Good morning. Can you believe itās already September? Same here.
You know what that means, itās time to lock inā¦
Anyways, hereās a few things that caught my eye.
š¤ Early ChatGPT ad tests are getting clicks without buyers
Yesterday, we shared that OpenAI said ChatGPT ads had reached a $1 billion annualized run rate. Now the early campaign results are rolling in, and theyāre messy. MediaPost spoke with one agency that spent $415 targeting Ontario and Alberta, received a 0.6% CTR, and found zero interested buyers. Another advertiser saw 57 clicks inside ChatGPT and fewer than 20 visits in Google Analytics.
That first campaign ran in June, before OpenAI added better targeting and conversion tools in August, so itās not a clean verdict on the current product. Practitioner results are also split: @irentdumpsters says $20,000 in spend brought back $69,000, while @colindougherty says agency owners are reporting poor performance and unreliable measurement. Keep the test budget small, and judge it using your own analytics.
š® Researchers say 73 of Temuās top 100 Meta creators may be fake
Online Risk Labs studied the accounts behind Temuās Partnership Ads in the UK and EU and found that 73 of the top 100 were likely fake. Many changed names repeatedly, operated from Russia or China while targeting European shoppers, and had no identity researchers could verify. One account appeared in nearly 109,500 Temu campaigns and generated more than one billion views.
The group estimates Temu spent up to $962 million on Partnership Ads during the 16-month study period, though that figure covers the whole format, not only the suspicious accounts. Thereās no evidence Temu created the accounts or knew they were fake. EU regulators have now been asked to assess the network, so brands running Partnership Ads should expect creator verification and disclosure to get a closer look.
š Google tests AI-generated social proof in sponsored search results
Google is testing a āWhat customers loveā section inside sponsored search results. The copy is generated from store-rating reviews and can appear beside your ad while someone is deciding whether to click. Google has not said whether advertisers can approve the summary or correct it before it runs. Search for your own branded ads, check the reviews feeding your store rating, and see what Google has decided customers love about you.
š¤ Anthropic released AI agents that can fill carts and recommend price changes
Anthropic has released retail agent blueprints ahead of the holiday season. Brands can build a Claude-powered shopping assistant that recommends products and adds them to a cart, plus a merchant agent that suggests changes to inventory, pricing, and marketing. Anthropic says one partner increased cart size by 30% to 35% and made shoppers about 60% more likely to finish their purchase.
The agents canāt complete transactions, so this is still a build project, not a magic checkout button. If you test one before Q4, start with the product questions already slowing shoppers down.
š” Nail your Q4 offer structure before BFCM hits
Q4 revenue rides on your offer structure. Structure it right and your average order value climbs while your margins hold.
Pilothouse's take: put your energy into AOV this quarter. Here's what to test, and when to start.
1ļøā£ Tiered offers
Set spending milestones: $100 gets 10% off, $150 gets 15% off, $225 gets 25% off.
Every tier gives shoppers a reason to add one more item, and you can stack in free shipping depending on your usual policy.
2ļøā£ Build-your-own bundles
Let customers curate their own set for a tiered discount: buy 2, get 10% off; buy 3, get 15% off. It feels personalized, and it's a natural fit for anyone building multiple gift sets.
3ļøā£ Mystery or holiday boxes
Pair your best-selling SKU with a couple of slower movers in an exclusive, discounted holiday box.
Shoppers feel like they got a deal, and you clear slow inventory without marking down your bestsellers.
4ļøā£ Start testing now
Don't wait until BFCM to find out if an offer lands. Set up a split test on Meta with an ABO campaign, run 2 to 3 offers against your best evergreen creative, and see which hook drives the highest AOV and CVR heading into November.
Test positioning alongside the offer itself. "Buy 2, get 1 free" and "$30 off bundles" can be the same discount and still perform completely differently.
Start testing before Labor Day weekend. You want your winning offer locked before BFCM traffic hits.
š§ 0.25% CTR and Not in Spam: What Do You Fix Next?
What do you fix when your email program has a 0.25% click-through rate and deliverability isnāt the problem?
This week on TWBERP, Jordan breaks down a pattern he sees across larger DTC brands: daily campaigns, oversized automated flow systems, low engagement, and plenty of attributed revenue that makes the program look healthier than it really is.
He walks through how to find your real click-through rate, strip out bot activity, identify which flow messages are actually generating revenue, and figure out where relevance can still create meaningful lift.
ā¶ļø Watch here | š§ Listen on Spotify
āš Your reel pops off. Now what? Your Instagram comments and DMs are packed. Wati replies and hands off the real ones to you. Start your free trial ā *
*Ā sponsored
Did you find this valuable? Support us by checking out our businesses:
Please note that items in this newsletter marked with * contain sponsored content.
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
