President Trump speaks during a joint news conference with Lebanese Prime Minister Saad Hariri at the White House in Washington on July 25. (Jabin Botsford/The Washington Post)

Once upon a time, Donald Trump would talk about his poll numbers constantly. Two years ago this month, he first surged past Jeb Bush in Republican primary polling and his success became a staple of his tweets and his campaign patter. Trump would bring up his poll numbers in the first few minutes of every speech, to huge cheers.

During the general election, that habit waned a bit, as did Trump’s support. Occasionally he’d break out a poll that showed him leading or competitive with Hillary Clinton, but it was more common that polls would be ignored or that Trump would claim that he had a massive flood of silent support. The national polls were right, and he lost the popular vote on Nov. 8 — but, as we all know, that didn’t matter.

Since Jan. 20, Trump’s relationship with polls has soured further. He started out with historically low job approval ratings and, in his first few months in office, watched them slip lower. In RealClearPolitics’ average of approval polls, Trump’s approval rating has gotten stuck somewhere in the 40 percent range for a while.

But there’s one poll that broke that tradition: The generally conservative survey from Rasmussen Reports. Rasmussen’s numbers were always above average, in part a function of the pollsters’ reliance on likely voters, a group that tends to skew to the right.

In mid-June, when Trump’s numbers were under 40 percent in many polls, he trumpeted Rasmussen’s having him at 50 percent.

The month before, an even grimmer citation:

A 48 percent approval rating is … not something a president usually celebrates as good news, given that it means that less than half of the country approves.

Notice, though, what happened to those Rasmussen numbers over time: They started trending down to where the RealClearPolitics average was hovering. Right now, the RCP average of polls has Trump at 39.2 percent approval. Rasmussen has him at … 39 percent.

That’s bad news. This is a Trump-friendly poll, with a Trump-friendly pool of respondents! And, after months of standing apart from the crowd, it’s now reflecting the same view of Trump as everyone else.

Why? The survey shows a significant erosion of a critical metric for Trump. At the outset of Trump’s term, nearly 4-in-10 respondents said that they strongly approved of Trump’s job performance. Now, only 25 percent of respondents hold that view, while nearly half strongly disapprove.

Again, this is among likely voters, a group that’s central not only to Trump’s reelection but to the chances of Republicans in next year’s lower-turnout midterm elections. Highly partisan voters make up a disproportionate share of the primary electorate. Rasmussen’s polling suggests that Trump’s approval with this group is headed south.

In short: Don’t expect Trump to tweet about Rasmussen poll. But, to be fair, it does make that 48 percent result look like something worth celebrating by contrast.