Pre-bid filtering is the screening of an advertising impression before any money is offered for it, so that a buyer declines to bid, or a seller declines to offer, when the opportunity fails a quality test. The tests address three risks: invalid traffic generated by bots and fraud schemes, content an advertiser does not want to appear beside, and supply judged wasteful, such as unviewable placements. The practice exists because programmatic auctions settle in milliseconds across pages no buyer will ever see. Removing an advertisement after purchase can prevent the placement. It cannot undo the purchase.

How a bid request is screened

The raw material is the bid request, the OpenRTB message an exchange sends to demand-side platforms (DSPs) describing each impression. Screening draws on a handful of its fields: the page URL or app bundle, the device IP address, the user-agent string and, since OpenRTB 2.6 was published in April 2022, a structured user agent object called sua, built from browser client hints.

Speed dictates the architecture. Content cannot be classified while the auction waits, so vendors crawl and score pages in advance. According to Peer39, a contextual data company, the DSP checks each request against an internal cache filled from the vendor's own, with results tied to a URL or a domain. A page never crawled carries no score. Display & Video 360 (DV360), Google's DSP, leaves the consequence to the buyer: its Integral Ad Science (IAS) settings include an option to block unrated URLs.

Fraud checks lean harder on live signals. According to HUMAN's documentation, a platform sends its MediaGuard service either selected fields or the whole bid request and receives a verdict it can use to drop the impression. White Ops, as HUMAN was then called, said that in most cases responses arrived in 10 milliseconds or less. DoubleVerify says fraud signatures inside its DSP integrations are updated roughly every 15 minutes.

Standards set the floor. OpenRTB implementation guidance recommends that exchanges filter known spiders by user agent and that bidders decline to bid on them, returning a reason code: 3 for a known web spider, 4 for suspected non-human traffic, 5 for a cloud, data centre or proxy IP. The Media Rating Council (MRC) separates general invalid traffic (GIVT), caught with lists and parameter checks, from sophisticated invalid traffic (SIVT), which needs advanced analytics. Its 2020 standard requires filtering invalid data centre traffic from Amazon Web Services, Google and Microsoft address ranges.

In DV360, third-party controls sit under Targeting, then Brand suitability, then Other verification services, at insertion order or line item level. According to Google's Structured Data File documentation, one vendor can be selected per entity, and charges appear on a Third-Party Rate Card. Material from the IAB and IAS states that pre-bid targeting fees are applied only to winning bids.

Who operates it

On the buy side, three layers stack. DSPs run native controls, such as DV360's digital content labels and sensitive categories. Verification vendors, chiefly IAS and DoubleVerify, sell segments activated inside those DSPs. Buyers add their own inclusion and exclusion lists, which can now reach below the domain: DV360 made seller ID blocklists available to all users in April 2025 at an extra 1.5% of media cost.

On the sell side, exchanges filter before requests leave. MoPub's marketplace integrated Pixalate's pre-bid filtering in November 2021, dropping ad calls that matched 38 types of invalid traffic across IP addresses, device IDs and user agents, which Pixalate said limited exposure for more than 180 DSP partners. Index Exchange said in August 2020 that its White Ops partnership protected inventory before a bid request was ever sent to a DSP.

Closed platforms, which expose no open bid request, apply vendor-supplied exclusion lists instead. IAS unveiled Content Block Lists for Facebook and Instagram on October 10, 2024, and DoubleVerify launched pre-bid video exclusion lists for TikTok on April 14, 2025. YouTube sits outside IAS's pre-bid settings in DV360 altogether, according to Google's help centre.

Origin and evolution

Screening entered real-time bidding (RTB) through contextual data firms. Peer39's quality, safety and category attributes became available through Brandscreen, a DSP operating in Asia Pacific, on August 30, 2011. By October 31, 2013, Peer39's then owner DG had brought the data to The Trade Desk, with Alex White of DG describing pre-bid data as "vital".

Fraud filtering followed the standards. The MRC issued its invalid traffic guidelines on October 27, 2015 and a comprehensive update on June 25, 2020. On March 30, 2016, IAS introduced what it called the first pre-bid brand safety segment for mobile apps, on AppNexus, reasoning that apps cannot be crawled like websites. In August 2017, The Trade Desk and White Ops announced a plan to scan biddable impressions on the platform in real time.

