In 2026 Google Ads crossed a line: AI Max, the agentic layer that expands targeting, rewrites assets and sets bids with minimal input, moved from Search into standard Shopping campaigns on 30 April, and Google began automatically upgrading eligible campaigns — those already using automatically created assets and broad match first. Google reports an average 7% lift in conversions from the upgrade. At the same time, Shopping ads started serving inside AI Mode from existing Shopping and Performance Max campaigns, which means your ads now appear inside conversational answers without a new campaign type.
Meta’s Advantage+ runs the same way, and a wave of standalone agents now handle targeting, bidding, creative and budget pacing across both platforms. Agentic advertising is no longer a forecast. The question for advertisers is not whether to automate but which decisions to keep.
What AI Max actually does
- Search term matching: extends keywords to related queries, including ones your account has never seen.
- Asset generation: writes headlines and descriptions from landing pages and existing assets.
- Final URL expansion: sends traffic to pages it judges more relevant than the one you set.
- Bidding: optimises toward the conversion goal with less room for manual CPC.
Each of these improves reach and usually volume. Each also removes a control point. The lift Google reports is an average; accounts with weak conversion tracking, mixed-value goals or thin landing pages frequently see cost rise faster than conversions.
The agentic ad-ops operating model
Automate the loop, own the inputs
The most reliable results come from teams that let the platform run the auction loop but take unusual care with inputs: conversion values that reflect margin, offline conversion imports, negative keyword lists and brand exclusions, asset libraries with real product claims, and landing pages that match the queries AI Max will expand into.
Set guardrails in writing
- Brand safety: exclude competitor and off-brand queries; review search term reports weekly even when the platform discourages it.
- URL discipline: turn off final URL expansion for regulated or pricing pages.
- Budget pacing: cap daily spend and set target ROAS floors per product group.
- Creative governance: approve generated assets before they run in regulated categories.
Use agents between the platforms
Standalone agents are most useful where the platforms do not see: reallocating budget between Google and Meta based on blended ROAS, pausing products out of stock across all campaigns, and generating creative variants from product data. Keep them read-only until their recommendations have been validated for a month.
Measurement that survives automation
Platform-reported conversions are the score the platform gives itself. Pair them with an independent view: marketing mix modelling for budget decisions, geo-holdout tests for incrementality, and first-party revenue by channel. When AI Max claims a 7% lift, the holdout tells you how much of it was incremental and how much was cannibalised from brand or organic.
Ads inside AI Mode and AI Overviews
Placements inside conversational answers are served from existing campaigns, so there is no separate opt-in — and no separate reporting yet. Expect lower click volume and higher intent. Review assets with the conversational context in mind: a headline that reads naturally inside a sentence performs better than a keyword string.
A 30-day plan for an account being auto-upgraded
- Week 1: audit conversion tracking and assign values; import offline conversions.
- Week 2: build exclusion lists and decide URL expansion per campaign.
- Week 3: refresh assets with specific, verifiable claims; align landing pages.
- Week 4: set up a holdout test and a blended ROAS dashboard before judging the lift.
Automation is winning because it is good. The advertisers who keep winning are the ones who feed it better data than their competitors and check its homework.
Keep reading: Google Advertising · Advertising