Search Partner Network in 2026: Transparency, Placements and Brand Safety

by Francis Rozange | Apr 4, 2026 | Google Ads

search-partner-network-transparence-2026-1

Search Partner Network in 2026: transparency, placements and brand safety

For two decades, the Google Search Partner Network operated as one of the most opaque corners of digital advertising. Advertisers checked the box, the impressions accumulated, the budget drained, and almost no one inside the buying chain could answer a simple question: where, exactly, did my ads run? The Search Partner Network was the textbook black box, and Google had no commercial incentive to open it.

That equilibrium broke in 2023. A series of investigations published by the analytics firm Adalytics, amplified by Search Engine Land, the Wall Street Journal and Tinuiti, documented Search Partner placements that ranged from absurd to legally problematic: parked domains, Made-for-Advertising sites, sanctioned Russian and Iranian properties, malware-distribution domains, pirate streaming sites. The reports forced Google into a defensive posture for the first time since the network was created, and they triggered a quiet but material program of refunds to specific advertisers exposed by the data.

By 2026, the Search Partner Network looks different on paper: parked domains were removed on February 10, 2026, placement reporting is finally available across Search, Shopping, App and Performance Max campaigns, third-party brand safety verification through DoubleVerify, Integral Ad Science and Zefr is wired into the buying flow, and a campaign-level opt-out is one toggle away. The black box is no longer a black box. It is a translucent box, and what is visible inside still warrants caution.

This guide explains what the Search Partner Network actually is, what Adalytics found, what Google fixed and what Google left untouched, how to opt out cleanly, and how to decide for or against participation depending on your account profile.

What the Search Partner Network is and why it exists

The Search Partner Network, often abbreviated SPN, is a collection of third-party websites and applications that display Google search ads alongside their own search results or content. According to Google Ads Help, the network includes hundreds of properties: legacy search engines like Ask.com, Dogpile.com and Lycos.com, retailer search boxes embedded in sites such as Walmart and Target, telecom and ISP portals, and a long tail of smaller publishers that integrate Google Custom Search. When a user types a query into one of these sites, the partner displays its own results plus a block of Google sponsored ads served from Google’s auction.

The network was created in 2003, two years after AdWords, on the premise that Google’s text-ad inventory could be syndicated outside Google.com without compromising relevance. Partners receive a share of the click revenue. Google extends its reach. Advertisers, in theory, get incremental volume at comparable cost per click.

The structure is simple and the commercial logic is sound. The execution, however, never received the public scrutiny that Google’s own Search auction got. Google.com is audited every day by every SEO and advertiser on earth. Search Partners were audited by no one because no one could see what was happening inside.

For advertisers, three points matter. First, Search Partner placements are bundled by default with Google Search in every new campaign. The opt-out exists, but the opt-in is automatic. Second, Search Partner clicks are billed at the same auction price as Google.com clicks, with the same cost-per-click model. Third, Search Partner traffic has historically converted at a meaningfully lower rate than Google.com traffic. Tinuiti and other agencies have published internal benchmarks showing the gap, sometimes a factor of three to ten depending on vertical.

None of this would matter if the network were merely lower-quality but legitimate inventory. The 2023 Adalytics reporting demonstrated that the problem went well beyond performance.

The 2023-2024 transparency crisis: the Adalytics report

Adalytics is an independent ad-tech research firm run by Krzysztof Franaszek. It is not an agency, not a verification vendor, not a Google partner. Its business model is forensic analysis of programmatic logs and crawled ad calls, sold to brands and regulators that want to know where their money actually went. The firm had already exposed YouTube ad placements running against demonetized child-directed content. In 2023, it turned its attention to Search Partners.

The report, covered in detail by Search Engine Land and the Wall Street Journal in the second half of 2023, documented Google Search Partner placements on properties that no rational advertiser would have approved. The list, published with screenshots and timestamps, included pornographic websites running Google search boxes, pirate streaming and torrent indexing sites, fringe political publishers, parked domains generated by typo-squatters, and Made-for-Advertising sites built specifically to harvest programmatic budgets through cheap clickbait articles surrounded by ad units.

