A certain reality TV star and presidential candidate needs a reality check.
On ABC’s Good Morning America Wednesday morning, Donald Trump remarked that “women don’t like Hillary.” Furthermore, he claimed that women “gave [him] great marks and a great percentage, but women do not like Hillary Clinton.”
While much has been written about the fact that Hillary Clinton has faired poorly with Millennial women voters, that matchup has been between Clinton and Sen. Bernie Sanders (D-VT). Saying that women don’t “like” Clinton and that they give Trump “great marks” is very, very far from being the “very, very true statement” that Trump claims it is.
Sign up: Click here to subscribe to the Broadsheet, Fortune’s daily newsletter on the world’s most powerful women.
In November of 2015, polling agency Morning Consult began to ask voters who they would vote for in a contest between Clinton and Trump. Female respondents consistently picked Clinton.
While Trump never managed to edge out Clinton with women voters, it has gotten close in a couple of instances. On February 11th, two days after the New Hampshire primary, Clinton lead the Morning Consult poll by just one percentage point (42% vs. 41%). Sanders won the Granite State by a 22-point margin, causing many in the media to question Clinton’s inevitability and point out her weaknesses, which may have had an impact on voters’ opinions.
However, just two months later, on April 11th, Clinton lead Trump with female voters by an 18-point margin. On that day, Clinton was in New York City, eating street food and making gaffes—could it be that saying the wrong things and not being able to swipe a Metrocard actually made her more relatable?
As for Tuesday night’s primaries, Trump was wrong about those, too—at least in the three states where exit poll data on gender was available. Clinton won 68% of female votes in Maryland, 60% in Pennsylvania, and 57% in Connecticut, according to ABC‘s exit poll analysis. Meanwhile, Trump won 50%, 55%, and 54% of female voters in those states, respectively.