A business broker can use AI to market listings by working only from a blind fact sheet: rounded figures, a codename and no names, addresses, customers or suppliers. The AI drafts buyer emails, portal copy and follow-ups from that sheet, and a leak check by a person runs before anything is sent. The seller's identity never enters a prompt.
Most AI marketing for listings goes wrong in one of two ways. Either it leaks (a detail in the draft points at the business), or it is generic ("a profitable, well-established business with strong growth potential") and no serious buyer reads past the first line. The workflow below deals with both. It treats confidentiality as a design problem you solve before writing, and specificity as something you supply from a legitimate source, the buyer's own criteria.
Why identity leaks happen even when the name is absent
Hiding the name is the easy part. Buyers, competitors and employees identify a business from combinations of ordinary facts. Before you use any AI tool, make a leak list for each listing: every detail that, alone or with others, could point to the seller. Typical items:
- The exact town, or a combination of area and unusual trade.
- Exact revenue or profit figures, particularly in small sectors where only a few firms exist.
- The year founded, particularly if round or notable.
- Named customers, suppliers, franchisors, software or equipment brands.
- Awards, certifications and licences specific to one operator.
- Number of staff in a small, distinctive team.
- A founder's career story, or phrases that appear on the business's own website.
The leak list drives two decisions. It tells you which facts to blur (a revenue of $1,873,400 becomes "around $1.9 million"), and which to drop entirely. It also becomes the checklist you run against every AI draft, because models often repeat details back in rewritten form.
Step 1: Write a blind fact sheet, by hand, once per listing
The blind fact sheet is the only document an AI tool sees. A broker writes it, not the model, from the confidential file. It has no more than 15 lines and uses ranges and categories.
BLIND FACT SHEET: "Project Harbour" (illustrative)
Sector: commercial cleaning services
Revenue band: $1.5m to $2m
Earnings band (after owner's salary): $300k to $400k
Years trading: 10 to 15
Staff: 15 to 25, mostly part-time
Customer mix: no single customer above 10% of revenue; mostly recurring contracts
Reason for sale: owner retiring
Assets included: vans, equipment, customer contracts
Growth angles: two sectors not yet pursued; no active sales effort in last 3 years
Transition: owner will stay 6 months
Asking range: $1.1m to $1.4m
Every line passed the leak list. This is the sheet you paste into prompts. If a buyer persona wants more than the sheet gives, that is a question for the NDA stage, covered in qualifying buyers with NDAs and AI screening.
Check the data rules of whichever assistant you use before pasting even this. Consumer ChatGPT and Claude accounts let you switch off training on your chats, and business plans such as ChatGPT Business and Claude Team do not train on content by default; both need at least two seats. The blind sheet reduces risk, but the seller agreement may restrict any use of a third-party tool, so check the wording with the seller and, if needed, a solicitor.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowStep 2: Segment the buyer list so the copy becomes specific
Generic copy usually means one email to everyone. Specificity does not need extra facts about the business; it needs a reason this buyer might care. Split your list by what they told you they wanted:
- Trade buyers in the same sector, who want customers, contracts or capacity.
- Adjacent operators who want to add a service line.
- First-time owner-operators who want a stable earner and a transition period.
- Investor groups who want earnings and low owner dependence.
Each group gets its own angle from the same blind fact sheet. That is the difference between "a profitable business is for sale" and a message that matches a buyer's stated aim.
If your buyer list lives in a CRM, how you tag it matters; see what to compare in a business broker CRM for the fields that make segmentation easy.
Step 3: Draft the outreach emails, one per segment
Here is a prompt that keeps the sheet as the only source and tells the model what to leave out:
You are helping a business broker write a first email to a buyer segment. Use only the blind fact sheet below. Do not guess the seller's name, area, customers or brands. Do not add facts. Segment: first-time owner-operators. Their stated aim: a stable business with a handover period. Write a 110-word email with a subject line under 50 characters. Tone: plain, respectful, no hype words such as "rare", "unique" or "opportunity of a lifetime". End with one clear next step: sign the NDA to receive the information pack. List any statement you were unsure about.
[paste blind fact sheet]
An illustrative output for that segment: "Subject: Cleaning business, owner retiring. Hi [first name], I am marketing a commercial cleaning business with revenue in the $1.5m to $2m range and recurring contracts, where no one customer is above 10% of sales. The owner plans to stay six months to hand over. If a stable, established business with a transition period fits what you are looking for, I can send an NDA, followed by an information pack. Would that be useful?" Fix: the model may add "established leader in its field", which is unsupported; delete it. Check the leak list again: "recurring contracts" and "10%" are on the blind sheet, and are safe.
The investor version, for contrast
For investor groups, the same facts become a different email: lead with earnings band, owner dependence (six-month stay, no sales effort in three years, so growth upside), and customer concentration. Ask the model to put the three numbers in the first two lines. Compare the two emails side by side; they should differ in order and emphasis, not in facts.
