After nearly five years of development and tweaking, John Cook and his colleagues debuted their project: a machine-learning algorithm that can detect climate misinformation on the Web.

The algorithm sounds like science fiction: It “reads” sites and flags those with claims presenting false or misleading information about climate change science and solutions.

But something ironic happened around two weeks after the software was made public in Nature Scientific Reports. The algorithm detected a blog post about its own methods, attempting to discredit them.

“It’s quite funny. It was kind of a bit of a meta moment,” said Cook, a research fellow at the Climate Change Communication Research Hub at Monash University in Australia.

The team programmed the artificial intelligence to detect false claims in five categories: global warming is not happening; human generated greenhouse gases are not causing global warming; climate impacts are not bad; climate solutions will not work; and climate science and scientists are unreliable.

Cook said the machine detected at least half a dozen claims about itself in the blog post.

“Ultimately, our goal is the Holy Grail of fact-checking, which is being able to detect and debunk misinformation in real time,” said Cook, who partly developed the framework previously at George Mason University. “Ideally, I would have social media platforms using it to detect misinformation in real time.”

Such a machine learning algorithm may seem unnecessary to fight climate misinformation as climate change increasingly manifests itself in tangible ways — fire seasons of unprecedented severity, record-breaking heat waves, extremely active hurricane seasons and freak cold snaps.

“We now see, for the first time ever, a majority of Americans saying that they think people in the United States are being harmed right now,” said John Kotcher, a professor at the Center for Climate Change Communication at George Mason who is not affiliated with Cook’s study.

According to surveys conducted in September, Kotcher and his colleagues reported that 55 percent of Americans think climate change is harming people in the United States. Fifty-two percent said they personally experienced the effects of climate change. Previously, most viewed global warming as a psychologically distant threat and far off in the future, affecting penguins and polar bears but not necessarily human health.

Now, most are concerned about a range of environmental hazards — among them extreme heat (76 percent are concerned), air pollution (76 percent), droughts (68 percent) and water shortages (64 percent).

Yet Kotcher said the public is divided on whether the issue should be a priority for the president and Congress.

“There’s been huge increases among Democrats, both liberal and moderate ones, over the last few years,” said Kotcher, who also publishes reports investigating climate change along with political ideology. He said there has been a slight increase in support from more moderate Republicans, but trends have essentially flatlined for conservative Republicans, who constitute a relatively larger proportion of the GOP.

And Kotcher said it is safe to say climate misinformation is playing a part in the differences of opinion. But not all climate misinformation is the same or is disseminated in the same quantity, the recent study showed.

Using the machine-learning algorithm, Cook and his colleagues conducted the largest content analysis of climate contrarian claims to date. The team analyzed more than 250,000 documents from 1998 to 2020 from 20 highly visited, mostly U.S.-based conservative think tanks and 33 central contrarian blogs — outlets the study called “key cogs in the so-called climate change ‘denial machine.’ ”

The team found that in recent years, contrarian outlets mainly countered climate information pertaining to science/scientists and solutions.

For instance, some may say the science is unreliable and there is no consensus, even though 99.9 percent of scientists agree that human activity is altering our climate. Others may say the movement is unreliable — environmentalists and media are alarmists; and climate scientists are biased, incompetent or lying. For solutions, common claims were that climate policies are harmful and clean energy won’t work.

The results were surprising to Cook, who has spent the past 15 years addressing science myths — it’s not happening, it’s not us and it’s not real. His previous research focused on how to counter misinformation about the greenhouse effect or the carbon cycle. While those may have been more common targets for contrarian outlets in the early 2000s, the focus has shifted.

“We’re over here fighting this battle and they have already moved over to this other battlefield. And we’re not even there, so that’s a real problem,” said Cook. “There is a dearth of research into understanding attacks on climate science and scientists themselves, let alone developing solutions.”

Other researchers have investigated ways to connect better with more conservative audiences. A study published in June showed Republicans in two congressional districts in Missouri and Georgia shifted their views after watching videos on the topic produced by people with similar conservative moral values. The study said the campaign increased the “understanding of the existence, causes and harms of climate change by several percentage points.”

Last year, Cook released a free game for phones and computers called “Cranky Uncle,” which allows users to play a humorous game where they must spot logical fallacies in scientific statements. He said lessons learned from the game could help address arguments against attacks on climate scientists, science and solutions.

He is also working on integrating the machine-learning algorithm into a browser extension, initially conceived by his students, that could detect misinformation and annotate a webpage in real time.

“Climate is such a holistic issue. It affects every aspect of our lives, so we don’t have to go down the science route. I don’t think that’s necessarily going to be the most effective way to … get people caring about the issue,” said Cook. “There is no one magic bullet, and I think we need to pursue multiple avenues from multiple messages to multiple audiences.”