Despite getting closer than before in 2024, the public polling industry has consistently underestimated support for President Donald Trump, in part due to more people ignoring unwanted calls and texts from pollsters.
And polling is unlikely to get a helping hand from Big Tech: A new feature in Apple’s incoming iPhone operating system could make it tougher and more expensive to get ahold of respondents.
An Apple press event from June describing new features in iOS 26 like “on-device spam detection” and “the ability to screen new senders” sent panic through the political polling and fundraising worlds.
“The modern problem of polling is reaching people in the first place, and everything that makes it harder to reach a potential respondent is damaging to our ability to poll,” Charles Franklin, director of the Marquette Law School Poll, told NOTUS.
When a filter is turned on, the Messages app sorts texts from unknown senders into a separate inbox. Users can choose to let through certain messages like urgent alerts, verification codes or texts from senders identified by Apple “as not sent by businesses or organizations,” the setting description says. Though a feature in Messages to silence “unknown senders” has existed since 2020, it is off by default and gone largely unnoticed until this upcoming Apple upgrade that includes a new specific folder for “spam.” The company touted the changes in their release of iOS 26.
“Unknown senders will appear in a dedicated folder where users can then mark the number as known, ask for more information, or delete. These messages will remain silenced until a user accepts them,” the company wrote. (Apple declined to comment for this story).
The setting could go far beyond filtering out robocalls or telemarketers. Already political fundraisers are worried that their solicitation texts could get marked as spam.
“Apple’s iOS 26 update introduces aggressive message filtering. Political texts — even from verified and compliant senders — will be treated as spam by default,” the National Republican Senatorial Committee wrote in a memo. “Apple isn’t just targeting cold outreach or spammy actors. Every political message — shortcode, long code, doesn’t matter — gets pushed into the dark.”
More worrying for public opinion polls, pollsters told NOTUS their calls and texts being filtered out could make polling more expensive and obtaining statistically random and reliable results more difficult, especially at the state and local levels.
Franklin has been conducting both statewide Wisconsin polls alongside national surveys since the Marquette poll’s inception in 2012. He said that for about a decade, Marquette tried initially to reach potential respondents by cold-calling, about 99% of which went unanswered. Of that 1% who actually picked up, 35% to 45% would agree to be interviewed for the poll.
“So you can see where the problem lies. It’s not that those 99 people were refusing to do a poll, it’s that they didn’t even know that someone was calling to do a poll,” Franklin said.
For pollsters, this means getting calls answered and texts read is critical. And for every call or text that goes ignored, pollsters have to keep trying to find a respondent. The time, cost and effort only goes up if they are automatically filtered out.
That also means smaller operations and newer candidates might struggle more. Republican pollster Conor Maguire said those campaigns are already more likely to rely on cheaper “snapshot” polling done over texts in a short period of time due to financial constraints, as opposed to lengthier live polls backed by bigger budgets. Texts from small campaigns also may go unread with the new iOS.
“The incumbents are in better shape. Generally, they can raise more money and pay for full, all-live polls,” Maguire told NOTUS. “But a lot of smaller campaigns and campaigns with lower budgets, they’re gonna have trouble affording all live. And it’s gonna raise the cost of these text -message snapshot polls.”
tate-level polling is also more likely to be affected than national, Franklin said, partially because national respondents often agree to be followed up with. It’s “a reasonably low-effort ask to have them input your contact information once they’ve agreed to be polled consistently.”
But for his Wisconsin statewide polling, Franklin finds his sampling from the state’s voter file. So even asking someone to save your number “still leaves out in the cold our state poll,” which is “just one more barrier to getting through to those folks.”
Beyond these constraints, a major concern is how accurate the work-arounds are. Many traditional academic surveys rely on “probability sampling,” which is randomly selecting from a representative overall list like the voter file. If pollsters know the whole voter file, they also know the statistical probability of each person being selected to be called is.
That fact underpins the math that makes random selection meaningfully representative of the whole, larger population.
But another method, one brought up by multiple pollsters on getting around the iOS 26 feature, doesn’t involve that. This “non-probability sampling,” like opt-in online polls, is when pollsters don’t know what the odds are that someone taking the survey could have been picked.
That kind of polling is generally cheaper and more accessible. But those polls also “just are not representative of the population as a whole,” Democratic pollster Adam Carlson told NOTUS.
“We’re always going to be leery about people that are self-opting into something,” Maguire said. “That’s not great scientifically, as we want a real cross section of whatever district or country that we’re researching.”
If you’re looking at a voter file, “you’re trying to contact people, and you can target people by different demographics that you’re lower on, that are hard to reach. That’s where it’ll be much, much more difficult to do,” Carlson added.
People opting in to a poll online are also reaching out themselves and are more likely to respond to a call or message they signed up for. Randomly selecting people, to align with probability sampling — the preferred statistical method of many pollsters — doesn’t allow for that.
That means this feature in iOS 26 might uniquely threaten the more accurate selection methods that pollsters like Marquette rely on.
“To the extent that this new Apple policy makes it harder or more expensive to do good, high-quality probability samples, it puts a little bit more of a thumb on the scale in favor of non-probability samples,” Franklin said. “Those of us that are committed to probability samples look at this as a bad thing. Anything that makes high-quality probability sampling harder to get makes the polling industry as a whole a little bit less reliable.”
However, not all hope is lost for the polling industry, which is already looking for ways to adapt to new technologies. Carlson said some of his peers have spoken to their text vendors, who seem more optimistic about iOS 26.
“It could make things more challenging, for sure, but it’s not quite as devastating as it seems at first glance. Users have to opt in,” Carlson said. “The default on iPhones will still be what it is now, which is not separating out known and unknown numbers.”
But even those who aren’t expecting the entire industry to be reshaped are still wary.
“I do think with the amount of people that are going to take the steps to sign up for it right now, it’s not going to be a real drastic, fundamental change to the business,” Maguire said. “But I do think it’s going to increase costs all around.”
Nuha Dolby is a NOTUS reporter and an Allbritton Journalism Institute fellow. Samuel Larreal is a NOTUS reporter and an Allbritton Journalism Institute fellow. This story was produced as part of a partnership between NOTUS and San José Spotlight.