ChatGPT advert focusing on is rubbish

I’ve been experimenting with shopping for ChatGPT advertisements for my occasion seating task software program. Listed here are the general outcomes from 05-Aug-2026 to 28-Aug-2026:

ChatGPT recommends bidding someplace round by £2 per click on, however I bid between £0.15 and £0.07 and was nonetheless in a position to get site visitors. Nevertheless 0.46% is a surprisingly low conversion charge. Given what I do know in regards to the lifetime buyer worth and set up to buy charges for PerfectTablePlan, I’m not shut to creating any cash on these advertisements.

Evaluating engagement with different sources in the identical interval:

So the issue is clearly the ChatGPT advert site visitors, fairly than my web site/product. It appears one among a number of issues may be taking place:

1. OpenAI is charging for extra clicks than they’re sending.
2. My context trace is unhealthy.
3. My advertisements are unhealthy.
4. A excessive proportion of the clicks are from bots, not people.
5. The site visitors could be very badly targetted.

The clicks reported by OpenAI are near what Google Analytics experiences, so we are able to rule 1 out.

My context trace appears uncontroversial, in order that guidelines out 2.

The advertisements are similar to ones I’ve been working in Google Adwords for years, so we are able to rule 3 out.

I reached out to digital advert fraud professional Dr Augustine Fou, half manner by way of the experiment, to ask him whether or not he thought bots may be chargeable for a excessive proportion of the clicks. He was type sufficient to assist me instrument my web site with fouanalytics.com. Taking a look at time on web page, mouse actions and different metrics have been in a position to set up {that a} majority of the clicks have been possible people. Right here you’ll be able to see the Fouanalytics evaluation, with 30% displaying ‘many crimson flags’ or ‘some crimson flags’ for being bots.

So there may be some bot site visitors, however not sufficient to account for the horrible conversion charge. So we are able to rule 4 out.

That solely leaves 5. The site visitors is poorly targetted. That is supported by information from Fouanalytics. I wrangled the information in Simple Knowledge Remodel to get a histogram of the time on the touchdown web page, the place every bin is 1 second:

Fouanalytics additionally reveals that 88% of the site visitors moved their mouse, however solely 10% clicked something. Which is an extra indication of poorly targetted site visitors.

Concentrating on is every part in the case of promoting. Clearly ChatGPT is aware of learn how to ship me targetted site visitors, as a result of their free referrals convert an order of magnitude higher than their paid advertisements. Perhaps they’re so determined to fill the yawning void of their funds that they’re producing as many paid clicks as they’ll, with out caring an excessive amount of in regards to the high quality?

I might be fascinated to know in what context they’re displaying my advertisements. However there isn’t any manner for me to search out that out. Consequently, I don’t know the way I can enhance this conversion charge. Even when I did, I wouldn’t be hopeful about enhancing the conversion charge by the order of magnitude required for profitability.

Right here is Dr Fou’s conclusion:

“Finest practices confirmed by this experiment — all the time use actual conversions because the measure of success. The issue with ChatGPT advertisements was NOT bots; it was low worth guests (ones that arrived on the touchdown web page however did nothing else and left). It’s not completely clear WHY these guests are so low in worth (e.g. have been the ChatGPT advertisements merely focused very poorly?) Make investments your advert budgets in sources that not solely yield extra clicks to your website, however most significantly extra precise conversions.” Dr Augustine Fou

(Edit: Up to date the Fouanalytics chart to the right chart)d