How-To Guide

How to Measure TV ROI in the AI Era

TV measurement has always relied on MMM. The AI era adds two complications: streaming TV measurement and the AI search halo. How to handle both.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: May 18, 2026

The TV Measurement Problem

TV has always been measured through MMM because there is no user-level signal. The AI era adds two complications. First, streaming TV and CTV produce partial user-level signal that some advertisers try to use for attribution, with mixed results. Second, AI search halo effects make TV's contribution harder to isolate from other upper-funnel channels.

Step 1: Capture Linear and CTV Separately

Linear TV and CTV behave differently in MMM. Linear has broad reach and long carryover; CTV has more targeted delivery and somewhat shorter carryover. Model them as separate variables with separate adstock and saturation parameters. Aggregating produces less accurate identification of each.

Step 2: Use GRPs or Impressions, Not Spend

Spend is a noisy proxy for TV exposure because rate cards vary by daypart, network, and season. GRPs (gross rating points) for linear and verified impressions for CTV are the better exposure measures. Most MMM frameworks accept either spend or exposure as the channel input.

Step 3: Set Long Adstock Priors

Linear TV adstock half-life: three to six weeks. CTV adstock half-life: two to four weeks. Longer than digital channels. Borrowing digital priors underweights TV in MMMs and explains why some brands incorrectly conclude TV is "not working."

Step 4: Capture the AI Search Halo

TV campaigns drive AI visibility through several pathways: increased branded search produces more content discoverable to AI, PR amplification of TV moments feeds AI training data, and consumer mentions of seen ads on social media affect AI assistant context. Include AI visibility as a separate variable in the same MMM; the TV variable captures direct TV effects, the AI variable captures the AI-mediated downstream effect.

Step 5: Calibrate With Geographic Lift

Run a geographic lift test on TV every two to four quarters. Pause TV spend in matched regions for eight to twelve weeks. Outcomes are branded search lift, direct traffic, and AI visibility movement in test versus control regions. The lift estimate calibrates the MMM's TV coefficient against causal ground truth.

Step 6: Distinguish CTV-Attributed From CTV-Lift

CTV platforms offer user-level attribution products that link CTV impressions to website conversions. These products produce attribution-style estimates that often overstate CTV's incremental effect because they take credit for viewers who would have converted anyway. The MMM coefficient is the cross-channel calibrated estimate; the CTV platform attribution is the tactical optimization signal. They will disagree; reconcile with periodic lift tests.

How Presenc AI Helps

Presenc AI provides the AI visibility data that disentangles TV's direct contribution from its AI-mediated downstream contribution. Without the AI variable, TV often takes credit for AI search lift that was downstream of the TV campaign; with the AI variable, the two effects are separated and the TV coefficient reflects its direct contribution alone.

Frequently Asked Questions

No. Different adstock, different saturation, different targeting precision. Model them as separate variables. Some MMM frameworks make this easy (separate paid_media_var entries in Robyn); others need explicit specification. Aggregating produces worse identification of each.
Bidirectionally. TV drives AI visibility through PR and branded search; AI visibility drives some of the consideration that TV creative is designed to produce. Both variables in the same MMM, with attention to multicollinearity, captures the joint dynamics. Omitting either causes the other to absorb its credit.
Two to four weeks geometric half-life, shorter than linear TV (three to six weeks) and longer than paid social (one to two weeks). CTV has more targeted delivery than linear, which reduces the broad-reach carryover but produces stronger immediate consideration lift.
For tactical within-CTV optimization yes; for cross-channel allocation no. CTV platform attribution measures within-channel performance reasonably well but takes credit for cross-channel conversions that would have happened anyway. Use MMM for the cross-channel view and run periodic lift tests to reconcile.

Track Your AI Visibility

See how your brand appears across ChatGPT, Claude, Perplexity, and other AI platforms. Start monitoring today.