In 2007 I was so broke I paid only half my electric bill and bet what was left on Google Ads.
A pathetic $60.
That's what I scraped together after the power company got theirs. I begged my girlfriend (now my wife Chaunna) to let me risk it, and I ran it at $2 a day. Talk about meager lol.
You know what that $2 a day bought? Keywords. Hand-picked, one at a time, by me.
That was the JOB. You found the exact phrases buyers typed, you bid on them, you wrote the ad that matched, and if you were good at it, you ate dinner that night. I got good at it. Real good.
Good enough to build AdSkills (7 figure training company) on it, trained 14,000+ students with it, and speak on the biggest stages all over the world.
But everything has changed...
This May, at Google Marketing Live, Google's VP of ads described the new reality of Search in five words: "You can't choose keywords anymore."
🤮🤮🤮
20 years of craft, erased in a sentence.
I won't pretend that didn't sting. The platforms took the steering wheel and handed us a suggestion box.
But I've survived every platform shift since 2005, and they all obey the same law: when the machine takes a job, the money moves to whoever feeds the machine best. The operators winning right now figured out what the machine EATS.
It eats signals. And the signals are still 100% yours.
So today you get the full damage report on both platforms, receipts on whether their machines actually work (I brought both sides of that argument), and the six levers you still own.
This is the free Monday edition, and it's the "why" behind everything we're building on Thursdays this month.
What Google Took
AI Max is now the center of Search. It's the campaign setting that lets Google's AI decide which searches your ads show up on, reaching way past your keyword list to anything it believes has the same intent.
Google's own claim is about 7% more conversions at similar cost.
And the old way is being retired, not just demoted.
Dynamic Search Ads (the campaigns where Google matched your website pages to searches) are being folded INTO AI Max. ðŸ˜
Google already slipped the deadline once: forced migration was supposed to be September of this year, and their developer blog moved it to February 2027 (one carve-out: campaigns running auto-created assets with campaign-level broad match still convert this September).
Most articles you'll read still print the old date. Check the primary source, always.
Google shipped an agent. It's called Ask Advisor, a Gemini-powered assistant that spans Google Ads, Analytics, and Merchant Center with a shared memory, so you can interrogate your account in plain English. Google watched everyone build agents on top of their platform and decided to build the agent themselves.
But if you've ever tested Gemini chat you know we are in trouble! That thing is as smart as a box of rocks. Great for images and coding, horrible at empathy and reasoning.
Our version (inside Build Club) uses Claude Opus 4.6 instead, and beats other chatbots in head-to-head comparisons question by question.
Bidding got more autonomous. Journey-aware bidding (in beta) learns from your whole lead-to-sale funnel, phone calls and form fills included, instead of stopping at the click. Smart Bidding Exploration stretched from Search into PMax and Shopping, with Google claiming 27% more unique converting users for the Search campaigns that use it.
And after years of advertisers screaming about lack of controls, Performance Max spent the last two years handing back controls... campaign-level negative keywords, channel-level reporting, brand exclusions, and real per-asset performance numbers to go with the vague Low/Good/Best labels.
But look at that list again. Every "new control" is a way to feed the AI or fence the AI. Not one of them hands back the wheel.
What Meta Took
Meta rebuilt the engine under the hood, and most advertisers never felt the swap.
The new retrieval engine is called Andromeda. It's the system that decides which ads even get CONSIDERED for the auction, and Meta's engineers claim it expanded model complexity 10,000x.
10,000x complexity added 🤯 by their own engineers!!!
On top of it sits GEM, a ranking brain Meta says is 4x more efficient at squeezing performance out of the same ads.
Their numbers, their test. Same caveat I gave Anthropic's 90.2% last week. But the practical effect is one nobody on either side disputes: on the new engine, your targeting inputs matter less and your CREATIVE matters more.
The machine finds the audience now. Not your own targeting.
Your job is making sure your ad content finds the audiences, with their AI.
(That's why Part 2 of our department writes ads, and why next Monday I'm testing every image model I can reach. The month was built in this order on purpose.)
Meta also quietly welded the escape hatch shut. As of Marketing API v25 this February, you can no longer create or update Advantage+ Shopping or App campaigns through the API. Meta's name for this is "Automation Unification," which is a gorgeous piece of corporate poetry for "the automated version is now the only version."
And they bought themselves an agent. Meta paid a reported $2 billion for a startup called Manus in December and had it working inside Ads Manager by February: conversational report building, audience research, campaign analysis. (China blocked that deal in April and forced Meta to unwind it in June.)
Hmmm... agents are good? Who'd a thunk???
Praise God for my gift of being early on things.
Meanwhile, the scoreboard from their side of the table: Q2 ad revenue of $59.363 billion, up 27% year over year. Impressions up 14%. Average price per ad up 12%.
