we've been running a lot of ai generated outreach copy through head to head testing lately, different models, different prompts, same audience, and the results keep teaching the same unglamorous lesson: the clever version rarely wins. the clear version almost always does.
it's tempting to reward the output that made you laugh or feel impressed reading it. that's your reaction as someone evaluating twenty variants in a row, not the reaction of a busy stranger seeing one email in a crowded inbox for three seconds. that stranger isn't looking for clever, they're scanning for whether this is relevant to them specifically, and clever often adds friction to that scan instead of removing it.
the copy that consistently wins in these tests reads almost boringly direct: here's the specific thing i noticed about you, here's the specific problem it implies, here's the specific next step if that's relevant. no wordplay, no big opening hook, just fast relevance. it feels like it shouldn't work as well as it does, and then the reply rate numbers come in and it always does.
this doesn't mean personality doesn't matter, it means personality is a seasoning, not the main dish. clarity is the main dish. get that right first, then let personality show up in how you say the clear thing, not instead of saying it clearly.
in your own outreach, are you optimizing for what impresses you rereading it, or for what a distracted stranger scans in three seconds?