Give The Algorithm Room To Roam
Google expands Smart Bidding Exploration to Performance Max and Shopping, and lets advertisers set how far the AI can stray from target
By Alexander Marshall
Search advertising's control surface has been shrinking for years: match types loosened, keywords automated, Performance Max hiding the levers. Google's June 2026 bidding update goes a step further. By expanding Smart Bidding Exploration into Performance Max and Shopping, and letting advertisers set a tolerance for how far the AI can stray from their ROAS target, Google is asking marketers to stop dictating the bid and start describing the outcome.
WHAT ACTUALLY CHANGED
On June 15, 2026, Google rolled out a set of bidding changes across its platforms, the headline being that Smart Bidding Exploration expands to Performance Max and Shopping campaigns, having previously been available only in Search.
The mechanism is a deliberate loosening of control. Smart Bidding Exploration is built to help advertisers discover new search categories and find new customer segments, using Google's AI to reach and convert demand that a tightly-targeted, keyword-bound setup would never have surfaced. A new ROAS tolerance setting is the key lever. It defines the percentage by which a target ROAS can be lowered, deliberately giving the system room to reach users it could not before, trading a measure of on-paper efficiency for incremental reach.
The predictability promise cuts against the loosening, which is worth noting. Google is at once giving the AI more room to explore and promising steadier delivery against fixed budgets, a pairing that only reconciles if its models have genuinely improved, so advertisers should verify steadier performance shows up in their own numbers rather than taking the claim on faith. Google also introduced Promotion Mode, now in beta. It lets marketers temporarily adjust bidding strategies during high-traffic seasonal events or sales, so a campaign can lean in hard during a peak moment and settle back to baseline afterward.
The exploration expansion is the structurally bigger of the two changes. Letting Smart Bidding Exploration loose in Performance Max and Shopping, where product feeds and multi-channel inventory live, means the AI can now discover demand across the formats that carry most retail spend, not just classic Search text ads. Bidding target optimization is being updated as well. Starting August 17 and rolling out over the following weeks, it is designed to deliver more predictable performance for campaigns that run on fixed spending limits.
The phased timeline gives advertisers a window to prepare. With bidding target optimization arriving in mid-August and review notifications appearing from early July, Google is signaling the changes weeks ahead, which gives teams time to set tolerances, brief clients and decide where to open exploration before the system starts behaving differently.
Google is shepherding advertisers through the shift. From July 6, notifications in Google Ads will let advertisers review historical campaign performance and quickly apply the updates, with Ads product liaison Ginny Marvin detailing the changes on LinkedIn and the Accelerate with Google blog. The stated rationale is that AI-driven insights are making search performance more predictable as people move beyond simple keywords toward complex, highly specific queries.
The bundle of changes shares one logic. Smart Bidding Exploration, ROAS tolerance and Promotion Mode all hand the system more room to act, which is consistent with Google's broader message that, in an AI-shaped search landscape, advertisers get further by guiding automation than by manually constraining it.
It also lands in the same window as Google's AI Mode ad formats. Having introduced Gemini-powered conversational ad experiences at its May Marketing Live, Google is now retooling the bidding underneath them, so the creative surface and the buying engine are being rebuilt for the conversational-search era at the same time.
WHAT IT MEANS FOR ADVERTISERS AND AGENCIES
The unit of control is moving from the keyword to the goal. When the system explores categories and segments on the advertiser's behalf, the job stops being to enumerate the right keywords and becomes to define the right outcome and how much latitude the AI has to chase it, a genuinely different way of running an account. ROAS tolerance is an explicit trade of efficiency for reach, and it should be treated as one. Lowering the target ROAS ceiling lets the system find incremental customers, but it also means a reported ROAS that looks softer on the surface, so the value only shows up if advertisers measure the incremental sales the exploration actually delivers.
Expanding exploration to Performance Max and Shopping is the consequential part. These are where most retailers' budgets sit, so giving the AI room to roam across shopping and multi-channel inventory materially widens where and how a brand's products can surface, for better reach and for less predictability. Promotion Mode formalizes the seasonal lean-in. Instead of manually rebuilding bids before a sale and unwinding them after, advertisers get a temporary, purpose-built setting, which is genuinely useful provided it is used for real peaks rather than left on as a permanent crutch.
Predictable performance for fixed budgets matters more than it sounds. Many advertisers operate hard monthly caps, and bidding that respects those limits while still optimizing removes a real source of overspend anxiety, which makes the automation easier to trust for budget-constrained teams. The shift also rewards a different kind of expertise. As the levers move from manual bid management to goal-setting, audience signal and measurement, the value an agency adds moves up the stack, toward defining the right outcomes, feeding the system quality data, and proving incremental lift rather than tuning keyword bids by hand.
The complex-query backdrop is the real driver, not a footnote. As AI Mode and conversational search push people toward longer, more specific queries, fixed keyword lists simply cannot keep up, so exploration is Google's answer to inventory that keyword targeting can no longer fully address. There is also a control-and-transparency tension worth naming. As Google folds more of the buying decision into its own AI and softens the targets advertisers set, it becomes harder to see exactly why the system spent where it did, which is why advertisers who keep their own measurement and first-party data are better placed to hold the automation to account.
The caution is the familiar one with looser controls. More AI latitude plus softer efficiency targets makes it easier to spend into low-quality demand if nobody is watching, so the discipline of independent measurement and clear guardrails matters more, not less, as the controls relax.
FROM KEYWORD TO GOAL
Google's June update trades fixed targets for explored ones. The advertiser's job is no longer to set the bid, but to define the outcome and the tolerance, then verify the lift.
WHAT WE WOULD DO ABOUT IT
Set the goal and the tolerance deliberately. We treat ROAS tolerance as a real strategic choice, opening it where a client genuinely wants incremental growth and keeping it tight where efficiency is the priority, rather than accepting whatever default the platform suggests. Use Promotion Mode for genuine peaks only. We deploy Promotion Mode for real seasonal moments and sales, then return to baseline, treating it as a precision tool rather than an always-on setting that quietly inflates spend.
Anchor every loosened bid in incrementality. We judge exploration on incremental sales using incrementality testing and media-mix modeling, not on the platform's reported ROAS, so a softer target is justified by real new customers rather than reshuffled credit. Feed the machine first-party data and community signal. We give Google's AI strong first-party data and clear conversion signals so its exploration chases a client's real audience, anchoring even automated bidding in the community-first approach at the heart of CAC to Community.
FURTHER READING
MediaPost - Google Ads, Platforms See Major Bidding Changes
Google - See the latest Google Ads product announcements
Search Engine Land - Google tests new conversational ad formats in AI Mode and Search