Two findings provoked the strongest reactions. First, Adalytics identified Search Partner placements on websites tied to entities sanctioned by the United States Treasury, including Russian state-aligned media properties operating under sanctions imposed after the 2022 invasion of Ukraine, and Iranian-government-linked domains. Running US advertiser dollars through sanctioned entities is not a brand safety question. It is potentially a federal compliance question under OFAC rules.

Second, Adalytics extended its analysis to DV360, Google’s enterprise demand-side platform, and showed that the same opaque placement problem existed there at scale. DV360 buyers, including Fortune 500 brands and federal agencies, were allocating budget to inventory they could not see and would not have approved. The Wall Street Journal coverage put names on the list of exposed advertisers, and the reputational damage was immediate.

Google’s initial response was defensive. The company disputed the methodology, argued that the placements represented a fraction of total Search Partner volume, and pointed to existing exclusion mechanisms. That position held for roughly six weeks. Then the refunds started.

Adalytics documented, with confirmation from named advertisers, that Google issued credits to specific brands whose ads appeared on the most problematic placements. The refunds were not announced publicly. They were processed quietly through standard make-good channels, and the existence of the credits was the strongest possible admission that the underlying placements should not have happened.

2024 reforms and the February 2026 parked domains removal

The reform program that followed unfolded in stages. Google did not announce a single overhaul. Instead, the platform shipped a sequence of incremental controls and removals across 2024, 2025 and the first quarter of 2026.

In March 2024, Google introduced account-level placement exclusions for Search and Performance Max. Before that change, exclusions had to be applied campaign by campaign, which made enforcement at portfolio scale impractical. The new feature let an advertiser maintain a single exclusion list, sync it across Search, Performance Max, Demand Gen, YouTube and Display, and update it from one console.

In August 2025, Google released full placement reporting for Search Partner Network across Search, Shopping and App campaigns. The report exposed, for the first time, the list of partner domains that displayed each campaign’s ads, with impression counts per domain. It did not include click counts, conversion attribution or cost data at the placement level, and it did not backfill historical data. But it ended the principle that the network was inherently invisible.

In April 2025, Google announced the SPN Pre-Screen Solution, a brand safety verification layer powered by three independent vendors: DoubleVerify, Integral Ad Science and Zefr. Each vendor builds and maintains its own classification of Search Partner inventory and offers exclusion lists that an advertiser can apply directly to campaigns. The verification runs continuously rather than as a one-time audit.

The most decisive change arrived on February 10, 2026: Google removed parked domains, the inventory category historically known as AdSense for Domains or AFD, from the Search Partner Network entirely. The removal applied retroactively across all accounts. The setting that controlled parked domain inclusion in Content suitability was deleted. Parked placements no longer appear in any campaign report from that date forward.

Parked domains were the single largest source of the cybersquatting problem documented by Adalytics. Typo-squatted versions of brand names, expired domains acquired purely to harvest residual traffic, automatically generated landing pages stuffed with affiliate links: this category absorbed real advertiser budget for years and converted essentially never. Removing it was the lowest-controversy, highest-symbolic move available to Google, and it was clearly informed by the documented findings of the prior three years.

In January 2026, Google extended Performance Max channel reporting with Search Partner Network segmentation. Advertisers running PMax could finally see, at the campaign level, what share of impressions, clicks and conversions came from Search Partners as distinct from Google Search, YouTube, Shopping, Display and Gmail. February 2026 added placement-level detail inside that segmentation, mirroring the August 2025 release for Search campaigns.

Where ads can appear: placement reporting in practice

The placement report is accessed inside the Google Ads interface under “When and where ads showed”. For Search, Shopping and App campaigns, the Search Partners segmentation has been available since August 2025. For Performance Max, the same segmentation became available in February 2026, after the channel-level rollout in January.

The data set has known limits. It shows the partner domain and the number of impressions delivered. It does not show clicks, cost or conversions per individual domain. It does not provide the user query that triggered each impression. It is not retroactive. An advertiser running a Performance Max campaign for two years will see partner domain data only from the February 2026 rollout forward.

