Every TV maker in this story sells advertisers the whole screen. Then I read what they told the SEC.

About The Investigation. Every Wednesday, ADOTAT publishes one long piece built from primary sources rather than press releases: written on-record answers, contracts, filings, transcripts, and the questions the trades tend not to ask. The format is deliberately narrow.

One subject, named companies, named executives, and every claim put to the people it concerns before publication, with their answer printed as they gave it or their silence noted. Nothing here is sponsored, nobody reviews it in advance, and the reporting runs whether or not the subject is a friend of the publication.

Vizio's sales page promises advertisers "real-time, glass-level Inscape ACR data from 28M+ opt-in devices." Glass-level is the whole pitch. If it's on the screen, the TV saw it.

Vizio's last annual report to the SEC, filed before Walmart bought the company, says something else. The company "currently do[es] not collect, and might not in the future collect, viewing behavior data regarding content streamed through SmartCast," its own app platform. Then the line that matters, which appears twice in the filing: "some of our agreements with third party content providers, including Netflix and Disney+, restrict us from using viewing data from consumers engaging with that third party's content."

Read those two sentences side by side. One is written for buyers. The other is written for the Securities and Exchange Commission, where the penalty for overselling is a lawsuit.

Automatic content recognition, or ACR, is the data underneath a huge share of TV advertising. The TV listens to or samples what's playing, turns it into a fingerprint, and matches that fingerprint against a library. That's how a TV maker knows you watched a show or saw an ad, and that data is sold for targeting, measurement and attribution. Samsung, LG, Vizio, Roku and Samba TV all sell some version of it, and much of the CTV "outcome" business is built on top.

The pitch is that it sees everything. The paper trail says it goes dark exactly where the audience went.

Netflix and YouTube alone account for more than a fifth of all US TV viewing, by Nielsen's own count. Inside the native apps on the TV, the evidence says ACR mostly can't see them.

Samsung tells advertisers its ACR "captures every pixel being passed back through the glass" and "recognizes content viewership, across linear TV and OTT." A Samsung Ads guide published with IAB Canada in 2022 says it "measures everything that is 'on the glass.'"

Samsung tells TV owners the opposite. Its consumer support page: "Samsung does not record or watch any of the content displayed on your Smart TV."

LG Ad Solutions told Beet.TV in 2021, "We understand everything that hits the screen."

LG itself, on Sept. 12, 2026, after independent testers raised privacy questions about its TVs: ACR "does not operate within third-party streaming apps such as Netflix or YouTube available through LG Apps." It runs on live TV, HDMI inputs and LG's own free channels.

Inscape, Vizio's ACR unit, now owned by Walmart: its data chief Ken Norcross told The Measure in 2025, "We see everything: streaming, linear, and what's in between." In the same interview: "some apps don't allow ACR."

Everything, except the parts that aren't included. That's not a gotcha. It's the product sheet.

None of this is a secret inside measurement. It's just not on the sell sheet.

CIMM, the Coalition for Innovative Media Measurement, the industry's own research group, put it in plain language in its October 2024 Strategic Review of Content Measurement: "due to contractual agreements between streaming app owners and TV set manufacturers, streaming viewership data captured using automated content recognition (ACR) technology on Smart TV sets is limited to OTT streaming consumption (e.g., through devices such as Apple TV, Chromecast, Fire Stick or Roku)."

In other words, the TV can see Netflix when Netflix comes in through a stick plugged into the HDMI port. When Netflix runs as an app on the TV itself, the contract turns the lights off.

CIMM said it again in 2021, in a study built with Janus Strategy and Insights: "Smart TV ACR providers generally cannot report on viewing within OEM apps." The same deck gives the vendor slogan with the qualifier the slogan leaves off: ACR "can report everything that hits the screen for which there is a library fingerprint."

Jon Watts of CIMM told ADOTAT the study identified "contractual restrictions imposed by some OTT apps that prohibited measurement within the TV manufacturer's native environment." He was careful about the limits: it "did not publish capture or matching rates by input," and it shouldn't be read as a statement about Netflix's or YouTube's current arrangements.

The people who built the data say the same thing. Kevin Krim, CEO of EDO, an ACR-based TV ad measurement company: "A bunch of the premium streaming apps are blocked from ACR collection by legal contracts. The top apps negotiate hard: if you want our native app on your TVs, then you have to agree to zero data collection when it's active. It's been this way from as early as I can remember."

Tom Weiss, who ran data science at Inscape: "Some streaming providers offer their streaming libraries for ingest, but most don't, and actively prevent ACR when their native app is running."

Nielsen, which licenses ACR data and sells the panel that calibrates it, wrote in 2023: "Some smart TV streaming applications block ACR from capturing the content on screen while the app is in use." Nielsen has a commercial interest in pointing at those gaps, since its panel is sold as the fix. Asked this month about ACR in its own measurement, Nielsen said it is "not able to comment on this right now."

A peer-reviewed study presented at the ACM Internet Measurement Conference in 2024, by a team including researchers from University College London and UC Davis, put Samsung and LG sets in the US and UK on a network tap and watched what they sent home.

On live TV and HDMI, the ACR traffic was heavy. An LG set in the US sent more than 4,000 KB an hour to its ACR servers while watching antenna TV. Inside Netflix and YouTube, LG's ACR traffic fell to idle levels, about the same as a TV sitting on the home screen. Samsung's US fingerprint server showed none.

Is "idle levels" the right way to put it, rather than zero? "Yes it is fair to write that," Yash Vekaria, the UC Davis computer science PhD candidate who answered for the paper's authors, told ADOTAT. Their read of the small spikes: likely the TV checking in to say it's still idle or still in an app. Asked whether their results fit both LG's statement and RTINGS' 2026 tests, which saw small periodic ACR traffic in apps: "Yes to both."

The paper measured what was sent, not what was matched; the traffic is encrypted. But a TV that sends idle-level traffic while you binge a series is not sending a fingerprint of the series.

One more line from the same study cuts the other way, and it belongs here: the TV makers' own free channels, LG Channels and Samsung TV Plus, drew ACR traffic comparable to live TV. The TV sees its own inventory just fine.

ACR is sold three ways: as an audience to target, as a measurement of who saw an ad, and as the exposure file behind attribution, the "your ad drove this sale" report.

If the exposure file can't see a fifth of viewing, three things happen.

The audience is built from what the TV could see. A household that streams mostly inside the native Netflix app looks lighter than it is.

The exposure file undercounts. An ad seen inside an app that blocks ACR never makes it into the "exposed" group.

Attribution gets graded against an incomplete control group. If some of the "unexposed" households actually saw the ad, the lift math is off, and nobody on the report can tell you by how much.

Josh Chasin, a former VideoAmp executive who now runs the consultancy KnotSimpler, put the buyer's question in one line: "Where are the gaps in your coverage, for both content and advertising, and why?"

Tom Weiss had his own: "What number of TVs have more than X hours a week detected?" A TV in a streaming household that shows up for two hours a week isn't a light viewer. It's a blind spot with a remote.

Ask both. Then ask for the answers in writing.