AI-generated ads perform roughly as well as human-created ads, unless consumers suspect they were generated by AI.
Researchers from the Technical University of Munich, Columbia Business School, Harvard Business School, and elsewhere analyzed more than 16 billion ad impressions and 116 million clicks from Tabola. In a novel quasi-experiment, they compared the performance of AI-generated and human-generated "sibling ads": ads from the same campaigns, for the same products, and with the same launch dates.
They find that CTR performance for AI-generated ads is statistically "indistinguishable" from human-created ads when controlling for campaign-level factors. But they also reached an equally important conclusion: ads that look AI-generated perform worse.
The researchers identified a set of visual traits that consumers identify as "AI-like," and found that perceived artificiality reduced click-through rates even when the ads were not explicitly flagged as AI-generated. Ironically, the authors find that a not-insignificant number of human-generated ads were perceived as AI-generated for possessing those visual traits.
This is an important result. The cost of creative production through generative AI is decreasing rapidly, especially for static images. But not all AI-generated creative performs equivalently. Teams need to consider an optimization for their AI-enabled workflow beyond mere volume, which is the perceived visual fidelity and provenance of the ads being generated.
Link to paper below. Thanks to @garjoh_canuck for flagging this for me.
@doodlestein@patio11 You may want to play with term and think say, 25 vs 30 year term, will make a big difference here. (Source: I built a life insurance rates index - https://t.co/prfMpYM1WV - but $600-ish is a pretty reasonable price for that.
@Danhightower@patio11 This might be bc they are thinking wellbutrin is a smoking cessation RX rather than depression? Either way, depression (which I assume is the issue here?) does get you rated up (some), depending on the carrier.
@patio11 And if you'd like to see what low rates are available (without putting in your email), please check out our CoverSavvy Term Rates Index - https://t.co/ldtLpiLieQ - and check out how much it will be. Everything Patio11 is saying is right on!
William is a great reminder that the people who win become obsessed with something very small and stick with it far longer than anyone else.
Enjoy three minutes of him explaining how to become the very best at one thing:
"I'm probably the best in the world at a couple small, boring stuff. I do think the world for builders is the world of specialists. And you have to go extremely, extremely deep into your area and that's where you find value.
One of the best determiners for success of founders is can they find the most boring thing humanly possible, interesting, over a multi-decade period.
How AI is going to disrupt software...These are generalist topics that I can probably find a thousand people that have interesting, compelling ideas and can go pretty deep on that. But you can't create value there.
You can create value if you're the number one person in the entire world at this little niche thing. And I think this niche thing can generate billions of dollars over time.
The problem is those places are really boring and requires you to read hundreds of thousands of pages that you cannot Gemini Deep Research your way through."
The biggest take off in AI isn’t a lobster, it is that millions of weekly phone calls to confirm trucking loads and appointments that used to be done by overseas call centers and are now done by AI callers.
It flipped in like 12 months.
The age of opaque lead sources and blind traffic in financial publishing is over.
Ad networks used to send leads and refused to provide any information about where the leads came from.
This allowed them to get away with shady business practices.
Media buyers are not putting up with that crap anymore.
What is the over/under on when they roll this back? I'm thinking 3 months.
Making people order on their phones during a movie is nonsensical, the current system works fine.
EXCLUSIVE: Alamo Drafthouse is getting rid of pen-and-paper food orders and is switching to mobile.
“We continue to evolve, as we must, to protect the Alamo movie-going experience as the entertainment landscape shifts around us."
https://t.co/Vln1Mm5eA6
Utah has become the first state to allow AI to renew medical prescriptions with no doctor involved. The company, Doctronic, also secured a malpractice insurance policy for their AI. Their data also shows that their system matches doctors treatment plans 99.2% of the time.
Big news for taxpayers: the Texas Bond, Tax, and Project Database is LIVE!
Now Texans can see exactly how much debt their local governments carry, what bonds voters approved, and how those dollars are being used.
BIG step for transparency and accountability! https://t.co/uPgg8qpBVg
Haus year in review. An exec summary of all the research that we published in 2025. We’ll be talking a lot more about what this means and how to use this to your advantage in 2026.
1. View based channels (YouTube, TikTok, & CTV) are systematically undervalued by click attribution because their impact is delayed and skews omnichannel.
- YouTube is a powerhouse: It drives 3.4x more incremental lift than Google Ads reports.
- Like YouTube, Tiktok builds your funnel and is best viewed as demand creation. It delivers 68% more lift in the post-campaign window as audiences take time to convert and drives 1.9x more value when you account for its impact on Amazon and retail sales.
- CTV has the most delayed return. During big moments like Cyber Week, Connected TV sees efficiency improve by 344% when you account for delayed conversions.
2. Meta is king but beware the automation trap
- Meta remains one of the most reliable sources of incremental performance. 77 of the top 100 highest lift Haus experiments ran on Meta.
- Automated tools are excellent at identifying intent, but they may prioritize capturing cheap, easy conversions over generating new ones.
- Meta Advantage+ vs. Manual: Advantage+ is great for speed, but it over-reports revenue by 12 percentage points compared to Manual campaigns. Manual campaigns frequently deliver higher incremental ROAS because they are less likely to cannibalize existing organic demand.
3. Branded Search & PMAX
Buying your own keywords is often not necessary; it is a defensive tactic that depends on competitive pressure.
- Competition dictates value: Branded search is highly incremental when you have 3+ competitors bidding on your terms. If competition is low, you may be paying for traffic you would get for free organically.
- Sales Channel Shift: Branded search tends to cannibalize your own Amazon sales, shifting that revenue to your DTC site.
- PMAX Brand Exclusions: If your goal is acquiring new customers, you should almost certainly exclude brand terms from Performance Max (PMAX); doing so lowered Customer Acquisition Cost (CAC) in 100% of tested cases. However, High-AOV brands (>$238) may need to keep brand terms included to give the algorithm enough signal to work.
4. Black Friday / Cyber Monday and Seasonality
- You may not see immediate payback on ads, especially in the lead up to holidays. That doesn’t mean media is not working.
- For Q4/Cyber Week, consumers add to cart early (early November) and buy later. Brands that ramped spend early saw 75% efficiency improvements when accounting for this lag. Apparel brands saw outsized impact.
- Don't cut a test too early: Cutting spend based on midpoint results runs the risk of "false negatives." Run longer tests and adding a Post-Treatment Window (PTW) that runs through your sale period to see the full payback.
Extremely interesting!
(source: https://t.co/KtqeWuhVD7)
Seems like more and more sites are blocking GPTBot, unclear to me what that will mean. Perhaps another long term advantage for GOOG.
for my ecomm friends, is there a trick to getting decent support from @klaviyo ? So far I've gotten wrong answers from the LLM and then wrong responses from support and tickets closed w/o resolving. Do you dial 1 for spanish or something?
If you have any doubts about who Prop Q benefits, the love Austin pac (pro prop q) campaign financial statement is out and it is chock full NGO'S trying to save their grifts.
Donation highlights: ECHO $25k
Foundation communities $25k
@KirkPWatson PAC $20k
Sunrise homeless center $5k
The Other ones Foundation $10k
@chitovela3 $1k #atxcouncil #txlege