What the report does enable is the construction of a credible exclusion list. An advertiser can identify partner domains receiving meaningful impression volume, evaluate them against brand standards, and add the unacceptable ones to the account-level placement exclusion list. The work is manual at first. It becomes routine after one or two cleanup passes.

The report also enables a sanity check on the volume question. Search Partners often represent a small share of total impressions for B2B accounts, but a meaningful share for consumer accounts running broad-match keywords or Performance Max with extensive asset groups. Quantifying this share is the prerequisite to deciding whether the channel deserves any participation at all.

Adalytics findings on MFA, DV360 and brand safety failures

Made-for-Advertising sites, abbreviated MFA, are a category of web property created specifically to harvest programmatic ad spend. They are not built to inform, entertain or transact. They are built to attract a click, load eight to twelve ad units per page, generate scroll events that trigger viewability beacons, and earn the resulting cost-per-mille fees. MFA sites typically rely on aggregated content from press releases, scraped news, low-effort listicles and arbitrage traffic purchased from social platforms.

The Adalytics reporting on Search Partners and DV360 placed MFA inventory at the center of the brand safety failure. The sites were not pornographic, not pirated, not sanctioned. They were merely worthless. Advertiser budget flowed to them in volume because programmatic systems optimize for impressions and clicks, both of which MFA properties manufacture efficiently. The conversion rate is negligible, but the upstream metrics look acceptable.

Tinuiti, in agency-side research published in 2024 and updated in 2025, quantified the cost of MFA exposure for several of its clients. The firm found that excluding identified MFA domains improved measured return on ad spend by between 10 and 40 percent depending on vertical, with no detectable loss of incremental conversions. The interpretation is straightforward: MFA placements were not driving incremental sales. They were absorbing budget that would otherwise have flowed to higher-quality inventory in the same auctions.

The DV360 strand of the Adalytics work was particularly damaging because DV360 is sold as the enterprise-grade alternative to programmatic open exchanges. Buyers using DV360 expect a curated supply path, vendor accountability and transparent placement data. The investigation showed that the platform’s default settings exposed buyers to the same problematic inventory categories as the open web, including Russian state-tied properties operating under sanctions and pirate sites operating under DMCA takedown notices.

The reputational consequence was that several large advertisers paused DV360 spend pending internal audits. The financial consequence was the credit program that Adalytics documented through advertiser disclosures. The structural consequence was the acceleration of the reform program described in the previous section.

The Adalytics methodology and why it held up

The Adalytics work merits a section of its own because the methodology is what made the findings impossible for Google to dismiss. The firm did not rely on hearsay or on advertiser screenshots after the fact. It built a forensic pipeline that crawled partner sites at scale, captured the live ad calls, decoded the parameters that identified the buying campaign, and matched them back to specific advertisers. Every claim in the published reports was anchored in a logged ad request with a timestamp, a partner domain, a creative ID and the chain of intermediary IDs that traced the placement back to its source.

This is the same evidentiary standard that brand safety vendors apply internally for incident response. Adalytics applied it externally and published the results. The defensive arguments that Google initially raised, principally that the placements represented a statistical fringe, did not survive contact with the data because the data were not statistical samples. They were enumerated incidents, each documented to the level required for a make-good claim.

The Wall Street Journal’s coverage extended the methodology by interviewing the named advertisers and obtaining confirmation of the make-good credits. Search Engine Land tracked the platform-side response week by week and documented the timing of each Google announcement against the reporting cycle. Tinuiti contributed agency-side benchmarks that quantified the financial impact on real campaigns. The combination of independent forensic research, financial press validation, trade press tracking and agency benchmarking produced a record that the platform could not credibly contest.

The refund programme that followed was the most consequential outcome. The credits were modest in absolute terms relative to Google’s revenue, but they constituted an admission of liability that altered the negotiating posture between platform and advertiser for every subsequent compliance discussion. Once a platform has issued make-goods on a category of placement, that category is permanently encumbered. The February 2026 parked domains removal is, in part, the direct operational consequence of that encumbrance.

How to opt out of the Search Partner Network

The opt-out is a single setting at the campaign level. The path is consistent across campaign types.

