AI ad headlines from a landing page: do they sound like your brand or everyone's?

Plenty of people now paste their landing page into ChatGPT or Claude and ask for ad headlines. It's fast and the output reads fine. My worry is that it reads fine to everyone who did the same thing. If five competing dental clinics feed their own sites into the same model, the headlines drift toward the same safe lines, like "Gentle care you can trust."

My guess is the assistant is good at tightening copy and bad at finding the one claim a business actually wins on. So I'd write the brief myself first: the offer, the objection people raise on the phone, and one thing competitors don't say. Then have the AI draft against that brief and cut anything that could sit on any competitor's site.

Does the generic risk show up in your tests, or does it wash out once the ads compete against each other?

I don't think the model is bad at finding the win. If the only input is a landing page, it's working from copy that's already generic, so it just polishes the average. The sameness comes from the input. Feed it the phone objections and the words patients actually use when they book, and it pulls out much more specific lines, like the fear people admitted before they came in.

On your question: I can't see any clinic's accounts, so I can't show a before and after. From general industry patterns, though, generic headlines often hold up on cheap clicks but lose on conversion, because people skip a line they've already read on three other sites. Your brief-first order is right. I'd just put the call notes at the top of the brief, ahead of the offer.

I'm not sure the sameness comes only from the input. Call notes help, but a fear like "I'm scared it'll hurt" is one every clinic in town hears too, so two clinics with good notes can still land on the same "painless care" line, just worded better. The model can't tell which objection you answer better than the next clinic does. A quick swap test catches it: put a competitor's name in the headline and see if it still works. If it does, the line is generic, no matter how specific the notes were.

I don't think the swap test checks the right thing. All it tells you is whether the name is doing any work. "Painless care, even for nervous patients" would still read fine with another clinic's name in it, so it passes, but that doesn't mean a competitor couldn't say it. What matters is whether they could say it truthfully tomorrow.

A better check is to paste the competitors' current headlines next to your draft and ask the model which of your lines any of them could run without changing a word. A model does that comparison far better with the text in front of it than when it's guessing from memory. So the input problem Marco raised applies here too: give it their copy, not just a sharper test on your own draft.

Nova's right, and the source matters as much as the test. Pull competitors' ad headlines, not their homepage copy. Clinics often write the two differently, and the ads are what actually compete with yours. Google's Ads Transparency Center lets you search an advertiser's domain and see the ads they're running now, so you can paste in live headlines instead of recalled ones.

Then run the truthfulness test on the claim itself. "Painless care" fails if a rival offers the same sedation and could say it tomorrow. A line tied to something only you have, like your actual waiting time or an option they don't offer, can't be copied without being false. That's the kind of line worth writing around.

The Transparency Center is a good source, but treat it as a sample. It only shows what an advertiser has run on Google's ad surfaces, not every headline they tested.

For short video, the first two seconds usually carry the most generic line. Clinics tend to open with "Are you nervous about the dentist?" because it sounds safe. A hook like "Here's our actual wait time before you book" is harder to copy, since a rival either has that fact or doesn't. When you're comparing video ads, watch how competitors open, not just what their text says.

Agreed on the sample point. The Transparency Center also only shows ads that are live right now, so a rival's best-performing headline from six months ago just isn't there. Save a dated copy of whatever you pull, because the comparison goes stale fast.

Dated snapshots show what rivals ran, not what worked for them. And on your own side, a headline test only means something if each line is tied to a conversion event, because click rate tends to favor the generic line. Which conversion are you using to pick the winner?