Smart Bidding Exploration trades a slice of efficiency for new queries.
Google announced Smart Bidding Exploration on May 20 as part of the GML 2026 bidding and budgeting drop. The pitch: opt in to a small reduction in your effective ROAS target, and Smart Bidding will bid more aggressively on queries it would otherwise have skipped because they did not meet the target 1. Google reports an average 27% increase in unique converting users for Search campaigns using it 1.
The setting is live in Search, AI Max, Performance Max, and Shopping as of this week. Operators who turn it on without thinking will see their search-terms report start filling with queries that look strange next to their account history. Operators who read those queries closely will get the lift Google is advertising. Operators who do not will get the queries and pay for them.
The work this week is reading the search-terms report differently.
How the setting works
Smart Bidding Exploration adds a configurable ROAS tolerance to your existing Target ROAS bid strategy 2. If your target ROAS is 200% and you pick a 10% tolerance, the campaign’s effective target ROAS drops to 180%. The system uses that extra room to bid on queries that historically would not have cleared the 200% bar but might still convert at a rate that earns the spend.
Tolerance is set at the campaign level. Range is 0% (off) to roughly 50%, though Google warns that going past 20% widens the query distribution faster than most accounts can absorb.
The mental model: it is the inverse of a strict target. You are telling Smart Bidding, “I am willing to accept this much lower average ROAS in exchange for finding queries you would have refused to bid on at the strict number.” Whether that trade earns out depends on the quality of the new queries.
What shows up in the search-terms report
Two things change. First, query distribution gets wider. Campaigns that were converging on a tight set of branded and high-intent terms will start showing up on broader category queries, informational queries, and queries with lower historical CTR. The shape of the search-terms report looks more like the early days of a campaign than the steady state it had settled into.
Second, the average position and CPC on the new queries will be higher than the campaign’s old average, because the system is now willing to outbid on queries it previously lost. Reported CPA on the new queries will be higher than the campaign’s old average CPA, because the queries are by definition the ones Smart Bidding rejected at the strict target.
The math still works in aggregate as long as the new queries convert at a rate that pulls the effective ROAS to your tolerance-adjusted number, not below. The risk is that the new queries include enough waste that the aggregate dips below the tolerance, in which case Smart Bidding will pull back the next learning cycle and you will have paid for the exploration.
How to read the report this week
Three reads, in order.
The first read is the intent split. Open the search-terms report after seven days of running with exploration on. Tag each new query (queries that did not appear in the prior 30 days) as either: a closer-funnel query you missed, a broader-category query that was a near match, or a query with clearly mismatched intent. The first two are the lift Google advertised. The third is the waste you need to negative out.
The second read is the converter quality. For the new queries that did convert, pull them into your CRM (for lead-gen) or e-commerce platform (for ecom) and check downstream quality. Lead-gen accounts should look at qualified-rate by query; ecom accounts should look at first-order AOV and 60-day repeat rate. Smart Bidding Exploration earns its keep when the new queries produce buyers whose downstream value matches or beats the account average, not just when they produce form fills.
The third read is the cannibalization check. Some share of the new queries are users who would have found you on a different query (or organically) and bought anyway. Compare total campaign conversions with and without exploration over a two-week stretch, controlling for seasonality. If the lift in unique converters matches Google’s 27% claim but total conversions are up only 8 to 12%, you are paying for a lot of users who would have converted without the new spend. That is still net-positive in most cases, but it changes the math on whether the exploration spend is worth the cycles to manage.
Where to start tolerance
The right tolerance setting depends on the account’s starting ROAS headroom and the operator’s appetite for negative keyword management. Three starting points:
If the campaign’s reported ROAS over the trailing 90 days is more than 30% above target (you are over-converting at your set target), start at 15% tolerance. The campaign has room to explore without compromising margin.
If trailing ROAS is roughly at target, start at 10%. Watch the search-terms report twice a week for the first month.
If trailing ROAS is below target, do not turn exploration on. The campaign is already struggling at the current target; loosening the target will widen the loss before it widens the find rate.
These are starting points, not destinations. The right tolerance for a given account is whatever number produces aggregate ROAS at or above the original target after 30 days of run time, accounting for the negative keywords you add along the way.
What the setting reveals about your account
Smart Bidding Exploration is, among other things, a diagnostic. The queries it surfaces tell you what your campaign was missing under the stricter target.
If the new queries are mostly broader versions of queries you were already winning, the prior target was reasonable and the exploration is finding incremental volume.
If the new queries are mostly informational or research-stage queries that someone in your category would naturally search but that your campaign was rejecting, you have been under-bidding on upper-funnel intent and Smart Bidding Exploration is correcting it.
If the new queries are mostly off-topic or competitor-adjacent, your campaign structure is too broad and the exploration is exposing a weakness in keyword selection that exists with or without the new feature.
The third case is the one to act on. Tightening keyword selection inside Smart Bidding Exploration looks the same as tightening it in any campaign: identify the mismatch pattern, add the negatives, and let the system learn within the cleaner query pool.
What this does not change
The leverage in a Google Search account still sits in the same four places: offer, creative (RSA combinations and extension assets), landing page, and conversion event quality. Smart Bidding Exploration is a knob on the bidding strategy. Turning the knob does not fix a campaign with a weak offer or a landing page that converts at 1%. It does not turn a bad account into a good one. It can, on a good account, find 20 to 30% more converters at a slightly worse unit economics. On a bad account, it finds the same percentage worth of additional cost.
This is the same pattern as every other Google Ads release for the past three years. Each new feature gives the system more room to optimize. The optimization is only as good as the inputs. The accounts that win are still the accounts with the better offer, the better creative, the cleaner landing page, the more accurate conversion tracking. The feature does not change the order of operations covered in the pMax search terms piece. The new queries get read the same way as any other.
The setup task for today
Open one Search campaign with healthy trailing ROAS (above 30% over target for 90 days). Turn Smart Bidding Exploration on at 10% tolerance. Block out 30 minutes on the calendar for each of the next three Fridays to read the search-terms report and add negatives.
Pick one campaign, not the whole account. The feature is new, the system is volatile in the first two weeks of any tolerance change, and you want a controlled comparison against your other campaigns running at the strict target.
Three weeks of disciplined reading on one campaign will tell you whether the feature earns out for your category and your account structure. Then expand to the campaigns where the math supports it. Spend less, on fewer things, including on which new features you give your time to.