For a Search or Shopping campaign, open the campaign, go to Settings, expand the Networks section, and uncheck the box labeled “Include Google search partners”. Save. The change applies on the next ad serve. The campaign now runs only on Google.com, Google Images, Google Maps, Google Shopping and YouTube Search, depending on campaign type.

For a Performance Max campaign, the same logic applies but the path differs. The Networks section in PMax exposes a Search Partners toggle that controls participation. The opt-out is reversible at any time.

For App campaigns, the Search Partners setting is bundled with the Search network setting and can be controlled separately under Settings.

For accounts that want a global default, Google does not offer an account-level opt-out. The setting must be applied to every campaign individually, which is one of the legitimate criticisms of the current architecture. A scripts-based or Google Ads Editor approach is the practical workaround for portfolios with dozens of campaigns.

Three details matter. First, opting out reduces total impression volume. The volume reduction is variable and depends on auction dynamics, but typical observed ranges run from 5 to 30 percent of total impressions for Search campaigns. Second, opting out does not change cost per click on Google.com inventory. The auction is independent. Third, the opt-out does not affect Display, YouTube or Discover placements, which sit in different network categories.

Brand safety controls: placement exclusions and content categories

Beyond the binary opt-out, Google exposes two additional control layers that an advertiser can use to shape Search Partner participation without disabling the channel.

Account-level placement exclusions allow an advertiser to specify domains, mobile applications or YouTube channels that should never display the account’s ads. The list can hold up to 65,000 entries. It applies across Search, Performance Max, Demand Gen, YouTube and Display campaigns simultaneously. The list can be uploaded as a CSV, edited manually, or generated programmatically through the Google Ads API.

Content suitability settings, accessed under Tools and Settings, expose a tiered control: Expanded inventory, Standard inventory, or Limited inventory. The default is Standard. The Limited setting excludes additional content categories such as mature content, profanity and sensitive social issues. The categories are defined by Google’s classification systems and audited by the SPN Pre-Screen vendors when those vendors are enabled.

The Pre-Screen Solution itself is the third layer. An advertiser activates DoubleVerify, Integral Ad Science or Zefr in account settings, configures suitability preferences with the chosen vendor, and the vendor builds and maintains an exclusion list applied to Search Partner inventory. The list updates continuously as the vendor’s classification of partner inventory evolves. Costs vary by volume and vendor and typically fall in the range of several thousand dollars per month for mid-market accounts.

The combination matters. Manual placement exclusions handle known problematic domains. Content suitability handles broad category filtering. Pre-Screen handles continuous third-party verification. Each layer addresses a different failure mode, and each is independently useful.

Recommendation framework by account type

The right Search Partner Network strategy depends on the account profile, the regulatory environment, the brand safety stance and the volume sensitivity. The framework below summarizes the defensible defaults.

B2B and enterprise software: opt out by default

B2B accounts rarely benefit from Search Partner volume. The channel’s incremental impressions tend to come from low-intent queries or partner sites that do not match B2B audience signals. The conversion rate gap between Google.com and Search Partners is typically widest in this segment. The brand safety stakes, particularly for vendors selling to regulated industries or government, are also highest.

The defensible default is to disable Search Partners on every Search and Performance Max campaign, monitor for any meaningful volume loss, and reactivate selectively only if a specific campaign demonstrates incremental value with the channel enabled.

Regulated industries: opt out by default

Pharmaceutical, financial services, insurance, legal services and healthcare advertisers face content-adjacency rules that go beyond ordinary brand safety. A pharmaceutical brand cannot afford its ads to appear next to unverified medical claims. A regulated financial advisor cannot afford to appear next to unregulated investment advice. The Adalytics findings on adjacent placement risks make Search Partners a poor fit for these verticals.

The defensible default is to opt out, document the decision in the account governance log, and reconsider only if Pre-Screen verification is implemented with a vendor whose classification is acceptable to the compliance function.

Ecommerce: test cautiously

Consumer ecommerce accounts can sometimes find incremental value in Search Partners, particularly when running Performance Max with Shopping feeds. Retailer search boxes, comparison sites and shopping aggregators sit in the partner network and can drive bottom-funnel traffic at acceptable cost per acquisition.