Step 4: Reuse the anonymous teaser without re-leaking
You will write or edit the anonymous teaser once, then reuse its facts across outreach, portals and social posts. Treat the approved teaser as a locked source: each reuse starts from it, not from a rewrite of a rewrite. Rewrites by AI drift, and details creep back in, such as a phrase from the seller's website that a model recalls. For the teaser itself, with before-and-after examples and a leak check, read the business teaser template tutorial.
When you reuse it, add a one-line instruction to every prompt: "Use only the approved teaser. Do not add any information that is not in it." This costs nothing and stops most drift.
Step 5: Portal copy, which is shorter and indexed
Listing portals publish your text to the world and, depending on the portal, to search engines. The copy has to be specific enough to attract and anonymous enough to be safe. A good structure: a 60-character title with sector and revenue band, a two-sentence summary, six bullet points, a closing sentence about the NDA. Prompt the model for three variants, then pick the safest.
An illustrative result for the Project Harbour sheet: "Commercial cleaning business, revenue $1.5m to $2m. Recurring contracts, no customer above 10%. Owner retiring; six-month handover. Two untouched service lines. Asking $1.1m to $1.4m. Sign an NDA for details." Check the title for phrases a search could trace: "commercial cleaning" is a category, so it is safe; a unique service name would not be.
Step 6: The follow-up cadence, without nagging
A cadence is a set of messages you send on a schedule. A simple one for a cold segment: day 0 teaser email; day 4 short nudge with one additional fact from the blind sheet; day 12 an update that is genuinely new (a date for information packs, or interest from other parties without naming them); day 21 a polite close. Four touches, over three weeks, to a list you have segmented.
Ask the assistant for each message in the series, with limits: 60 words for the nudge, 70 for the update, 40 for the close. For example: "Hi [first name], one quick follow-up. The owner is happy to stay on for six months, which several buyers have asked about. Shall I send the NDA?" Note that "several buyers have asked" must be true. Do not let the model invent interest, since false scarcity is dishonest and damages your reputation with the buyers you most want to keep.
When a buyer says no, log it, and exclude them from further touches. A tool or spreadsheet should track status, so automation never emails someone who has declined. For routines that run themselves, make sure the tool can read your do-not-contact list first.
What the AI may see, and what it may not
| Material | In a prompt? | Why |
|---|---|---|
| Blind fact sheet (ranges, categories) | Yes | Written by the broker; passes the leak list |
| Approved anonymous teaser | Yes | Already cleared with the seller |
| Buyer segment descriptions ("first-time owner-operators") | Yes | No personal data |
| Individual buyer names, emails and notes | No; use [first name] | Personal data; add at send time |
| Seller's name, address, website or marketing copy | Never | Direct and indirect identifiers |
| Financial statements, tax returns, contracts | Never in outreach prompts | Belongs to the NDA and information-pack stage |
| Employee names or customer lists | Never | Third parties who never agreed to it |
Pin that table above your desk. It settles most "can I paste this" questions in seconds, and it gives a new team member a rule they can follow without asking.
More outreach examples you can adapt
The adjacent-operator email
For an operator who already runs a related service, the angle is capacity and customers. An illustrative prompt tail: "Segment: adjacent operators who run residential cleaning and want commercial contracts. Stress recurring contracts, no customer above 10%, and the six-month handover. 100 words." A fair output: "Hi [first name], you run a residential cleaning firm, so this may be of interest: a commercial cleaning business with recurring contracts is for sale, with revenue between $1.5m and $2m and no customer above 10% of sales. The owner will stay six months. If adding commercial contracts is on your list, I can send an NDA and information pack." Fix: if the model writes "perfect fit for you", cut it. You do not know that.
A subject-line bank, tested against the leak list
Ask for 15 subject lines under 50 characters, with a ban on the sector's most distinctive word, the area and any number more precise than a range. Good illustrations: "Established service business for sale", "Owner retiring, six-month handover", "Recurring contracts, no big-customer risk". Poor ones, to reject: anything with an exclamation mark, "confidential" in capitals (it reads as bait), or a figure to the nearest thousand.
Handling "who is it?" replies
The most common reply to any anonymous outreach is a request for the name. Draft the answer once and reuse it: "I can't name the business until we have a signed NDA, and a short note on what you are looking for. The NDA takes about five minutes to complete, and the information pack follows within a day." Ask the assistant to write three versions in your tone, then keep one. The firm line protects the seller and, in practice, filters out buyers who are not serious. For what happens after they sign, see the NDA and screening tutorial.
A re-engagement email for a quiet buyer
After a buyer has gone silent for a month, send something genuinely new rather than "just checking in". If the seller has agreed a longer handover, or a new customer contract has been renewed (told in general terms), say so in one line. Illustrative: "Hi [first name], one update since we last spoke: the owner has agreed to stay for nine months rather than six. If your timing has shifted, I'd be glad to reopen the conversation. If not, no problem, and I'll note that." The model drafts, and you verify that the update is true and cleared with the seller.
