
CTV audience targeting works at the household level. A streaming ad reaches a television that shares a home network, and often a login, with the phones and laptops in the same house. Most targeting options are rules for deciding which households receive the impression, and a rule is only as good as the identifier behind it.
Buyers have moved targeting to the top of the list. In the IAB's 2026 Digital Video Ad Spend and Strategy report, 49 percent of 360 decision-makers named targeting capabilities as a top criterion for where TV and video money goes. Content quality followed at 46 percent. Targeting was up ten points from 2025, and the report ties the rise to IP degradation and AI-driven traffic.
Streaming took 49.0 percent of US television time in Nielsen's July 2026 Gauge.
The IAB's September 2026 standard for classifying video advertising sets the terms. Impression-based delivery goes to a known audience: a logged-in person, or a household identified by a smart TV with an IP address. That identity supports targeting, frequency capping, and measurement. Deterministic signals come from logins, automatic content recognition, and set-top boxes. Probabilistic signals are modeled from behavior and context when no resolved identity exists.
Platform documentation names three identifiers:
Identity graphs link these identifiers. MNTN's identity graph documentation from May 2026 lists the IP address and the device identifier among its inputs and covers more than 120 million households. It caps a household at 50 devices, since larger networks are usually hotels, airports, or offices. The Trade Desk's explainer, updated September 2026, clusters cookies, mobile ad IDs, CTV IDs, UID2, and RampID at the household and individual level. CTV and streaming device IDs sit at the household level because several people use them.
Exclusions come with several of these options. Roku's location and genre targeting both support exclusion lists, and MNTN excludes households by location.
Household identity decides how many times a home sees the ad. DV360 caps frequency across CTV apps on the same device when the IFA or another supported identifier type is passed in the bid request, and impressions without one are excluded from capped line items. Amazon says household recognition lets it cap across every streaming app in use and reports 57 percent fewer redundant exposures, per its internal data from March 2026.
In the IAB report, 37 percent of buyers would pay extra for cross-device or household targeting and measurement, and 32 percent for frequency capping across partners. Premion's 2026 CTV/OTT survey with Advertiser Perceptions is a vendor-commissioned study of 151 agency or brand-side marketers at US advertisers spending at least $250,000 a year, fielded in October 2025. In it, 49 percent named better frequency control across platforms among the most valuable innovations for CTV campaigns, and 33 percent said it is likely in 2026. Innovid's August 2026 case study for one political campaign suppressed households above 18 impressions and raised bids on those below three, across two platforms.
On the browser side, Google announced in October 2025 that it would retire Chrome's IP Protection, and its status page in August 2026 lists the feature as scheduled for phaseout.
Notice and consent now sit on the screen. California's CCPA regulations took effect January 1, 2026. A business that sells or shares personal information collected through a connected device, such as a smart TV, must provide the opt-out notice before or at the time the device begins collecting it. The Texas attorney general announced agreements with Samsung on February 26, 2026 and with LG on May 11, 2026. Samsung must obtain Texas consumers' express consent before collecting ACR viewing data, and LG's agreed judgment requires affirmative express consent once a 180-day period after the judgment takes effect has run. Comprehensive privacy laws in Indiana, Kentucky, and Rhode Island took effect January 1, 2026. Connecticut's amendments on July 1, 2026 extended its law to any business processing data on at least 35,000 residents, selling personal data, or processing sensitive data.
SmartReach, in AI Dash, starts with the product, not just the audience. It organizes product and audience intelligence across five categories and more than 500 tags, covering audience intents, use cases, traits, content contexts, and product model, and builds on first-party data and conversion signals. Household Resonance extends a CTV campaign to the eligible mobile, tablet, and desktop devices in the same household. Retargeting applies SmartReach to existing users, combining first-party audience data with engagement and conversion signals. For how a household match proves an outcome, see Starti's guide to CTV ad measurement.
Book a demo to see SmartReach.
CTV audience targeting selects which households receive a streaming TV ad. The signals include geography, daypart, content genre, demographics, third-party segments, publisher first-party data, advertiser lists, ACR viewing data, and site or app retargeting. Most options depend on an identifier: an IP address, a device IFA, a login, or a graph that links them.
The television, phones, and laptops in a home usually share one IP address, and streaming apps add logins and device identifiers. Identity graphs cluster those signals into a household record. Segments, frequency caps, and attribution are applied to that record.
Common options are state, ZIP, and DMA geography; dayparts; genre and content category; and age, gender, parental status, and household income. Data-driven options add behavioral and in-market segments, publisher and retailer audiences, advertiser customer lists, ACR-based viewing audiences, and retargeting of site visitors or app users.
A pixel or app event records a visit or action, and the platform matches the visitor's IP address or device identifier to households that can be reached on streaming TV. MNTN documents pixel-based retargeting segments, and AppsFlyer added retargeting measurement for CTV app campaigns in August 2026.
Both. The IAB's September 2026 classification standard treats logins, ACR, and set-top box data as deterministic and modeled inferences from behavior and context as probabilistic.
California's CCPA regulations effective January 1, 2026 require a business that sells or shares data collected through a smart TV to show an opt-out notice before collection begins. Texas reached ACR consent agreements with Samsung and LG in 2026. New comprehensive privacy laws took effect in Indiana, Kentucky, and Rhode Island on January 1, 2026, and Connecticut's amendments on July 1, 2026.