Running Political Ads on Google That Spread Disinformation Is Still Really Easy
Why won’t Google change the algorithm?
--
In July, I performed an experiment to see how easy it was to run ads on Google that made false claims about Joe Biden.
First, in the Google Ads system, I bought the keyword “should I vote for Biden?” Then I told Google I wanted to run this ad:
Within an hour, Google accepted the ad, and it was up and running. It served (usually in the top position) on the results pages for people across the United States when they Googled “should I vote for Biden?”
Here’s a slight variation of the ad, in the wild:
The top line of the ad — the part that’s blacked out — is the name of the website where I chose, in the Google Ads system, to direct clickers of the ad. In this example, I sent clickers to a conservative news site. (The news site didn’t actually write the ad, but to the searcher, it seems like it did.)
After I ran the Biden “pedo ad,” I ran other disinformation ads about him. One ad claimed he would destroy the country. Another said he would ruin Americans’ savings.
Then I ran the same ads but for the search “should I vote for Trump?” I directed these clicks to left-wing sites.
Then I ran more general political ads. One claimed Election Day was postponed or that it was to be held “virtually.” Another offered made-up details about a second stimulus check. Another established fake meetup locations for Black Lives Matter protests — this one was triggered when people Googled “BLM protest near me.”
In every experiment, I ran the ads just long enough to make sure Google accepted them. I paused them before too many people saw them.
I was allowed to write nearly anything I wanted in the ads, and I was able to send the clickers to nearly any website, lending authority to the ads’ bogus claims.
The ease of this made it feel like there was no check from Google on what I was doing, no hurdle or loophole I had to pass through. It seemed like anything was fair game when it came to political ads (or “election ads,” as Google calls them).
This surprised me. I expected the ads to be disapproved right away. After all, ever since the 2016 election, Google was supposed to be policing election ads more closely. For example, in 2020, if you want to run an election ad on Google, you’re supposed to first verify your identity. But I never had to do that.
Google’s policy says election ads include those that feature a candidate, officeholder, party, or ballot measure. My Biden or Trump ads didn’t contain the words “Trump” or “Biden.” They mentioned “Joe” or “Don” or “Donald.” (When I tested ads mentioning “Trump” or “Biden,” they were indeed disapproved.)
The policy was flawed and easily exploitable because Google was letting ad copy decide an election ad, rather than the intent of the searcher — the words typed into Google.
On its face, the policy makes sense. An ad that says “you shouldn’t” isn’t political. But it is political within the context of the search intent “should I vote for Biden?”
When Google learned of my experiment, in August, the company vowed to step up its evaluation of websites where election ads direct to. But it didn’t say anything about closing the loophole of a searcher’s intent.
Then Google went into my account and disapproved everything and stopped the ads. I saw a new label next to the ads: “Election advertising in the United States.”
I let some time pass. In early October, I turned the Biden ads back on, just to see whether they’d be accepted again. Some of them were. They ran for a couple hours before I paused them. Then I ran them again yesterday, a week out from Election Day. It looks like Google never changed the policy — I was directing the ads to the same conservative news sites, and it still didn’t matter.
Due to the quirk of (still) being able to direct ad clickers to any website, one can pit two entities against each other — two news outlets, for example. One might run an outlandish fake news ad as if it had been written by a right-wing news outlet, and then pitch a left-wing outlet with a screenshot of the ad in the wild. All the while, neither party knows who is really responsible for the ad and the public takes the ad’s claim as truth. (This is a 2020 take on a tactic that marketer Ryan Holiday used to manipulate the media to get exposure for his clients — companies like American Apparel.)
In September, Google doubled down. It announced that it would ban election ads altogether for a week after the polls close on November 3. That some of my Biden ads were again accepted this month gives me little hope that Google will be able or willing to enforce this. Why leave loopholes at all? Why not just ban election ads outright and keep them banned? Is it that Google can’t change its algorithms in time for the election? Is it that there’s too much money on the line? By some accounts, Google stands to make only a few hundred million dollars from election ads on search this election. Is it that the company doesn’t want to ruffle the feathers of the wrong beast?
Whatever the reason, I think Google needs to ban any type of political ad until:
- Anyone can’t direct anyone’s click to any website.
- Anyone can’t claim an ad is written by any entity.
- The words people type into Google determine whether an ad is political or not.
- The majority of people can differentiate an ad from a non-ad.
- People view search engines as marketplaces rather than information resources and approach search engine ads with the same skepticism as social media ads.
It doesn’t take millions of dollars to bankroll disinformation ads on search engines. Google’s ad system is a pay-per-click system. The ad runner pays Google only when their ad is clicked. But the ad can serve tens of thousands of times for just a few hundred bucks.
