Smart Bidding Exploration is an opt-in setting in Google Ads that lets a Target ROAS (return on ad spend) bid strategy accept a somewhat lower return in exchange for bidding on searches it would normally pass over. It exists because automated bidding learns from history: queries that have converted before attract confident bids, while plausible but unproven queries are bid down or ignored. Exploration loosens that loop deliberately. Google's help centre describes it as a feature that "uses Google AI to bid on more high-performing searches within your existing keyword and product targeting". It changes how aggressively the bidder pursues traffic, not which traffic a campaign is eligible for.

How it works

The setting sits inside a campaign's bid strategy, or inside a portfolio bid strategy shared across campaigns. An advertiser running Target ROAS switches it on and sets a ROAS tolerance, the percentage by which the system may undershoot the stated target. Google's worked example is plain arithmetic: a 200% target with a 10% tolerance produces an "effective" target of 180%. Ad group-level ROAS targets are lowered by the same proportion, according to Google's eligibility documentation.

What the system does with that headroom is the distinguishing part. Lowering a ROAS target in the ordinary way raises bids broadly and fairly uniformly across every matched query. Exploration, according to Google, "selectively raises bids on specific high-converting traffic slices", concentrating the extra spend on query groups where its models predict conversions are plausible but where history is thin. Google uses the image of a pond and fishing gear: targeting defines the pond, and exploration is better gear for fishing in it.

The unit of measurement is the search category, a cluster of queries grouped by intent and meaning. Raw search term reports treat misspellings and plurals as separate entries; categories collapse them. Google's chosen success metric, traffic diversity, counts how many unique search categories generate impressions, clicks and conversions. Three campaign table columns report it: unique search categories with impressions, with clicks, and with conversions. The bid strategy report adds a light green timeline bar marking when exploration was active and an "Avg. target ROAS" figure showing the tolerance-adjusted target. Google says actual ROAS should land within roughly 15% of that effective target.

Eligibility is narrow. Google's documentation for Search lists four conditions:

  • Bid strategy: Target ROAS only. Maximize conversion value without a target is excluded, because it has no target to relax.
  • Budget: uncapped. If a campaign becomes budget-limited, the exploration mechanism switches itself off and an alert appears.
  • Targeting: broad, meaning broad match keywords, Dynamic Search Ads or AI Max for Search should carry most of the traffic.
  • History: an established campaign. Testing on a new campaign without conversion data is "strongly discouraged".

It works with shared budgets, manager-account portfolios, seasonality adjustments and Search Ads 360 campaigns using Google Ads portfolios, though multi-target portfolios do not get its benefits. Google advises against pairing it with CPC (cost per click) bid limits, and against raising the ROAS target at the same moment exploration is switched on, which would cancel the headroom.

A worked example shows the trade. A retailer bidding at a 400% Target ROAS that enables a 10% tolerance gives the bidder permission to operate at 360%. On EUR 10,000 of monthly spend, attributed value could fall from EUR 40,000 to EUR 36,000 if nothing else changed. The bet is that conversions from new categories raise total value, so that spend rises and the ratio dips while absolute returns still grow. Google's guidance asks advertisers to judge exactly that: whether added volume justifies "somewhat lower ROAS".

Google says there is no learning period, but also that the system may need one to two weeks to recognise valuable new traffic, partly because of conversion delay. Its evaluation advice is to exclude the first one to two conversion cycles, run for at least six weeks, and gather at least 50 conversions before drawing conclusions.

Origin and evolution

The feature has an awkward birth certificate. On September 30, 2024, PPC News Feed reported a beta called Bidding Exploration, shown at Google's AI Innovation Day event, which lowered a Search campaign's tROAS by up to 10% and refused to run on budget-limited campaigns. Google then presented Smart Bidding Exploration publicly at Google Marketing Live on May 21, 2025, in a blog post titled "Expand your universe of conversions with Smart Bidding Exploration". The post called it the largest bidding update in over a decade and cited Google internal data from March 11 to April 11, 2025: an average 18% rise in unique search query categories with conversions and a 19% rise in conversions. PPC Land's report on the event listed the feature alongside AI Max for Search with those figures.

Tooling followed. Google Ads API v22, released October 15, 2025, added date segmentation for exploration's query-cluster metrics, and Google Ads Editor 2.11, dated November 6, 2025 in PPC Land's timelines, described it as "a new opt-in feature for Search campaigns".

The 2026 phase widened scope. In a March 11, 2026 Ads Decoded episode, Google product manager Kristina Park called it "the biggest advancement in bidding in years", while colleague Carlo Buchmann said "it's still maintaining performance and making sure you deliver on your core traffic, but it also explores new areas to help find growth so that the system learns." On May 7, 2026, Josh Braverman, Group Product Manager at Google Ads, announced that exploration would extend to Performance Max with product feeds and standard Shopping, citing 27% more unique converting users for Search campaigns using it. On June 15, 2026, Google made it available in all languages for Performance Max campaigns without a product feed and opened a beta for feed-based Performance Max and Shopping.

