Much to her surprise, New York magazine reporter Olivia Nuzzi found herself sitting in the Oval Office on Tuesday with President Trump, Vice President Pence, Secretary of State Mike Pompeo, White House Chief of Staff John F. Kelly and press secretary Sarah Huckabee Sanders. Sanders had asked Nuzzi to come speak with Trump; the others sort of materialized as the conversation progressed.

The subject soon turned to how well Trump’s presidency was going, as it does when Trump is talking to a reporter or, really, any other human.

“Doing four rallies this week,” he said. “I think the rallies have, frankly, built up our poll numbers very greatly. What am I now in Rasmussen? 52?”

(Trump has held five rallies this month, excluding one Wednesday evening.)

Sanders confirmed that, yes, Trump was at 52 percent approval in Rasmussen Reports' polling. He wasn’t actually; he was at 51 percent, a figure that has since fallen to 49 percent in polling conducted by Rasmussen, which has consistently generated Trump’s best numbers.

He continued.

“Plus there’s 10 percent, they think, where people don’t respond, unfortunately,” he said. “I’m not sure if this is nice or not nice, but when they don’t respond, that means it’s an automatic Trump vote. But it’s a 52, and we’re doing very well in the polls.”

This is a remarkable claim.

Trump’s relationship with polling has consistently been dodgy since the spring of 2016. Before that, the polls that mattered were Republican presidential primary polls, and Trump was dominating. So at campaign stops, Trump would often pull a sheet of paper out of his jacket pocket to recite the latest numbers — or just relay them from the top of his head.

But then the primary ended and national and state polls consistently showed Trump trailing. Suddenly, polling became suspect, with Trump and his allies theorizing that his support wasn’t being accurately or fully measured. Trump voters were wary of talking to pollsters, the theory went, or they lied to pollsters from the hated mainstream media. On the national level, though, the 2016 polls were on the money: He trailed Hillary Clinton in the polls and then trailed her in the popular vote.

Once inaugurated, Trump’s rhetoric around polling shifted a bit. While consistently lifting up Rasmussen’s results (usually once those numbers passed 50 percent before receding again), Trump would insist that smart analysts recommended tacking on a few percentage points to more accurately capture opinions about him.

At a rally in North Dakota in June, Trump made this point.

“By the way, our people, they call it the base, they used to say it’s 35,” Trump said, referring to the percentage of the country. “Then they said it’s 40. Then they said it’s 42. Then they have these polls go — we’re driving them crazy. Now they say it’s over 50 percent. And then they said — some great people — they said, ‘Anytime Trump gets a poll, add 12 to it.’ Really."

Who said this, if anyone, isn’t clear, nor is it clear why 12 percentage points should be added.

In Michigan in April, Trump gave some more detail.

“Some genius analyst said, ‘But he’s got at least 10 percent of the people that don’t want to say they are voting for him,’” he said. “And you know what I say to that? We’ll take them anyway, whatever it takes.”

So we come back to Trump’s most detailed version of this missing-support idea, offered to Nuzzi: 10 percent of people don’t respond, and “when they don’t respond, that means it’s an automatic Trump vote.”

Trump pretty clearly doesn’t mean the people who don’t respond to pollsters' calls. That’s far more than 10 percent of the calls made, as anyone watching the New York Times’s real-time polling can attest. He appears instead to mean the 10 percent (or however many) who tell pollsters that they have no opinion on whether they approve of Trump’s job performance.

Take CNN’s recent poll with SSRS. It found that 41 percent of Americans approved of Trump and 52 percent disapproved — with 7 percent offering no opinion. Apparently the president believes that those 7 percent get added to the 41 percent because they are “automatic Trump votes.” (Note that Trump is conflating approval ratings with voting here for some reason.)

Is that fair? Well, no. Some of those “no opinion” respondents might quietly support Trump to some extent, sure. Others, though, may pay only passing attention to politics and don’t feel comfortable expressing a strong opinion. “No opinion” is also often a way station between support and opposition. There may be a lot of people who supported Trump but were questioning that position or who opposed him and are giving him a second look but haven’t settled on an opinion.

It is certainly not the case that they should be wrapped up in a bow and added to Trump’s total.

What Trump is doing here isn’t subtle. He’s trying to bolster his low approval ratings (lower at this stage of his presidency than for any president besides Harry Truman, according to FiveThirtyEight) by doing some clumsy sleight of hand. Trump’s goal is rarely to accurately represent himself and his administration, and it’s often to inflate how well he and it are faring. Approval numbers are frustratingly precise for his needs, so he introduces some uncertainty.

Nuzzi, to her credit, was clearly not convinced.