Quick sums for planning a campaign
Illustrative numbers, to show how to plan: suppose your segmented list has 120 buyers. A first email to the best 30 might produce 6 replies, 4 NDA requests and 3 signed NDAs. That would be a 20% reply rate and a 10% NDA rate. Those percentages are placeholders, and your own history will differ. The useful part is the method: set the expected number of NDAs you need to get two serious conversations (say, 8), work backwards to the emails required (about 80), and decide whether your list is big enough. If it is not, widen the segments rather than blasting the same buyers again.
Time matters too. Writing each segment email by hand might take 20 minutes; with a blind fact sheet and a tested prompt, a first draft takes five and your edit ten. Four segments saves about an hour per listing, which is real but modest. The bigger gain is consistency: every listing gets the same leak check and the same cadence.
Record-keeping that protects you later
Keep, per listing, the blind fact sheet, the approved teaser, the leak list, the sent emails and a status column for each buyer. If a seller ever asks what was said about their business, or a buyer claims they were misled, this record answers it. Store them in your CRM or a locked folder. Do not keep AI chat histories containing even the blind sheet longer than you need; delete conversations when a listing closes, as your data rules require.
The leak check that runs before every send
This step is a person with the leak list, not another AI. Read the draft against the list and tick each item: location, exact figures, dates, names, brands, unusual features, wording lifted from the seller's own materials. Then do the "search test": paste three distinctive phrases from the draft into a search engine in a private browsing window. If any returns the seller's website, rewrite the phrase. It takes about three minutes, and it catches lifted phrases that a read-through misses, because it tests what a stranger could actually find.
A realistic mistake and how it showed up
An illustrative case: a broker asked an assistant to "make the teaser more compelling" and pasted the seller's website About page for tone. The draft came back with "fifth-generation family business", a phrase from that page. Anyone searching it would have found the seller in seconds. The leak check caught it, because "generation" was on the leak list under founder story. The lesson is to paste only the blind fact sheet, never the seller's own marketing.
A second mistake: false precision
Another trap is the model sharpening a range into one number: "revenue of around $1.75m" from "$1.5m to $2m". That is a fact you never supplied, and a buyer will hold you to it. Add the instruction "keep every figure as the range given" to your prompt and check the output.
A week-by-week plan for a new listing
Week one: build the leak list and blind fact sheet, and agree them with the seller in writing. Draft the teaser, and have it checked by the seller. Week two: segment the buyer list, generate one email per segment, run the leak check, and send the first batch to your best-matching 20 to 30 contacts. Week three: send the nudge, post the portal copy, and log replies. Week four: send the update and the close, review who signed NDAs, and decide whether the listing needs repositioning. The numbers in this plan are illustrative; adjust to your own list size and sector.
How to tell whether it is working
Track four figures per listing: emails sent, replies, NDAs signed and conversations booked. Compare them with your last few listings that you marketed without AI. If replies are up but NDAs are flat, your emails are attracting the wrong buyers, so tighten the segment. If NDAs are up but no calls follow, your information pack or your response speed is the problem, not the outreach. And if any seller asks "how did they know it was me?", stop the campaign and go back to the leak list.
The valuation figure that anchors your asking range is also worth checking before anything is published, because outreach that quotes a range you cannot defend costs credibility with exactly the buyers who know the sector.
Questions brokers ask about AI and confidential listings
Can I paste a listing's financials into ChatGPT?
Not into a consumer account, and not with identifying detail on any plan unless your seller agreement allows it. Work from a blind fact sheet with rounded figures and no name, address, customer or supplier names, so nothing in the prompt can reveal the seller.
Will AI outreach emails look like spam?
They do when they are generic. Segment your buyer list first, give the model one real fact about each segment's interest, and edit every email. Short, specific, plain emails sent to a small relevant list perform better than long polished ones sent to everyone.
How often should I follow up with a buyer?
A common cadence is three or four touches over about three weeks: the teaser, a short nudge, a relevant update and a polite close. Stop at once when a buyer says no, and record it so it is never repeated.
Should the portal listing match the teaser?
It can reuse the teaser's facts, but portal copy should be shorter, keyword-led and checked separately for leaks, because portals index text and sometimes show it in search results.
Further reads
- Confidential Information Memorandum AI: A Business Broker's Workflow — The next document buyers see after the NDA.
- Due Diligence Checklist for Selling a Business: Build One With AI — Prepare the sale for buyer scrutiny.
- Seller Discretionary Earnings With AI: Recast and Find Add-Backs — Recast numbers before they go into any outreach.
- Do AI Business Valuation Tools Work? What They Miss for Brokers — Why AI valuation tools can mislead on pricing.
- How to Automate Buyer and Vendor Follow-Up With AI — Follow-up routines you can automate safely.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ChatGPT Business and Claude pricing pages and OpenAI's data-controls help page (checked 4 November 2026).