Customisation came next. Google opened DV360 to custom pre-bid filtering in October 2020, with IAS as the first vendor; until then, third-party filtering there had covered standard categories such as alcohol and violence. DoubleVerify brought Authentic Brand Safety to DV360 on November 2, 2020, citing more than 75 avoidance categories in more than 44 languages. In June 2022, IAS released Quality Sync, a single pre-bid segment mirroring an advertiser's post-bid settings, removing the manual work of maintaining both.

Why it matters to marketers

For buyers, these controls are where a written brand policy becomes a spending decision, repeated billions of times a day. For vendors, they are commercially central: DoubleVerify reports revenue from measurement services and from higher-margin activation services, the category that includes its programmatic avoidance products.

Publishers feel the settings directly, because conservative thresholds withhold revenue from legitimate pages, news above all. Vodafone and DoubleVerify reported in July 2025 that revised suitability settings cut block rates by 41% and opened 10% more news inventory, figures supplied by the two companies. Every exclusion is also a decision about whose content goes unfunded.

Limitations and disputes

Missing signals are the core technical weakness. Adalytics, a research firm, published findings on March 28, 2025 that pre-bid services from DoubleVerify and IAS do not receive browser, device and IP address data from some of the largest buying platforms, leaving them unable to separate declared bots from data centre traffic. The report documented US Navy ads carrying a "charge-allDoubleVerifyBotAvoidance" reference being served to declared bots on Google Cloud servers. DoubleVerify has called Adalytics research "misleading and manufactured" and says bots missed before bidding are often caught afterwards, so advertisers are not charged. A securities class action followed on May 22, 2025.

Privacy measures remove the same inputs. AdSense bid requests carry only the first three octets of an IP address by default. A US settlement approved on March 26, 2026 requires a Google control that strips IP addresses and device advertising IDs from bid requests for users who activate it, and generalises user agents.

Classification carries its own dispute. The MRC decided on October 18, 2025 that keyword or domain-level analysis cannot be positioned as brand safety unless vendors also examine images, video and audio, with a grace period to April 18, 2026.

Politics arrived through exclusion lists. The Global Alliance for Responsible Media, whose framework standardised risk categories, shut down in August 2024 after X sued. On April 15, 2026, the Federal Trade Commission and eight states alleged that WPP, Publicis and Dentsu conspired from 2018 to impose a common brand safety floor that demonetised certain conservative publishers.

Independence is the remaining question. Nielsen agreed on August 6, 2026 to acquire DoubleVerify for about $2.15 billion, placing the largest listed verification vendor inside the company that sets US television audience currency. IAS had already been taken private by Novacap in a $1.9 billion deal.

Not the same as

Post-bid verification examines what was actually delivered and reports back. A pre-bid control prevents a purchase; a post-bid measurement describes one. Post-bid blocking, a hybrid, suppresses the creative after the impression is won, and is not possible everywhere: DoubleVerify has cited VAST video as an environment where blocking is unsupported.

Prebid, one word, is the open-source header bidding project maintained by Prebid.org, which gathers bids before a publisher's ad server decides. It organises auctions rather than judging quality.

Traffic shaping and bid throttling reduce request volume for cost and latency reasons. DSPs filter incoming requests they are unlikely to bid on, and publishers throttle duplicate calls; Will Doherty of The Trade Desk said supply-side platforms pass on 30% of a publisher's traffic, the portion they judge best, and "they send it three times".

Bid-time valuation prices inventory instead of excluding it. Google began feeding Adelaide attention scores into DV360 bidding from mid-July 2026, while the same scores packaged as Amazon DSP segments work as pre-bid exclusions.

Recent developments

Controls are narrowing inside one platform and multiplying outside it. Google will withdraw digital content label exclusions and most sensitive category exclusions from the DV360 API and Structured Data Files on October 1, 2026, directing buyers to inventory modes and content themes that do not map one to one. IAS put a pre-bid segment against low-quality AI-generated content into open beta on April 2, 2026.

