Nano Interactive, a London-based company that sells advertising audiences built without cookies or personal identifiers, published survey findings on September 22, 2026 stating that 51 percent of undecided registered US voters dislike being tracked online. The release landed 42 days before the November 3 midterm elections and framed its results as a problem for the $13.8 billion it says political advertisers will spend this cycle, a figure it does not source.

In Short

A company that sells cookie-free ad targeting asked American voters how they feel about being followed around the internet, and says about half of those who have not picked a side dislike it. That matters because political campaigns are about to spend billions on ads aimed at exactly those undecided voters, often using tracking to find them. Nothing changes overnight, though: the survey does not say how many people answered or when, so the figures stand as the company's case for its own product rather than an independent measurement.

Five figures, one comparison

Nano Interactive commissioned a survey of registered US voters "across party lines" on online privacy and digital advertising, according to the company. Among respondents it classifies as undecided, 51 percent said they dislike being tracked online. Thirty-eight percent said the use of their personal data for advertising does not feel safe, and 31 percent said it feels like an invasion of privacy. Asked about cookie prompts, 37 percent said they do not accept them.

The fifth figure carries the comparison on which the release rests. Eighteen percent of undecided voters said they actively go out of their way to avoid personalised ads, according to the company, against 14 percent of Democrats and 16 percent of Republicans.

Niall Moody, Nano Interactive's chief revenue officer, went further. "Undecided voters are an example of a larger cohort of audiences who are missed by the targeting methods of many campaigns," he said. "More than a third are declining cookies, and nearly one in five are actively avoiding personalized ads. A meaningful portion of this audience has already opted out of conventional digital targeting. Campaigns built on cookie-based data are, in many cases, missing a large portion of audiences."

The release describes undecided voters as "the segment most opposed to being tracked online." The published evidence for that ranking is a single metric, with gaps of four points over Democrats and two over Republicans. For the other four figures, no partisan baseline is given. Is 51 percent high? Without the equivalent Democratic and Republican numbers, there is no way to tell.

Read from the other side, the same figures describe a majority that has not opted out. Forty-nine percent of undecided voters did not say they dislike tracking. Sixty-three percent did not report refusing cookies, and 82 percent did not say they avoid personalised ads. By the company's own count, most of the audience it describes remains within reach of the methods the release characterises as failing.

What the release leaves out

Nano Interactive has not published a sample size. Nor does the release give fieldwork dates, the firm that ran the survey, the panel it drew on, how responses were weighted, or a margin of error. The definition of "undecided" is absent too: the release does not say whether respondents were undecided about a House race, a Senate race, party identification or whether to vote at all.

Those omissions are not incidental to a two-point gap. Were the undecided subgroup 300 people, the sampling error on an 18 percent result would be roughly plus or minus four points at the 95 percent confidence level, wider than either margin the company reports. A larger subgroup would narrow that band; the release gives no way to know which applies.

Other political surveys this cycle published their mechanics. Adtaxi's 2026 Economic and Political Survey polled more than 1,100 US adults through SurveyMonkey, and the VAB's research on voter trust in television news drew on a December 2025 Dynata survey of 2,319 US respondents. Disclosure does not guarantee accuracy, but it lets a reader judge scale.

Demographics are the other unknown. In the YouGov data the release borrows, 74 percent of Americans aged 65 and over called limiting access to their personal data very important, against 47 percent of those aged 18 to 29, according to YouGov. If the undecided group skews older or younger than the partisan groups, age alone could explain part of the gap the release attributes to indecision. No age breakdown was published.

Borrowed benchmarks of different vintages

The release cites two outside sources as evidence of "a broader national sentiment." Both figures are genuine; they are not equally recent.

The YouGov number, 61 percent of Americans saying that limiting access to their personal data is very important, comes from YouGov Profiles data published for Data Privacy Day in January 2026 and drawn from more than 400,000 consumer responses, according to YouGov. The same dataset carries a qualifier the release leaves out: a third of Americans said they take only moderate care in protecting their data.

The Pew Research Center figure is older. Pew's finding that 81 percent of US adults are concerned about how companies use the data they collect about them comes from a survey fielded from May 15 to 21, 2023 and published on October 18, 2023, according to Pew. The comparable figure in 2019 was 79 percent. A lawsuit against the AI note-taking app Granola cited the same Pew figure, alongside a 2021 industry analysis finding that 94 percent of US users declined cross-app tracking once Apple's iOS 14.5 update began requiring permission.

When Apple's single system-level prompt offered a clear refusal, refusal rates reached the 90s. When voters are asked to recall their own behaviour at website cookie banners, the self-reported rate in this survey is 37 percent. Stated attitudes, recalled behaviour and observed behaviour are three different measurements, and the release relies on the first two.

