Meta Advantage+ and Google Performance Max are taking over targeting and bidding. Here is what still belongs to you, and how to test automated campaigns honestly.
Ad platforms are removing the levers that performance marketers used to pull. Audience selection, placements and bidding are increasingly handled by the platform's own models. That changes the job, but it does not remove it.
The direction of travel
Meta is the clearest example. A Wall Street Journal report, summarised by Social Samosa, said the company aims to offer tools by 2026 where an advertiser provides a product image or website and a budget goal, and Meta's AI generates the creative, picks the audience and suggests how to spend. Meta had already been narrowing manual options. eMarketer reported that it was phasing out manual targeting from some Advantage+ catalog campaigns in favour of automated targeting, while still allowing custom audiences and retargeting.
Google has moved the same way with Performance Max, which assembles ads and bids across Search, YouTube, Display and other surfaces from the assets and signals you provide. Both platforms trade control and visibility for scale, and both are free to use inside your ad spend.
Dentsu's global report goes further, projecting that algorithms will drive more than three quarters of ad spend in 2026. Treat the exact figure with care, since it is a forecast, but the direction matches what we see in accounts.
What you still control
If the platform picks the audience and the placement, four things are still yours. They decide who wins.
1. Creative and offer. The model can only rearrange what it is given. Distinct angles, strong hooks, real product proof and a clear offer are now the main lever. There is also a known risk: when everyone uses the same generation tools, ads start to look alike and fatigue sets in faster. A deliberate creative system, with fresh concepts shipped on a rhythm, protects performance.
2. Data signals. Automated bidding learns from the conversions you report. Send clean, deduplicated events through server-side tracking and the Conversions API or enhanced conversions, and pass on value, not just a count. Poor signals produce confident, wrong optimisation.
3. Measurement. Platform dashboards grade their own homework. Use holdout tests, geo experiments or marketing mix models to check what the spend caused. Compare blended cost per acquisition and revenue, not only in-platform ROAS.
4. Guardrails. Set brand exclusions, negative lists, budget caps and review rules. Performance Max, for example, can cannibalise branded search unless you exclude it. Automation does exactly what its goal says, so write the goal carefully.
A practical way to test automation
1. Run the automated campaign alongside your current structure, with an agreed budget split. 2. Keep creative and offers equal across both so you test the system, not the ads. 3. Judge on business results over a full buying cycle, not three days of platform data. 4. Review search terms, placements and asset-level reporting weekly for waste or brand risk. 5. Scale what holds up against a control and switch off what does not.
Where AI helps beyond the platforms
Outside the ad platforms, AI is most useful where output is easy to verify: producing creative variants for review, building audience and keyword research, writing first-draft reports, spotting anomalies in spend, and answering leads quickly on WhatsApp or email. The principle is the same as in media buying. Let machines do volume and speed, and keep people on judgement.
What this means for Indian advertisers
Industry reports show Indian digital advertising growing quickly, with online video and programmatic taking a larger share of spend. As more of that spend flows through platform automation, the four controls above become more valuable, not less. Brands that get creative, signals and measurement right will beat rivals who simply switch the automation on.
If you want help with this, see our AI advertising service. If your agency's pitch is mainly that it knows the platform's settings, the settings are being automated. Look for one that can show how it builds creative, protects data quality and proves incremental results.