In the digital age, high-profile crimes often kick off crowdsourced investigative efforts among volunteers eager to use the internet to solve a mystery. But they have rarely sparked this volume of backlash due to the identity of the victim, or this kind of empathy with the killer.

Social media posts and news reports have been flooded with dark jokes and tidings of goodwill for the gunman some have compared to vigilante-justice figures from comic books. One viral comment on a TikTok video of the killing said, “Praying for this man and his family … he must be so scared being on the run.”

The shooter remains on the loose after gunning down Thompson outside a hotel on Wednesday morning and reportedly escaping through Central Park. In the hours since the shooting, police officials have released surveillance-camera footage of the suspect inside a Starbucks before the killing and riding away on an electric bike.

When officials early on Wednesday said they suspected the gunman had fled on a Citi Bike, software engineer Riley Walz quickly pulled up data he had been collecting on the mass movements of New York’s popular bike-share fleet.

Walz, who has used data to visualise YouTube history and map out the changing prices of Big Macs, thought the bikes’ GPS co-ordinates would make a cool online art project. For a few weeks, he had run a script that automatically logged the ID numbers and other details every minute from 16,587 bikes at all of the New York City area’s 2234 docks.

Walz looked up the bikes taken from the six docks closest to the midtown hotel where Thompson was killed and found one that seemed to match the police time frame: bike No. 421-6511, which was pulled just before the shooting and docked eight minutes later near Central Park. He tweeted his findings and alerted the police.

“It would be so cool if I can use this data to figure out where the murderer went,” Walz told the Washington Post in an interview on Wednesday afternoon. “And, like, I’m just some random guy.”

The data, however, was a dead end, and police officials have since said the gunman hadn’t used a Citi Bike after all. Critics laid into Walz not just for getting it wrong but for attempting to help in the first place: An X post with more than 130,000 likes called him “the biggest narc/nerd combo” in the world.

Anant Sinha, who posted a video to X in which he interviewed Walz about the data, said he and Walz were pummelled with online death threats, and he shared a screenshot of his X inbox filled with messages calling him a “pig” and a “snitch”.

Since the killing, users have flooded social media with their grievances over the healthcare industry at large. Some made light of the shooting, celebrating the news and encouraging New Yorkers not to assist police in their search. One meme, showing a smiling star and the words “CEO DOWN,” was shared widely on social media; a TikTok video posted by the Daily Mail claimed a balloon with the meme was also left near the hotel where Thompson was shot.

The shooter was also widely glorified. When the New York police published images of what they said was the suspect’s smiling face online, some people suggested he was handsome.

Thompson’s wife, Paulette, said in a statement to the Minnesota Star Tribune, “Our hearts are broken, and we are completely devastated by this news.”

Policing experts, speaking generally and without direct knowledge of this investigation, said the internet’s wealth of information has traditionally been a gift to modern detective work.

Joseph Courtesis, a former inspector for the New York police who ran its real-time crime centre from 2016 to 2019, said, “Social media, quite frankly, is the greatest thing to ever happen to a criminal investigation,” thanks to possible leads, like a suspect’s old photos and posts.

But there’s a big difference, he said, between unearthed digital clues and the shaky “internet theories” that tend to bubble up online. “If somebody hits on a rabbit hole that they’re going down, yeah, they’ll look at that,” he said. “But most of that can be ruled out in seconds because you know where your investigation is going.”

The early evidence shared by police has largely been composed of images from New York’s vast infrastructure of surveillance cameras, a mix of public and private recording devices that investigators routinely access to identify and track criminal suspects throughout the city.

In 2021, as part of a crowdsourcing project, volunteers with the human-rights group Amnesty International counted more than 25,000 cameras on buildings, poles and streetlights across New York City. Stanford University researchers that year estimated that New York’s camera density was nearly four times higher than Los Angeles.

The scene’s location in one of Manhattan’s busiest districts probably ensured the man was recorded from many angles, said Ralph Cilento, a former commander of detectives with the New York police who retired in 2021 and now teaches police science at John Jay College.

“Midtown is like the Iron Dome of cameras,” Cilento said, referencing the rocket-repelling air-defence system that blankets the Israeli skies. “You cannot get into Manhattan at all now without being caught on camera.”

But finding and gathering all that visual evidence can require considerable effort – and take more time than some sleuths on social media are prepared to give. “They will track the guy all the way through the city,” he said, but “it’s extraordinarily tedious work.”

The crowdsourced online manhunts that have arisen alongside police work have often raised their own concerns because of the chaotic nature of early investigations and the risks of amateurs getting it wrong.

In one infamous example from 2013, users on Reddit wrongfully accused a missing college student, Sunil Tripathi, of participating in the Boston Marathon bombing, comparing his photo to a suspect photographed before the attack and flooding his family’s social media pages with threats of revenge. Later, after finding Tripathi’s body, police said he had died by suicide before the bombing. (The man caught on camera was actually Dzhokhar Tsarnaev, now in Colorado on death row.)

Manhunts following high-profile crimes can drag on for days, even when suspects are quickly identified. It took nearly 30 hours for New York police to arrest Frank James, a man who later pleaded guilty to shooting 10 people in a crowded subway car in Brooklyn in 2022. By that time, James had been photographed by a bystander who tweeted the image; James had even called the police himself, reporting that he could be found in the East Village inside a McDonald’s.

Christian Quinn, a former deputy chief of cyber and forensics for the Fairfax County Police Department in Virginia who now consults on public-sector technology issues, said he suspected the investigators aren’t “really affected by online chatter of … people with opinions who don’t truly have the expertise.”

But the possibility that a random tip could help unravel the killing – or give a clue to the gunman’s preparations or motives – forces them to keep watch, just in case.

“A lot of what you’re going to get is well-intended folks who aren’t really contributing to information,” Quinn said. But sometimes “you get that one little tidbit” that can turn everything around.