Whose $13.8 billion?

The release does not say where its spending projection comes from. The number matches an estimate from Madison and Wall, reported by MediaPost on May 20, 2026, which put midterm-focused political advertising at $13.8 billion, 9 percent above the 2022 midterms. That analysis assigned 50 percent of political advertising to digital platforms, up from 46 percent in 2024, with local television's share falling to 36 percent from 40 percent, according to MediaPost.

AdImpact, a political ad tracking firm, works from a lower total. Its projection stood at $10.8 billion in September 2025 and was raised to $11.6 billion in June 2026, according to CNBC, with $5.6 billion expected on broadcast, $1.4 billion on cable and $1.68 billion on digital. Its projection for political spending on CTV, the streaming side of television, rose to $2.7 billion from an earlier $2.4 billion, and MediaPost reported that political spending had reached $4 billion by early June.

The two houses slice the market differently. Madison and Wall's 50 percent digital share implies roughly $6.9 billion, about four times AdImpact's digital line. Neither figure measures how much political money depends on tracking, and a large part of either total buys broadcast and cable television, where no cookie is involved.

Timing favours late sellers. Roku said its second-quarter political advertising exceeded the comparable quarter of the 2024 presidential cycle, with spend building into the fourth quarter. Nano Interactive's pitch arrives inside that window.

Refusing a cookie banner has narrower consequences in the United States than the release implies. US state privacy laws are built largely on opt-out rights rather than prior consent, so a consent prompt of the European kind is not a legal requirement on most US websites. Google confirmed in April 2025 that Chrome would keep third-party cookies, then retired most of its Privacy Sandbox technologies in October 2025. Safari and Firefox block third-party cookies by default. Exposure to cookie tracking depends on the browser as much as on what a voter clicks.

Automated refusals are another route. Global Privacy Control, a browser setting that sends a one-bit opt-out with every request, carries legal force in a growing number of states. Google began treating the signal as a trigger for restricted data processing in Delaware and Oregon in November 2025, joining nine other states including Colorado, Texas and New Jersey. Compliance is uneven. A webXray audit of 7,634 popular sites, run from a California address in March 2026, found 55 percent set advertising cookies despite an active signal.

The larger gap concerns how political campaigns actually target. Voter files are public records in the United States, and much political digital targeting starts from them rather than from browser cookies. Deep Sync and MiQ said in April 2026 that matching voter file records and onboarding them for activation could complete in under an hour in many cases. JamLoop's ActiveVoter, opened in February 2026, draws on L2, i360, Aristotle and TargetSmart files and claims access to more than 125 million US households.

None of those pipelines asks a voter to accept a cookie. A household can refuse every banner on every news site and still be matched from a voter file to a television set through an identity graph. Nano Interactive's cookie figure measures one channel. Its framing of "cookie-based data" describes a subset of the political stack, not the whole of it.

The survey does not ask whether undecided voters object to those methods too, arguably the more consequential question for campaigns. Identity-based targeting has accuracy problems of its own, which lends Nano Interactive's argument some weight: Truthset's State of Data Accuracy 2026 report found age segments accurate 13 percent of the time and gender segments 61 percent, while IP-to-postal address linkages tested at 13 percent.

What Nano Interactive is selling

The second half of the release is a product pitch. According to Nano Interactive, its proprietary LIIFT platform reaches the audiences identified in the research across channels including CTV, using off-the-shelf persona segments built for undecided voters. The company says those segments are developed with human-verified panels that combine AI-driven signal analysis with human validation, so advertisers can activate them without cookies, tracking or personal data. It adds that its NanoQ tool lets advertisers build custom audiences to add directly to a political campaign stack.

The release states that LIIFT "has been proven to effectively reach" these audiences, but names no campaign, metric, measurement partner or time period. Panel sizes, segment counts, pricing and the demand-side platforms carrying the segments are not disclosed.

Moody set out the technical premise. "Advances in artificial intelligence have expanded what's possible in audience targeting. Instead of relying on personal data, AI can now identify interest based on real-time contextual signals, what someone is reading, watching, or searching for in the moment, and use that to inform ad delivery without collecting personal information," he said. "Our platform processes more than 4.9 billion real-time signals daily to understand audience intent as it forms, without cookies, tracking, or personal data."

The 4.9 billion figure is a standing claim on the company's technology page, which does not define a signal. NanoQ predates the release as well: Nano Interactive's website describes it as an agentic media planner drawing on a library of more than 800 pre-built and custom segments. The panel method is older still. In June 2024, according to ExchangeWire, the company said its segment taxonomy was verified with panel research.

