Buyers say they can spot AI-written text, and the confidence is the first thing worth noting. What they are usually recognising is not authorship but style: a particular smoothness, a habit of hedging before every claim, and a tendency to restate the brief back at them in slightly grander language. None of those is unique to a language model. A junior associate under deadline pressure produces the same tells.

The cues people point to

Ask a buyer what gave it away and the answers cluster around a handful of things: sentences that all run the same length, transitions that lean on the same few connective words, praise for the buyer's business that could apply to any business in the sector, and a structure that answers the brief point by point without ever pushing back on it. These are real patterns in a lot of generated text. They are also patterns in proposals written by a tired human working from a template, which is why the cue is weaker evidence than it feels.

Where the tell breaks down

The reliable signal is not style, it is specificity. A proposal that names the buyer's actual constraint, a number from their last set of accounts, a detail from the conversation that only that buyer could have supplied, reads as considered regardless of what wrote the first draft. A proposal that could be sent to three other companies with the names swapped reads as generated regardless of who typed it. Buyers who say they caught an AI-written proposal are usually catching a generic one, and the two overlap a great deal but are not the same thing.

What this means for what you send

If a draft comes from a model, the work that follows matters more than the drafting did: pulling in the specific figures, the stakeholder's own words from the meeting, the objection that came up and has to be answered rather than smoothed over. That editing pass is what separates a proposal that reads as considered from one that reads as templated, whichever tool wrote the first version. The full guide on how AI is changing the pitch covers where these tools help in the process and where they quietly cost you the deal.