The defensible approach is to enable Search Partners on a subset of campaigns, instrument channel-level reporting in Performance Max, evaluate the Search Partner contribution to conversions versus cost over a 30 to 60 day window, and either retain, exclude specific domains, or disable the channel based on the data. Treat the inclusion as a hypothesis to be tested, not a default to be accepted.

Local services: usually opt out

Local service businesses, particularly those running geographically constrained campaigns, rarely benefit from the partner network. Most partner sites lack the geographic precision that local campaigns require. The volume gain is small and the conversion gain is smaller.

The defensible default is to opt out and direct budget to Local Services Ads or Google Search proper.

High-volume direct response: keep with controls

Affiliate-style or high-velocity direct response advertisers operating on tight CPC economics sometimes find Search Partners acceptable when paired with aggressive bid adjustments and exclusion lists. The strategy is to keep the channel enabled, apply a negative bid modifier in the range of -30 to -50 percent if the campaign type allows it, build an exclusion list of identified low-quality partner domains, and treat the channel as residual volume rather than primary inventory.

Common mistakes

The transparency improvements of 2024 to 2026 do not protect an advertiser who fails to use them. The recurring mistakes observed across audited accounts are predictable.

The first mistake is leaving Search Partners enabled by default without reviewing the placement report. The default is checked. The advertiser has never looked. Budget flows to inventory the advertiser has not approved and would not approve if asked. The fix is one campaign-settings review and one bulk exclusion list.

The second mistake is treating the August 2025 and February 2026 placement reports as final. The reports show impression counts but not conversions or costs at the domain level. An advertiser who concludes that a partner domain “performs well” because it has 5,000 impressions is reading the wrong signal. Impression volume is not performance. Without clicks and conversions per domain, the report is a starting point for exclusion decisions, not a basis for inclusion decisions.

The third mistake is applying placement exclusions at the campaign level when account-level exclusions exist. Campaign-level exclusions create maintenance debt: every new campaign requires the list to be reapplied, and inconsistencies accumulate. Account-level exclusions, available since March 2024, propagate automatically.

The fourth mistake is conflating Search Partners with the Display Network. They are separate networks with separate inventory, separate reporting and separate controls. An advertiser who excludes Display does not exclude Search Partners. The opt-out toggles are independent.

The fifth mistake is assuming that the parked domains removal of February 10, 2026 solved the brand safety problem. It removed one category of low-quality inventory. MFA sites, low-authority publishers and edge-case adjacency risks remain inside the network. The Adalytics methodology that found the original problems would still find problems if reapplied today, in different categories and at lower volume but in the same structural pattern.

The sixth mistake is overspending on third-party verification for accounts whose volume does not justify it. Pre-Screen vendor contracts in the range of several thousand dollars per month make sense for accounts spending six or seven figures on Search and Performance Max. For smaller accounts, the cleaner answer is to opt out of Search Partners entirely and avoid the verification spend.

Conclusion

The Search Partner Network of 2026 is a different artifact than the Search Partner Network of 2022. Parked domains are gone. Placement reporting exists across every major campaign type. Account-level exclusions consolidate enforcement. Third-party verification through DoubleVerify, Integral Ad Science and Zefr is wired into the buying flow. The opt-out is one toggle.

What did not change is the underlying commercial structure. Google still bundles Search Partners into new campaigns by default. Google still bills partner clicks at the same auction rate as Google.com clicks. Google still does not provide click, cost or conversion data at the partner-domain level, which means the reporting available to advertisers is incomplete by design. The Adalytics methodology that exposed the original problems can still find problems, in fewer categories and at lower volume, but in the same structural shape.

The right posture for 2026 is not blanket distrust and not unconsidered acceptance. The right posture is a deliberate policy decision recorded in the account governance log: Search Partners are off by default for B2B and regulated industries; on with controls for ecommerce and tested through 30 to 60 day windows; off for local services; and continuously monitored everywhere they are on. The transparency tools exist. They were paid for, in reputational and financial terms, by the Adalytics findings and the refunds Google issued in response. Using them is the minimum standard.