What is new is the political packaging. The company, which says it operates in more than 100 markets, has been extending the approach into television. On August 11, 2026 it released Screen Graph, which merges partial CTV content records held by different publishers into single listings, with no coverage figures, partners or pricing disclosed. Earlier in the year, Index Exchange listed Nano Interactive among the partners building on Index Cloud, its containerised platform, then in closed beta.

How a panel-built persona reaches anyone

The release does not explain the mechanics. Contextual and persona products of this kind do not find an undecided voter and serve that person an advertisement. They find content where a defined group is more likely to be present, and buy impressions there. Nano Interactive's website refers to ratio matching for its persona targeting. A panel establishes which content a group consumes; a model then scores live inventory against that pattern.

The output is a composition, not a list. Some impressions delivered against an undecided-voter persona will reach undecided voters; the rest will reach others consuming the same material. The release gives no composition rate, the figure that would let a buyer compare a persona's efficiency with a voter-file match.

Television adds a further constraint. Contextual methods on streaming inventory depend on knowing what is playing, and Peer39 analysis cited by Gracenote found that only 40 percent of CTV bid requests carry usable programme-level signals. A persona built on content signals is limited by the share of inventory that describes itself. Ogury extended persona-based targeting to CTV in February 2026, combining zero-party survey data with contextual signals rather than individual identifiers.

A crowded stack for a short season

Nano Interactive enters a political market that ad tech vendors have been building towards all year. Comscore and Yahoo DSP opened Proximic Political Audiences on March 5, coordinating linear and streaming exposure for contested races. Nexxen brought ADvolution's influencer fandom segments into its DSP on May 4. Basis and DAX gave political buyers exclusive programmatic access to DAX audio inventory, with L2 and Experian voter data, on May 18, and Digital Remedy assembled a political desk with more than 17 data partners three days later. In June, PharosGraph and AdImpact paired ad tracking with narrative measurement, and Gray Media connected its 117 local markets to Madhive's AI DSP.

Most of those products add data to identity-based targeting: richer voter files, faster matching, more household signals. Nano Interactive argues the reverse: that part of the audience has left those methods and must be reached through context. The two are not mutually exclusive, though the release presents them as alternatives.

Why this matters for the marketing community

The direction of Nano Interactive's findings matches years of public data on American privacy attitudes. The magnitude, and the comparative claim about undecided voters, cannot be checked. A survey without disclosed methodology is a positioning document, and positioning documents tend to reappear in planning decks as though they were measurements.

The commercial pressure behind the release is real. When political budgets fill streaming supply from September through October, brand advertisers face tighter availability and higher CPMs on the same inventory. Every vendor with a political product has an incentive to publish something before November 3, and six weeks leaves campaigns little time to test a new audience source against an established one.

Publishers face a different question. Readers who refuse consent, block cookies or send opt-out signals still generate impressions; what changes is how those impressions are priced and who can buy them. The shrinking pool the release describes is, in industry terms, a question of addressable reach, the share of a target audience that a given method can recognise at all. Contextual and persona demand is one of the few ways non-consented traffic is monetised at scale, which gives vendors in that category a reason to describe the refusing audience as large.

For brand advertisers outside politics, Moody's description of undecided voters as an example of "a larger cohort" is the most portable claim in the release, and the least supported. The data behind it covers one audience, in one country, ahead of one election.

Timeline

Summary

Who: Nano Interactive, a London-based targeting company that says it operates in more than 100 markets, with chief revenue officer Niall Moody providing the statements. The survey concerns registered US voters, with results reported for undecided voters, Democrats and Republicans.

What: A commissioned survey reporting that 51 percent of undecided voters dislike being tracked online, 37 percent do not accept cookies when prompted, 38 percent say the use of their data for advertising does not feel safe, 31 percent call it an invasion of privacy, and 18 percent actively avoid personalised ads against 14 percent of Democrats and 16 percent of Republicans. The release pairs the findings with a pitch for the company's LIIFT platform, undecided-voter persona segments built on human-verified panels, and its NanoQ tool. No sample size, fieldwork dates, survey vendor, weighting, margin of error or definition of "undecided" was disclosed, and the $13.8 billion spending figure is unsourced.

When: The survey was published on September 22, 2026, 42 days before the November 3, 2026 US midterm elections.

Where: The United States, across display, video and CTV inventory, in a political market where AdImpact projects $11.6 billion in total spend and $2.7 billion on CTV, and Madison and Wall projects $13.8 billion.

Why: Nano Interactive argues that campaigns relying on cookie-based targeting miss a growing share of privacy-conscious voters, and that contextual persona targeting can reach them instead. The comparative claim rests on one published metric with gaps of two and four points, much political targeting runs on voter files and household identifiers rather than browser cookies, and the company's own figures show most undecided voters have not opted out.