Straight answer: if you’re using ChatGPT or Claude to write your full proposals, your close rate is probably dropping even though the writing looks better than ever. Clients can’t tell your proposal apart from three others that read exactly the same, so they default to picking the cheapest one. The fix isn’t dropping AI, it’s using it for the 80 percent of a proposal that doesn’t matter and writing the 20 percent that wins the job yourself.
Worth reading next: Which Parts Of Your Marketing AI Is Good At (And Which It Isn’t).
The four months I didn’t notice my close rate was falling
In early 2025 I started using AI to draft every proposal I sent out. Structure, pricing rationale, the opening paragraph, all of it. It was faster. It read well. Grammar clean, tone professional, no typos at 11pm when I was finishing one after a long day.
What I didn’t clock for about four months was that my close rate had dropped from roughly 40 percent to 22 percent. Same offer, same prices, same me on the discovery call. I only noticed because I finally sat down and reread six rejected proposals side by side, and they all opened with a version of the same line: “Thank you for taking the time to discuss your goals, I understand you’re looking to…” Competent. Forgettable. Interchangeable with anything a competitor’s AI would have produced for the exact same brief.
A client I coach in Leeds, who runs a small design studio, told me almost the same thing had happened to her. She’d started using AI for proposals in autumn 2025 and noticed prospects were coming back with “can you match this quote” emails far more than before. Her work hadn’t changed. Her writing had, and it had changed in a way that made her replaceable.
What’s happening: the sameness problem
Here’s the uncomfortable bit nobody selling you an AI writing tool wants to say out loud: when everyone uses the same three models to write their proposals, everyone’s proposals start to sound like the same person wrote them. Not badly. Just identically. The client isn’t comparing your expertise anymore because there’s nothing distinctive left to compare. They’re comparing numbers, because numbers are the only thing left that’s different between the three PDFs sitting in their inbox.
I did a rough audit of 40 proposals sent to me by small business owners I coach, asking them to send me their last one whether it won or lost. 34 of them opened with a near-identical structure: gratitude line, restated problem, three bullet benefits, pricing table, next steps. None of that is wrong. But when 85 percent of proposals share a skeleton, the client’s brain stops reading for quality and starts reading for the total at the bottom.
This is the same trap brands fall into when they copy a competitor’s playbook without understanding why it worked. I wrote about this when I looked at how Revolut built a brand that wins on distinctiveness rather than sameness, and the principle holds just as well for a one-page proposal as it does for a fintech app.
The three tells that give an AI proposal away
- The opening line thanks the client for their time before saying anything specific about their business.
- The benefits section lists three bullets that could apply to almost any client in the same industry.
- There’s no reference to anything that was said on the call, no specific number, name, or detail the client mentioned.
Clients might not consciously spot all three. But they feel the lack of specificity even when they can’t name why. It reads like a template with their company name dropped in, because that’s exactly what it is.
Why this costs you money, not just polish
The contrarian bit that most people writing about AI and proposals skip: it’s not that AI writing is bad. It’s that AI writing is good enough that it stops being a signal. For years, a well-written proposal was itself a small proof point, it showed you were careful, thoughtful, paid attention. Now that anyone can produce a well-written proposal in ninety seconds, being well-written no longer proves anything. It’s the baseline, not the differentiator. Which means the client’s decision moves entirely onto price, because price is the only variable left that clearly differs between you and the next quote in their inbox.
This is exactly the mechanism Dale Carnegie was describing decades before any of us had heard of a large language model. I wrote about his core idea in my piece on business lessons from Dale Carnegie, and the one that applies here is simple: people don’t buy from the most competent-sounding option, they buy from the person who made them feel specifically understood. AI, left to its own devices, produces generically competent. It cannot produce specifically understood, because it wasn’t on your call.
What I do instead now
I didn’t stop using AI for proposals. I changed which 20 percent I write myself and which 80 percent I hand off. Here’s the process, roughly what I’ve been running since late 2025:
- Step 1, within an hour of the call: I write three sentences by hand, no AI involved. One states the actual problem in the client’s own words, if they said “we’re bleeding leads on weekends” I use that phrase, not “improve lead generation.” One references a specific number or detail they gave me. One states what changes if we fix it, in their language, not mine.
