Do Chat GPT Paid Ads Work?
More than 6,000 brands are now buying ads in a channel nobody can fully measure yet.
That's ChatGPT advertising this week. Digiday reports ChatGPT now handles roughly a fifth of the search queries Google does. Adthena counted 1,082 brands advertising there in June and 6,333 by August. Similarweb puts ad reach at about 43% of ad-eligible U.S. users. The balance is shifting from Search to LLMs.
The proof hasn't caught up. A working paper cited by Digiday found paid ChatGPT traffic converting at 0.37%, against 1.17% for paid Google Search. Revenue per session: $0.89 vs. $2.45.
Most people will read that as "ChatGPT ads don't work." I read it differently.
People don't really open ChatGPT to buy. They open it to think, compare, window shop, and decide. That's where trust gets built and shortlists get made. Judge it on last click and it will almost always lose to search. Ignore it and you hand the shopper to someone else.
The mistake isn't testing a new paid media channel. It's testing it with the wrong scorecard.
Three moves for Marketing leaders for this week:
- Decide what the channel is for before you fund it. Shortlist share or direct sales? Pick one, and set the target.
- Run a holdout. Compare branded search, direct traffic and pipeline in exposed vs. control groups, not just clicks.
- Fix the answer before you buy the ad. If ChatGPT describes your brand poorly in its organic answers, a paid placement beside it won't save you.
I've seen this firsthand pattern with Facebook, podcasts and streaming TV. The early winners didn't wait for perfect measurement. They built their own.
Reach is easy to buy. Anybody can buy media.
Proof you have to build.
