Y Combinator Founders Say AI Applications Are Starting to Sound the Same — Here's What Gives Them Away
The Same Sentence, Different Startup
Are you noticing the same three sentences show up in every YC application you review? Do the pitches all somehow sound like they came from the same founder, even when the companies have nothing in common?
Y Combinator partners have started flagging this pattern out loud. Founders submitting applications for the batch are leaning on AI to draft their pitch, and the writing has developed a fingerprint of its own. Reviewers describe reading dozens of applications a day and hitting the same rhythm over and over: a problem statement that opens with "founders spend countless hours," a solution paragraph that promises to "streamline" and "unlock," a traction section that leans on words like "robust" and "scalable" without ever naming a number.
The unsettling part isn't that the writing is bad. It usually reads fine on a sentence level. The unsettling part is what's missing. A partner reading a hundred applications a week starts to notice when the person behind the words has gone quiet. The application says everything a good application should say, and somehow says nothing about who is actually building the thing.
That gap between polished and present is what this article is actually about.
What Actually Tips Reviewers Off
Three things give a generated application away, and none of them require a plagiarism checker to spot.
The first is pacing. Every paragraph runs the same length, three to four sentences, each one building toward a tidy close. Real founders write unevenly. They spend six sentences on the problem because they lived it and one flat sentence on the go-to-market plan because they haven't figured it out yet. AI writing doesn't have that lopsidedness. It distributes attention evenly across sections regardless of what the founder actually knows cold.
The second is the absence of numbers. A founder who has talked to forty potential customers says forty. A founder who has closed three pilot contracts names the three. Generic AI output tends to gesture at traction with words like "strong early interest" or "growing demand" because the model has no number to reach for. When a reviewer sees vague quantities where specifics should sit, that's usually the first flag.
The third is portability. Swap out the product name and the industry, and the answer to "why now" or "what's your unfair advantage" still reads fine for a fintech app, a biotech app, or a dev tool. That interchangeability is the tell. A founder-specific answer breaks the moment you try to move it.
Why Generic Writing Reads as Dishonest
This bothers reviewers more than a typo ever would, and the reason comes down to what an application is actually for.
A YC application isn't a writing sample. It's a proxy. Before a partner ever gets a founder on a call, the application is the only signal they have for how that person thinks under constraint, what they notice, what they choose to leave out. A founder who writes "we talked to forty restaurant owners and thirty-one said scheduling was their biggest pain point" is showing you their brain working through a problem. A founder who writes "there is strong demand in the market" is showing you nothing, even if every word is technically true.
When the writing goes generic, that signal disappears. Reviewers aren't offended by AI use itself. Plenty of strong founders run their draft through a model to tighten grammar or cut length. What reads as dishonest is when the application no longer carries any trace of the specific person behind it, because the whole point of asking someone to write four hundred words about their company was never really about the four hundred words. It was about getting a preview of the founder before the Zoom call. Strip out the voice, and the preview is blank.
Sounding Like Yourself Is the Actual Advantage
None of this means founders should stop using AI to draft. The problem was never the tool. A model that helps you cut a rambling paragraph down to size is doing what editing has always done. The problem is when the draft becomes the final version, and the founder's actual way of explaining things never makes it back in.
The fix is smaller than it sounds. Read your application out loud before you send it. If a sentence sounds like something you'd never say to a friend over coffee, cut it. Put the specific number back in, even if it's a small one. Three pilot customers is a better sentence than strong early interest, and it's the sentence only you can write, because you're the one who ran the pilots.
This is the same problem PostMimic was built around, just in a different room. A pitch deck, a LinkedIn post, an investor update — they all get flattened the same way once a model does the first draft and nobody puts the person back in. Learning what your own writing actually sounds like, and knowing when a sentence has drifted away from it, is worth more than a smoother paragraph. Reviewers can tell the difference. So can customers.