This post has been updated with a statement from the company.

One thing that people do on the Internet is they make fake images and pass them around as real, for fun. Why do they do this? Sometimes it is funny. Sometimes they are trying to troll people.

That is context for this tweet.

It appears to show that the globally trending #TrumpWon began in Russia — specifically, in St. Petersburg. There's resonance to that. When Adrian Chen of the New York Times explored the world of online Russian trolls, his search brought him back to an office building in that city. The image was quickly picked up on Twitter and passed around. But there's good reason to think it's not real.

The person who tweeted the image said he'd spotted it on TrendsMap.com, a site that catalogues Twitter trending topics. Here's what the map of Twitter trends around St. Petersburg looks like now.


If #TrumpWon trended there earlier, it isn't now. But there's no evidence it did start there.

Update: A spokesperson for TrendsMap confirms that the map isn't theirs -- and that the hashtag started in the U.S., according to their data.

"This is certainly not from any of our tools and do not know of any tools that look this way," Kathy Mellett said in an email. "Based upon our analysis, #TrumpWon primarily came from the US.  There was an initial spike just after the debate followed by a much larger one a few hours later.  In particular, around 97% of the initial spike of approximately 6,000 tweets came from the US." The next highest countries? Canada and the U.K.

Notice for one thing that the map provided by @DustinGiebel doesn't look like the TrendsMap map. The map I posted above is from a premium account for the service; if there's another tool that overlays blue dots, I couldn't find it. What's more, the map in the tweet appears to be based on a Google Maps image, not one from TrendsMap. (Notice the city’s name in both Cyrillic and Latin letters.) If the trend started from a handful of trolls in St. Petersburg, it's odd that they would be geographically distributed across the city instead of all in one location. It's odd that the graph would show a slowly, evenly increasing number of tweets from that location. (The axes are obscured, making it hard to tell what's being measured.)

But TrendsMap does show the timeline of #TrumpWon tweets in St. Petersburg — and they began well after the hashtag took off in Washington, for example. There was a spike as interest in the topic peaked globally, which is what you'd expect in a local region from something that's trending globally. (The graphs in other cities, like London, peaked at the same time. London, though, saw more sustained traffic on the hashtag.)


TrendsMap's list of the top hashtags in the city for the past week doesn't include #TrumpWon, which could just mean that it wasn't as popular there as were some other hashtags. But it also suggests that it's unlikely that so many #TrumpWon hashtags originated in St. Petersburg that it created a global trend.

Screen Shot 2016-09-27 at 4.58.40 PM

We know that online trolls were trying to game the politics of the debate, but that was American Reddit and 4chan users who were trying to move the online polls that sprang up after the debate was over. But this seems unlikely. Over Twitter, Chen told me that "none of the trolls that I have catalogued tweeted #trumpwon," suggesting that if St. Petersburg trolls were tweeting the hashtag, it wasn't trolls that Chen has been tracking. Others reinforced that.

Remember, too, that Trump started using the phrase "Trump Won" in ads even before the debate ended. Most major news sites, including The Post, had big banner ads reading "TRUMP WON" in capital letters. The effort didn't need much help from a few trolls in Russia.

But, again, we only have this one image that suggests that's who was behind it in the first place. It plays nicely into the idea that Russian agents are doing their best to rig the election for Trump, but in this case, the evidence simply isn't there.

The lesson, as always, is this: Nothing on the Internet is real.