- Step 2: I feed those three sentences to AI along with my standard pricing structure and ask it to build the scaffolding, timeline, deliverables table, terms. This is the part AI is fast at and the part clients never read closely anyway.
- Step 3: I rewrite the opening paragraph and the closing paragraph myself, every time. These are the only two paragraphs a client reads rather than skims. If they read like everyone else’s, you’ve lost before the pricing table.
- Step 4, the ninety second test: I read the finished proposal aloud. If I could delete the client’s name and send the exact same document to a different client in a different industry without changing anything but the logo, I rewrite it. That’s the sameness test and it catches almost everything.
Since making that change my close rate has climbed back to around 36 percent, not quite the original 40 but close, and the proposals still take me under twenty minutes to send.
Where AI still earns its place in proposal writing
I’m not telling anyone to stop using AI here, that would be a bit rich coming from someone who writes about AI for a living. It’s useful for the parts of a proposal that are mechanical rather than persuasive:
- Formatting the pricing table and payment terms consistently.
- Drafting the scope of work bullets once you’ve given it the specifics.
- Checking tone consistency across a long document.
- Generating a second and third pricing option quickly once you’ve decided the numbers.
What it shouldn’t be trusted with is the opening line, the closing line, or anything that’s meant to prove you were paying attention. That’s the human 20 percent, and it’s the part that decides whether you win.
The same problem is showing up in your marketing, not just your proposals
This isn’t limited to sales documents. I’ve watched the exact same sameness creep into small business email newsletters, LinkedIn posts, even website copy, all written with AI, all technically fine, none of it memorable. Brands that stand out right now are doing the opposite of blending in. Look at how Duolingo built a brand around a deliberately odd, unmistakable voice, or how Calm has stayed distinctive by refusing to sound like every other wellness app, or how Hotjar built trust through specificity rather than polish. None of that came from a prompt. It came from a person deciding what the brand was allowed to sound like and then holding AI to that line rather than the other way round.
If you want a low-effort way to make your own site feel less like everyone else’s AI-generated funnel, something like an interactive calculator that gives a visitor a specific, personal number does more for differentiation than another paragraph of generic benefit copy ever will, because it’s something a template can’t easily fake.
If you’re not sure where to start
Most small business owners I speak to know something has shifted, they just can’t pin down what to fix first, the writing, the process, or the tools. If that’s you, it’s worth getting a second pair of eyes on where AI is helping your business and where it’s quietly flattening it, which is exactly the kind of gap I work through with clients as an AI implementation coach rather than trying to fix it alone at 11pm before a deadline.
Related reading: ai lead generating calculators small business.
Frequently asked questions
Is it bad to use ChatGPT to write a business proposal?
Not on its own, but if you let it write the whole thing, including the opening and closing paragraphs, it will read like every other AI-written proposal your client has received that month. Use it for structure, pricing tables and scope bullets, and write the first and last paragraph yourself using specific details from the actual conversation.
Why do my AI-written proposals get compared on price?
Because when the writing itself no longer signals anything distinctive, price becomes the only clear difference left between competing quotes. A generically well-written proposal used to prove you were careful. Now that everyone can produce one in under a minute, that proof is gone, so clients default to whichever number is lowest.
How do I make an AI-drafted proposal sound less generic?
Run the ninety second test: read it aloud and ask whether you could send the exact same document to a completely different client without changing anything but the name. If yes, rewrite the opening and closing paragraphs with a specific phrase, number or detail the client said on your call.
Should small businesses stop using AI for sales and marketing writing altogether?
No, that’s overcorrecting. AI is fast and reliable for the mechanical parts of writing, formatting, structure, first drafts of scope or terms. The part worth protecting is the persuasive core, the opening line, the pricing rationale, the closing ask, because that’s the part clients notice and the part that currently gets flattened into sameness across every AI-written document they receive.
Related reading: Why Your AI Still Writes Like Everyone Else’s (And the 45-Minute Fix That Changes That) and AI-Written Proposals Are Losing You Clients (Here’s What to Send Instead).
If you want the full breakdown, here is everything I know about AI marketing.
Related: ai email sameness small business voice.
