worked. What to stop doing. This is the week most agents skip, and it's the one AI makes genuinely easy.

Steal these

The prompt library

Copy these. The difference between useful output and obvious slop is almost always specificity โ€” give it your voice, your market, and a constraint. Every prompt on this page is written to be pasted as-is once you fill in the brackets.

Prompt one ยท Do this first

Teach it your voice

Here are five things I've written โ€” captions, texts to clients, an email. [paste them] Analyze how I write. Sentence length, punctuation, how formal I am, words I use and words I'd never use, how I open and close. Write it as a style guide I can paste at the start of future chats. Then write one sample text to a past client checking in, using that guide.
Do this once, save the output, and paste it into every future prompt. It's the single highest-leverage ten minutes in this whole page โ€” it's the difference between AI sounding like you and sounding like AI.
Prompt two ยท Week 2

Twenty-five openers, not one template

[paste your style guide] I'm reaching out to past clients and people I met recently. Below is a list of names with one detail I remember about each. [paste list] For each person, write a short text โ€” under 40 words โ€” that references their specific detail and asks one easy question. No real estate pitch. No "just checking in." Each one should sound like I actually thought about them, because I did.
The constraint matters. "Under 40 words," "no just checking in," and "one easy question" are what stop it producing five paragraphs of nothing.
Prompt three ยท Week 3

Voice notes into CRM entries

Here's a voice memo transcript from after a client conversation. [paste] Pull out: their timeline (month and year if stated), what's driving the move, price range, what they said they want, what they seem worried about, and the single next action I owe them with a date. Format it so I can paste it straight into my CRM notes. If something wasn't mentioned, say "not discussed" rather than guessing.
That last line matters more than it looks. Without it, AI fills gaps with plausible invention โ€” and invented client details are worse than no notes.
Prompt four ยท Content

Make the Hive caption yours

[paste your style guide] Here's a caption template with brackets to fill in. [paste caption] My city is [City], my neighborhoods are [list], and the local details worth using are [your spots, your market facts]. Fill in every bracket with real specifics. Keep the CTA on the first line. Then flag anything that still sounds generic and give me a more specific alternative.
The last instruction is the useful one โ€” asking it to critique its own output catches the lines that read like a template.
Prompt five ยท Week 5

Score your own month

Here's my September: [touches sent, replies received, appointments set, what worked, what got silence]. Find the pattern. Which approach got the most replies, which got ignored, and what would you change for October? Be specific and be honest โ€” if something isn't working, say so plainly. Then give me my October Now list with the next action for each name.
"Be honest" and "say so plainly" earn their place. Left alone, AI is relentlessly encouraging, which is useless in an evaluation.
Prompt ยท Weekly content batch

One theme, every asset

This week's theme is [topic]. My market is [City] and my audience is mostly [buyers/sellers/both]. Using my brand system, create: one Instagram carousel, two single social posts, three story frames, and a short script for a talking-head reel โ€” all on that theme, all in my voice, all different enough that they don't feel repetitive. Put the CTA on the first line of every caption. Two hashtags maximum. Then tell me which one you'd expect to perform best and why.
Batching one theme is how people who post constantly actually do it. The last question is worth asking โ€” its reasoning tells you whether it understood the audience.
When you're ready for more

The next level, honestly described

Two bigger moves worth knowing exist. Neither is a September project โ€” but knowing they're possible changes what you plan for.

Turn your workflow into a saved instruction

Once you've done something three times โ€” a listing launch, a weekly content batch, a new-client welcome โ€” write it down as a standing instruction your AI keeps. Then you stop re-explaining and start saying "do the thing."

This is the realistic upgrade for most agents and it takes an afternoon.

Building your own dashboard

You pay for a CRM, an IDX, e-sign, a follow-up tool โ€” and none of them talk to each other, so you are the integration. Some agents are now building their own dashboard that pulls it together and answers the weekly questions that matter: where am I against goal, which new leads haven't I reached, which past clients are due.

The honest version: this is an iterative build across multiple sessions, not one afternoon. Tools run around $20 a month. Worth it if your stack is genuinely a mess โ€” skip it if you're just curious.

The failure modes

How agents get this wrong

01

Publishing the first draft

Unedited AI output is recognizable, and audiences reject it. Instagram now down-ranks accounts that don't post original content. The draft is the starting line.

02

Automating every reply

Auto-responses are fine for a generic inquiry. Use one on somebody you've met and you've taught them you're not really there. Reserve it for cold, first-touch volume only.

03

Asking it for strategy

A general AI doesn't know your market, your inventory, or your competition. It will produce confident, generic advice. Use it to execute a strategy you decided.

04

Trusting the numbers

AI states wrong square footage, invented school ratings, and outdated rates with total confidence. Every figure gets verified before a client sees it.

05

Pasting private information

Client names, addresses, financial details, contract terms. Assume anything you type into a public tool could be seen. Anonymize before you paste.

06

Using it to avoid the hard part

If AI is writing your ask, your apology, or your condolence, it's doing the part that was supposed to be you. Efficiency on the admin. Never on the relationship.

07

Redesigning instead of prospecting

Making a fourth version of a flyer feels like work and isn't. If your marketing has never looked better and your pipeline is empty, the tool is being used against you.

08

Quitting on day one

The first output is rarely right, and people abandon it there. You wouldn't fire an assistant over a first-day mistake. Give it feedback and go again โ€” that's the whole skill.

You already have one of these

Hive AI Caption Pro knows The Realtor Hive voice and what actually converts. Paste a caption from this month's bank, tell it your city and niche, and it finishes it sounding like a person wrote it. Start there before you start from scratch.

Open Hive AI Caption Pro โ†’
Research notes
  • HousingWire โ€” the fair housing checklist for AI-generated listing copy, and the specific high-risk phrases AI produces.
  • AI tool guides organized by agent workflow โ€” where AI helps by stage, and the finding that it performs poorly at negotiation, local judgment, and relationship work.
  • Coffee & Contracts โ€” AI workflows for Instagram, and the warnings on over-reliance, canned replies, and using general AI for strategy.
  • Instagram / Adam Mosseri โ€” original content as a ranking signal, and why unedited AI output costs reach.
  • AI Marketing Academy (Pantana & Ferry) โ€” the design-system-first approach, the one-prompt listing asset pack, connector safety, and the warning about marketing time eating prospecting time.
  • AI photo and virtual staging disclosure requirements now vary by state. Check your state and your MLS rules โ€” this page can't do that for you.

Start with this month's one thing

Everything here works better once AI knows how you sound. That's September's assignment, and it takes fifteen minutes.

Go to September's AI Strategy โ†’
Consistency Compounds. XOXO, The Jenns