One of the landlords we’d sent a proposal to told me this:

“Yeah, I uploaded your proposal to ChatGPT and was asking it about you guys.”

That one sentence explained the last few months.

Something was off

Over the last few months, our conversations with landlords and property investors started to change. Several of my salespeople came to me with the same thing: a handful of very specific new questions kept coming up. They came from different clients, but they were the same questions.

At first I put it down to coincidence. Then it kept happening, and it stopped looking like one.

I’ve been in property for a decade and I’ve done thousands of sales calls. There’s very little I haven’t heard. These questions were new.

The test

So I tried something. I took our first emails and proposal emails, dropped them into ChatGPT, and pretended to be a client weighing up our services.

It came back with the same objections our clients had been raising.

Then I noticed something else. It was also handling a lot of the common objections we used to handle ourselves. Without us knowing, it had been doing some of our selling for us. It had also been doing some of our unselling.

Your proposal has a second reader

You’re not only writing to a person any more. Every email and proposal you send can also be read by whichever AI your prospect asks about you. You’re selling to both at once, and they don’t read the same way.

The person on the call remembers how it went: your answers, your tone, the moment they got excited. The AI only sees the paperwork.

That’s why a proposal can be perfect for a human and still make an AI sceptical.

ChatGPT seems built to make people feel they got what they wanted to hear. Ask it what to worry about, and it will find you something to worry about.

Two things stood out:

  1. It can read silence as concealment. If you don’t bring something up in the proposal, ChatGPT can assume you’re leaving it out on purpose.
  2. It can scare people. The client who got excited on the call can read an AI’s take later and get cold feet. You’ll never hear why.

If you use AI every day with custom instructions and long chat histories, you may be the worst-placed person to spot this. Your client types one basic question into a fresh chat. Most of us are out of touch with what that basic question produces.

Try it yourself

Before your next proposal goes out, open a fresh ChatGPT chat with no custom instructions and no history. Paste in the proposal and ask:

“I’m thinking of hiring this company. What should I be worried about?”

Then read the answer the way your client will: alone, after the call, with nobody there to explain.

What to check before you hit send

  • Numbers that don’t match. The figure in your email differs slightly from the one on your website. A human might not notice. An AI comparing the two will, and it will ask which one is true.
  • Things you didn’t mention. Fees, contract length, notice periods, what happens if it goes wrong. If it isn’t there, the AI may assume there’s a reason.
  • Names that don’t line up. The proposal uses one company name and the public record shows a different registered name. That’s often perfectly normal, but if you don’t explain it, it looks like something to dig into.
  • Dates that don’t line up. “Established in” on your website against the incorporation date on the public record.
  • Vague promises. “Competitive fees”, “strong returns”, “up to” figures with nothing behind them. A human fills the gap with trust. An AI flags it.
  • Claims nobody can find. A track record stated in the proposal that appears nowhere public.

A careful human buyer might catch some of these too. The difference is that the human had the call with you. The AI only has the paperwork.

Some of it isn’t a copy problem

Not everything it flagged could be fixed with better wording. Where the gap was real, the fix had to happen in the business, not in the proposal.

Close the gap. Don’t hide it. A hidden gap is exactly the thing it’s looking for.

The upside

I’ve spent the last three weeks rebuilding our offers and proposals with that second reader in mind.

If you can convince ChatGPT, it starts doing part of your selling for you. It answers the objection when the client actually has it, and that’s the one moment your salesperson isn’t in the room.

The test also sent me down a second rabbit hole: what gets an AI to recommend you in the first place. I’m in the middle of that one now.

I’m still working on all of this. Next, I’ll write up exactly what we changed.

Moemen

More about my work