Screening is also moving upstream. Mediavine and SWYM extended AI supply shaping across more than 18,000 publisher sites on June 16, 2026, evaluating inventory before any bid request is sent. On September 9, 2026, BidSwitch, the exchange owned by Criteo, offered supply partners free Adalytics analysis of their inventory on top of its existing malware and invalid traffic filters, without saying when it will be available or what it will measure.

Timeline

  • August 30, 2011: Peer39 page-level quality, safety and category data becomes available through the DSP Brandscreen in Asia Pacific
  • October 31, 2013: DG brings Peer39 pre-bid page data to The Trade Desk
  • October 27, 2015: The MRC issues Invalid Traffic Detection and Filtration Guidelines Version 1.0, establishing GIVT and SIVT
  • March 30, 2016: IAS launches a pre-bid brand safety segment for mobile apps on AppNexus
  • August 2017: The Trade Desk and White Ops announce real-time scanning of biddable impressions
  • June 25, 2020: The MRC issues updated invalid traffic standards, including data centre filtration requirements
  • August 12, 2020: Index Exchange expands its White Ops partnership to screen inventory before requests reach DSPs
  • October 2020: DV360 opens custom pre-bid filtering, with IAS as the first vendor
  • November 2, 2020: DoubleVerify launches Authentic Brand Safety on DV360
  • November 2021: MoPub integrates Pixalate pre-bid invalid traffic filtering
  • April 2022: OpenRTB 2.6 adds the structured user agent object
  • June 27, 2022: IAS launches its Quality Sync pre-bid segment
  • August 2024: The Global Alliance for Responsible Media shuts down after an antitrust suit from X
  • October 10, 2024: IAS and DoubleVerify announce pre-screen content controls for Facebook and Instagram
  • March 28, 2025: Adalytics publishes findings on pre-bid bot filtration
  • April 2025: DV360 makes seller ID blocklists available to all users with a 1.5% media cost fee
  • April 14, 2025: DoubleVerify launches pre-bid video exclusion lists for TikTok
  • May 22, 2025: A securities class action is filed against DoubleVerify
  • July 15, 2025: Vodafone reports suitability results with DoubleVerify
  • October 18, 2025: The MRC restricts brand safety terminology to vendors with content-level analysis
  • March 26, 2026: A US federal judge approves the Google RTB settlement requiring identifier stripping
  • April 2, 2026: IAS opens a low-quality AI content avoidance segment to open beta
  • April 15, 2026: The FTC and eight states sue WPP, Publicis and Dentsu over brand safety coordination
  • April 18, 2026: The MRC grace period on brand safety terminology ends
  • June 16, 2026: Mediavine and SWYM extend supply shaping across more than 18,000 publisher sites
  • Mid-July 2026: DV360 begins using Adelaide attention scores in bidding
  • August 6, 2026: Nielsen agrees to acquire DoubleVerify for about $2.15 billion
  • August 27, 2026: Google announces the withdrawal of two DV360 exclusion controls
  • September 9, 2026: BidSwitch opens Adalytics bidstream analysis to supply partners
  • October 1, 2026: Digital content label and most sensitive category exclusions are scheduled to leave the DV360 API and Structured Data Files

Summary

Who. Demand-side platforms, verification vendors including IAS, DoubleVerify, HUMAN, Pixalate and Peer39, exchanges and supply-side platforms, and the advertisers and agencies that set the thresholds. The MRC and the IAB Tech Lab write the standards it is measured against.

What. Pre-bid filtering is the rejection of an impression before a bid or offer is made, when bid request signals or pre-computed page, app and fraud scores fail thresholds for invalid traffic, content suitability or supply quality.

When. Page-level pre-bid data reached DSPs by 2011, fraud filtering was standardised after the MRC's 2015 guidelines, and custom controls spread from 2020. Between 2025 and 2026 its accuracy, its terminology and its political use were all contested.

Where. Inside DSP targeting settings, in exchange pipelines before requests are sent, in publisher supply-shaping layers, and inside closed platforms such as Meta and TikTok through vendor exclusion lists.

Why. A check applied after an impression is bought cannot recover the spend. Screening first protects budgets and brand policies, but it depends on signals that privacy rules and platforms increasingly withhold, and on classifications that remain disputed.