Your output scales with AI now. Your judgment still has to come from you. This issue covers three places where that plays out: taste, Reddit visibility, and cloud economics.
This week's edition covers:
- Kevin's Take: why taste beats the algorithm
- The Signal: ChatGPT is routing answers through Reddit, plus the cloud split reshaping your programmatic vendors
- Tools & Tactics: the content formats AI engines actually cite, plus why the fixed 60/30/10 PPC split stopped working
- Quick Links: dynamic take rates squeezing CPMs, 63% of media spend chasing short-term wins, Google's AI Brief coming for the keyword
Kevin's Take
Taste Beats the Algorithm
A year ago, a big positioning call meant weeks of research and a deck to defend it. Last week I made one in an afternoon. I came in with a strong opinion already formed and used the model to poke holes in it, rather than asking it what to think. Tony Fadell, the guy who built the iPod and the iPhone, put words to why that works in a recent interview with Lenny Rachitsky: opinion has to come before data, and taste and judgment are still what separate a great product from a forgettable one. This may be even more important in marketing than it is in product.
The easy thing to do with AI right now is to flatten every decision into a probability. Run it through the model, look at the score, ship the winner. The trouble is that when everyone runs the same prompts against the same models trained on the same web, you get the same campaign as your competitor, the same landing page, the same email sequence, the same boring positioning. The model is an average of everything that has worked before, and an average, by definition, isn't distinctive. You have to come to it with a take, not ask it for one.
One thing I've found helpful is to stop asking AI what we should do, and instead tell it what I believe and have it argue with me. Here's my position, now find where it's weak, where I'm wrong, what I'm not seeing. The taste, the conviction, the specific opinion about your customer, that still comes from us. The model pushes on it, it doesn't author it. If you lead marketing, what's going to set you apart over the next couple of years is having a real opinion about what your brand stands for and being willing to ship it before the dashboard agrees. Having the opinion in the first place is still on us, and I don't think you can prompt your way to that.
That's it for this week. Talk soon.
— Kevin Kerner, CEO, Mighty & True
The Signal
ChatGPT is quietly routing its answers through Reddit (1 min read)
New Profound data shows ChatGPT explicitly queries Reddit by name before generating responses, making Reddit a primary upstream source for AI answers. This follows similar findings on Yelp's dominance in local AI discovery. (via Foundation Team)
Why it matters: If your brand isn't showing up in relevant Reddit threads, you're invisible to the AI tools your buyers are now asking for recommendations.
The cloud hardware split reshaping your programmatic vendors (1 min read)
Independent ad tech vendors are splitting into two camps: those owning cloud infrastructure versus those renting from hyperscalers, with the divide shaping cost structures and competitive positioning in programmatic. The hardware question is becoming a strategic fault line for DSPs and SSPs.
Why it matters: If you're buying programmatic, your vendor's cloud economics will shape what they charge you and how fast they can ship features — worth asking on your next QBR.
Tools & Tactics
The content formats AI answer engines actually cite (1 min read)
Analysis of HubSpot's State of AEO 2026 report and Wix Studio's AI Search Lab research identifies which on-page content formats LLMs cite most frequently. The findings give marketers concrete formatting guidance for getting picked up by answer engines. (via HubSpot Staff)
Why it matters: If your content team is still optimizing for blue links, you're missing the formats that actually get cited in AI answers your buyers see.
Why fixed PPC budget splits don't hold up (1 min read)
Fixed PPC budget split ratios across funnel stages rarely hold up under real market conditions. The piece argues for continuous funnel health diagnostics and dynamic spend reallocation instead of static percentages.
Why it matters: If your team is still defending a 60/30/10 PPC split in your QBR, you're optimizing for a spreadsheet instead of your pipeline.
Quick Links
News & Trends
- The hidden programmatic fee shift squeezing your CPMs — Ad exchanges including Google and Index Exchange are quietly shifting to dynamic take rates, variable fees that adjust per auction rather than a fixed cut (via AdExchanger Guest Columnist)
- Gartner: 63% of media spend now chasing short-term wins — Gartner's latest marketing survey finds 62.6% of media spend now goes to awareness and conversion as CMOs chase short-term results amid the AI rush (via Sara Karlovitch)
- Google's AI Brief may finally replace the keyword list — Google is rolling out AI Brief, a structured input that lets advertisers describe intent and context to AI-driven campaigns instead of relying on keyword lists (via Frederick Vallaeys)
- Why Claude users are turning on Anthropic — Anthropic has gone from beloved AI darling to community backlash target in under six months, with Claude users increasingly frustrated by performance and policy changes (via Daniel Nest)
Work with us — Growth Consulting
Working through a strategic pivot? Mighty and True Growth Consulting partners with tech CMOs and their teams on positioning, GTM rewires, and AI operating model work focused on outcomes, not vanity metrics. Bring us the hard question.
Join the group we're building
We are building something for marketing leaders who are tired of vendor-sponsored best practices content. A small group. Real conversations. No pitching. If that sounds like your kind of thing, reply and I will share the details.