You're paying 12% more per ad WHILE they remove your controls. It's a new world people.
The Receipts, Both Directions
Google's showcase numbers sparkle. Lufthansa credits AI Max with 24% better return on ad spend. IKEA claims 65% more non-brand clicks at a 28% lift in incremental ROAS.
Now the other side of the ledger.
An analyst named Mike Ryan at Smarter Ecommerce (an ads-automation firm, independent of Google, not of the industry) studied 250+ retail campaigns running AI Max.
The data is from late 2025, so this is the early-days report card, and I'm dating it for you because vendors never date anything. The median campaign: revenue up 13%... and cost per acquisition up 16%.
The full spread ran from +42% ROAS on the happy end to 35% WORSE on the ugly end.
His summary line: "Turning on AI Max is essentially a coin toss." YAY gambling!
The same study found up to 63% of AI Max's new matches simply recycled keywords the advertiser already owned, and caught one account where AI Max spent 69% of its Search impressions on COMPETITOR brand terms.
All of this buried in search term reports that had grown too large for a human to audit.
A report too big for the person paying the bill to read?? Who exactly does that arrangement serve???
Then there's the Search Partner Network, the long tail of non-Google sites where your Search ads can also run. PPC Land reports partner-network return running 37% below Google Search proper, and the Ryan study documented one campaign throwing 500,000 monthly impressions at partner sites that converted at 0.07%, against 3.04% on actual Google Search.
In July, Google started a limited alpha letting SOME PMax advertisers finally switch Search Partners and Display off. Ad buyer David Dweck called that inventory "remnant (aka trash) supply that Performance Max advertisers had been compelled to accept."
Compelled. To accept. For years. That should be criminal!
Meta's automation keeps its own rap sheet. One exec told Digiday in May about settings flipping themselves on mid-campaign: "we would turn on a campaign, we would never have Advantage+ Audience on, and then suddenly, two weeks into our campaign, Audience is on, and we've reached 15 million people."
Another exec in the same piece, on a different setting Meta flipped: "They don't send you an email or anything."
PPC Land quoted an early Manus tester catching it invent phone-call attribution: it "thinks that people are calling us from the ad, even though there's no phone number in the ad."
And back in 2024, advertisers reported Advantage+ Shopping inflating costs as much as 10x and draining full daily budgets in hours. Meta said it fixed the technical issue. The complaints outlived the fix.
Here's my honest read after 20 years of watching the ad industry under a microscope...
The automation isn't a scam, it's a tax. It's a powerful employee with zero supervision instinct, on the payroll of a company whose earnings grow when your price per ad does.
You wouldn't let that employee run unreviewed, and you can't fire him. So you manage him.
That's the whole game now. Babysitting robots.
Play the game well, make millions. Ignore the game, better odds at Vegas.
The Suggestion Box
Search Engine Land put the 2026 rule in one sentence: automation "runs on signals, not settings." And their warning belongs on your office wall: feed it poor or misleading data, and it will "efficiently automate failure."
Read that again...
IF you feed the ad platforms poor or misleading data then it will EFFICIENTLY AUTOMATE FAILURE.
The machine is a furnace. It doesn't inspect what you shovel in, it just burns it. Shovel clean fuel, it heats up your bank account balance. Shovel garbage, it burns your budget with exactly the same enthusiasm.
Here are the six levers you still have control of...
1. Conversion signal quality. The machine optimizes for whatever you count as a win. Count newsletter signups in the same bucket as $5,000 sales and it will happily flood you with the cheap stuff, then send Google a report card that says it's crushing it. Strip the soft conversions out of your primary actions. Count money.
2. Audience data freshness. Your customer lists and CRM uploads are the machine's seed corn. A list you uploaded 14 months ago teaches it to find people who bought from who you USED to be.
3. Campaign structure. Separate brand from non-brand, and high-ticket from low. Blended campaigns let the AI take credit for buyers who were already coming, and the coin-toss study showed you exactly how it happens: expanded matching quietly eats your own branded searches, then calls it growth.
4. Creative direction. On Meta, after Andromeda and GEM, this is close to the only real lever left. The machine needs volume and variety to explore with, and most accounts starve it. Next Monday's issue makes this one concrete, with prices.
5. Offline conversions and value-based bidding. If the sale closes on a phone call or a contract three weeks later, feed that back in. Journey-aware bidding was literally built to eat this data, and almost nobody serves it.
6. Drift detection on a 5-to-10-day window. Single days are noise (Thursday's issue taught you the Wet Paint Rule). But weekly windows catch the slow bleeds: CPA creeping, partner-network share growing, spend drifting onto competitor terms. The Daily Ad Brief Agent we built Thursday exists for exactly this job. This is the lever that guards the other five.
