What did Meta and Google actually remove?
Over 2025 and into 2026, both platforms pushed hard toward automation. On Meta, Advantage+ became the default way to run the main campaign types, and manual controls over placements, audiences and creative were folded into single automated settings. Manual audience targeting still exists, but it is increasingly treated as a hint the system can override rather than a hard boundary you set.
Google went further on search. AI Max, the successor to Dynamic Search Ads, drops keyword targeting entirely and uses Gemini to read your landing pages and match them to what people are actually searching. The keyword, the unit paid search was built on for two decades, is no longer the thing you point the campaign at.
Meta is even testing end-to-end campaigns where you hand over a URL and a budget and the AI generates the creative, picks the audience, chooses placements and manages bids. That is the direction of travel: you set the goal, the machine does the mechanics.
- Meta Advantage+ is now the default for Sales, App and Leads campaigns
- Manual audiences on Meta are treated as a signal the system can override, not a hard rule
- Google AI Max removes keyword targeting and matches intent from your landing pages
- Meta is testing URL-plus-budget campaigns where AI handles creative, targeting, placement and bids
Were keywords and manual audiences ever really the moat?
It is worth being honest about what we are actually losing. For years, campaign skill was partly about knowing which audiences to stack and which keywords to bid on. That knowledge felt like the job. But think about whether it was ever defensible. Your competitor could target the same interests and bid on the same keywords. Anyone could.
The targeting settings were never a secret. The platforms had the data all along, and automation is simply them saying they can now read intent better than a human clicking checkboxes. On a lot of accounts, that is true. The machine sees signals a media buyer never could, and it reacts faster.
So the loss stings less once you accept that manual targeting was table stakes, not an edge. If the thing being automated away was something any competitor could copy, it was never what set you apart. That frees you to focus on what actually did.
What are the four inputs that now decide your ROAS?
When the platform runs targeting, placement and bidding, your results are decided by the quality of what you feed it. There are four inputs, and they are where the real work now lives. Get these right and Advantage+ or AI Max performs. Get them wrong and no amount of manual tweaking would have saved you anyway.
The uncomfortable truth is that these four are harder than picking audiences. You cannot fake your way through them with platform knowledge. They take craft, data and judgement, which is exactly why they are where a good marketer now earns their fee.
- Creative: with targeting automated, your ad is the targeting. The image, hook and message decide who the system shows it to. Volume and variety of creative is now the biggest lever you have.
- First-party data: the AI is only as smart as the signals you send it. Clean Pixel, Conversions API and accurate conversion values tell it who a good customer looks like. Feed it junk and it optimises toward junk.
- Offer: no algorithm fixes a weak offer. What you are selling and at what price still decides whether the click becomes a sale.
- Landing page: the AI delivers the click; the page converts it. Speed, clarity and message-match between ad and page decide the return, and Google's AI Max literally reads the page to match intent.
Where does full automation still burn Indian ad budgets?
Automation is not magic, and handing it a URL and a budget with nothing behind them is a fast way to waste money. The AI needs signal to optimise. A new account with no conversion history, thin creative and broken tracking gives it nothing to learn from, so it spends your budget exploring while it guesses.
I see this most with smaller Indian advertisers who switch on a fully automated campaign expecting it to sort everything out. Without a feed of clean conversion data and a few strong creatives, the system optimises toward the cheapest, easiest action, which is often a low-quality lead or a cart that never checks out. The dashboard looks busy. The sales do not come.
Full auto rewards accounts that already have good data and good creative, and punishes accounts that are hoping automation will paper over the lack of both. That is the trap: it works best exactly where you least needed the help.
- New accounts with no conversion history give the AI nothing to optimise toward
- Thin creative starves an automated campaign of the signal it needs to target well
- Broken tracking teaches the system to chase the wrong action, like junk leads
- Full automation amplifies good inputs and exposes weak ones, it does not fix them
What does a good media buyer actually do in 2026?
The job moved up a level. Less time in the settings, more time on the inputs that feed them. A good performance marketer now spends their hours producing and testing creative, making sure the tracking and first-party data are clean, sharpening the offer, and building landing pages that convert the traffic the AI sends.
That is the work I do across accounts now: I build the site, design and cut the creative, run the ads against it, and keep the data flowing back so the platform learns from the truth. Fewer handoffs, faster iteration, and the ad and the landing page finally saying the same thing. The automation handles the mechanics I used to do by hand. I handle the things automation cannot.
If anything, the shift makes a marketer who can do all of that more valuable, not less. The person who only knew how to pick audiences is being automated. The person who can feed the machine well is who the machine now depends on.