Why it matters

For marketers, exploration addresses a structural complaint about automated bidding: it tends to keep buying what already worked. A campaign that wins "mortgage" queries but never bids seriously on "how to buy a home", to use Google's own example, may be missing earlier-stage demand. Exploration offers a sanctioned way to buy that demand without dismantling targets.

It also turns ROAS tolerance into a control in its own right. The same concept now runs through promotion mode, which schedules temporary tolerance changes for 3 to 14 days, according to Ads Liaison Ginny Marvin. And it arrives as Google repositions targets as the main efficiency lever: after August 17, 2026, targets govern spend in budget-capped campaigns, while exploration itself requires that budgets not be capped.

Limitations and disputes

Every performance figure comes from Google. The 18%, 19%, 27% and later percentages are platform averages from internal data, without published methodology or independent replication. A newer help-page figure, from January 1 to February 20, 2026, says campaigns allowed a 10% lower ROAS saw 13% more search categories with conversions and 10% more conversions than campaigns given a similar plain target reduction. That comparison is the more relevant one, and the gain is smaller.

The tolerance range is reported inconsistently. Google's eligibility documentation describes only an approximately 10% reduction, and its best-practice page says to "start with a 10% ROAS tolerance". Search Engine Journal's May 21, 2025 report put the flexibility at generally 10% to 30% below the usual threshold. The interface presents the control as a slider. Which values each account sees is not documented in one place.

Launch dating conflicts too. PPC Land articles from June 2026 say exploration has been available for Search "since 2024", reflecting the beta, while Google's own launch post dates from May 2025 and Editor 2.11 still called it "new" in November 2025.

The economic criticism is direct. Lower ROAS means more spend per unit of value, and Google collects the spend. Jonni Lomax of Lomax PPC said, according to Search Engine Land, that the optimistic reading is that "data quality is preserved", with the pessimistic view being that ROAS falls overall. Reporting remains aggregate: category counts rather than query-level detail, with some terms withheld for privacy thresholds. Change history support for Performance Max was still "planned for a future release" in Google's measurement documentation.

Disambiguation

Lowering the ROAS target. A plain reduction raises bids across all matched queries. Exploration applies a similar headroom selectively, according to Google, and adds diversity reporting.

AI Max and broad match. These are targeting tools that expand which queries a campaign can match. Exploration changes bidding within that pool. Google treats them as complements and recommends broad targeting as a precondition.

Learning period. The learning period is the calibration phase after a strategy change. Exploration is an ongoing setting, and Google says it triggers no learning period.

Seasonality adjustments and promotion mode. Seasonality adjustments tell the bidder to expect a conversion-rate shift for a few days. Promotion mode schedules temporary tolerance and budget changes. Exploration is a persistent tolerance aimed at new categories, and it is compatible with seasonality adjustments.

Recent developments

The June 15, 2026 package placed exploration beside promotion mode and the August 17 target change, with Jyll Saskin Gales calling the Performance Max expansion "a welcome update". PPC Land's explainer on Smart Bidding, published October 17, 2026, places exploration within the wider family of auction-time strategies. As of October 2026, no general availability announcement for the feed-based Performance Max and Shopping version had been found, and Google's documentation still recommends enabling exploration at least six weeks before major holiday sales.

Timeline

  • September 30, 2024: Bidding Exploration beta for Search tROAS campaigns reported after Google's AI Innovation Day, lowering targets by up to 10%
  • March 11 to April 11, 2025: Period of Google internal data behind the launch figures
  • May 21, 2025: Google announces Smart Bidding Exploration at Google Marketing Live, citing 18% more converting query categories and 19% more conversions
  • October 15, 2025: Google Ads API v22 adds date segmentation for exploration metrics
  • November 6, 2025: Google Ads Editor 2.11 adds Smart Bidding Exploration as an opt-in for Search
  • January 1 to February 20, 2026: Period of Google's comparison against a plain target reduction (13% more converting categories, 10% more conversions)
  • March 11, 2026: Google product managers discuss exploration on Ads Decoded
  • May 7, 2026: Google announces expansion to Performance Max with feeds and standard Shopping
  • June 15, 2026: Exploration available in all languages for Performance Max without a feed; beta opens for feed-based Performance Max and Shopping; promotion mode beta launches
  • August 17, 2026: Google's target change for budget-limited campaigns takes effect

Summary

Who: Google Ads operates Smart Bidding Exploration; advertisers and agencies using Target ROAS opt in, and Google product managers including Josh Braverman, Kristina Park and Carlo Buchmann have presented it.

What: A bid strategy setting that lets Target ROAS undershoot its target by a set tolerance, usually starting at 10%, and spends that headroom selectively on search categories with plausible but unproven conversion potential.

When: Tested as Bidding Exploration in September 2024, announced at Google Marketing Live on May 21, 2025, and extended to Performance Max and Shopping between May and June 2026.

Where: Search campaigns with broad targeting, Performance Max campaigns, and, in beta, feed-based Performance Max and standard Shopping campaigns, configured in campaign or portfolio bid settings.

Why: Automated bidding favours traffic with conversion history; exploration trades some reported efficiency for wider reach, a trade whose value rests largely on Google's own, unreplicated performance data.