And so you know what "normal" looks like while you're pulling those levers: WordStream's 2026 study of 13,474 US search campaigns puts the medians at 6.64% click-through, $5.42 per click, 8.18% conversion rate, and $66.69 per lead.
Very beatable numbers.
First cost-per-lead decline in five years, which is the good news. The decade view is the real story though: average CPC was $2.32 in 2016 and it's $5.42 now. The auction got 2.3x more expensive in ten years.
I called this in 2009!
Sloppy signals were affordable in 2016. They are not affordable at 2026 prices.
Your Suggestion Box Audit
One hour this week. Go lever by lever:
- Open your conversion actions and demote anything that isn't money to secondary
- Find every campaign blending brand with non-brand, and split them
- Export a search term report and hunt for competitor-term spend you didn't choose
- Check PMax channel-level reporting for your Search Partner and Display share (and if Google offers you the opt-out alpha, take it)
- Screenshot your current Meta Advantage+ toggles, so you'll KNOW if one flips itself on
- Re-upload fresh customer lists on both platforms
- Set up offline conversion import if any part of your sale closes off-platform
- Count your live ad variants on Meta (under 5 and you're starving the machine)
- Put a weekly 7-day-vs-prior-7-day review on the calendar, or let Thursday's brief agent run it for you
- Compare your account to the medians above and mark the biggest gap. That's September's project.
I've made money in every version of this game since 2005, and the version where the machine does the grunt work while I control what it eats might be the best one yet. Not because it's easier... it's not! But because smart men and women will now easily beat the lazy ones.
Your competition doesn't want to learn this.
Heck, let's be honest, you barely made it through this article yourself 🤣
The good news is, the rest of August is about making sure your machine eats gourmet signals.
Sources
- PPC Land, Google Marketing Live 2026 coverage (May 20, 2026): https://ppc.land/google-marketing-live-2026-every-announcement-that-actually-matters/
- Google Ads Developer Blog, DSA automigration timeline (Jun 11, 2026): https://ads-developers.googleblog.com/2026/06/dynamic-search-ads-dsa-automigration.html
- Google, the DSA upgrade to AI Max (home of the +7% claim): https://blog.google/products/ads-commerce/dsa-upgrade-to-ai-max-2026/
- Google, bidding and budgeting announcements (May 7, 2026): https://blog.google/products/ads-commerce/bidding-budgeting-google-marketing-live-2026/
- Search Engine Land on the Smarter Ecommerce AI Max study (data from late 2025): https://searchengineland.com/google-ai-max-revenue-higher-cpa-study-470928
- PPC Land, PMax Search Partners opt-out alpha (Jul 28, 2026): https://ppc.land/google-lets-some-pmax-advertisers-switch-off-search-partners-and-display/
- Meta Engineering, Andromeda (Dec 2, 2024): https://engineering.fb.com/2024/12/02/production-engineering/meta-andromeda-advantage-automation-next-gen-personalized-ads-retrieval-engine/
- Search Engine Land, Meta's Andromeda + GEM system (Jan 28, 2026): https://searchengineland.com/meta-ai-driven-advertising-system-andromeda-gem-468020
- Meta developer blog, Marketing API v25 (Feb 18, 2026): https://developers.facebook.com/blog/post/2026/02/18/introducing-graph-api-v25-and-marketing-api-v25/
- CNBC, Meta acquires Manus (Dec 30, 2025): https://www.cnbc.com/2025/12/30/meta-acquires-singapore-ai-agent-firm-manus-china-butterfly-effect-monicai.html
- NPR, China blocks Meta's acquisition of Manus (Apr 27, 2026): https://www.npr.org/2026/04/27/g-s1-118892/china-blocks-meta-from-acquiring-ai-startup-manus
- PPC Land, Manus inside Ads Manager (Feb 21, 2026): https://ppc.land/manus-ai-lands-inside-meta-ads-manager-changing-how-advertisers-work/
- Meta Q2 2026 results (Jul 29, 2026): https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Second-Quarter-2026-Results/default.aspx
- Digiday, agency execs on platform automation (May 11, 2026): https://digiday.com/marketing/google-doesnt-care-that-its-terrible-brand-agency-execs-air-frustrations-with-the-trade-desk-googles-performance-max-metas-advantage/
- eMarketer, Advantage+ budget complaints (Apr 30, 2024): https://www.emarketer.com/content/meta-faces-backlash-automated-ad-system-drains-budgets-with-little-payoff
- Search Engine Land, "everything is a signal in 2026" (Feb 4, 2026): https://searchengineland.com/in-google-ads-automation-everything-is-a-signal-in-2026-468218
- WordStream 2026 Google Ads benchmarks (May 19, 2026): https://www.wordstream.com/blog/2026-google-ads-benchmarks
Continue with Build Notes+
This Monday deep dive is free in full. For complete Thursday builds and supporting files, join Build Notes+.