The problem is volume of data - for example, “pressure cooker” or “fertilizer” is likely too broad. However, some items, like “acetone” are a lot less likely to be googled and have a more specific application, but may still be too broad. (Although when I was young, the pro shop where I caddied used acetone to wipe down the fibreglas golf carts. And it is actually nail polish remover. ) Trying to match multiple different hits to an IP address is more likely to get you a large corporation or a hotel or Starbucks.
I have even googled the floor plans of the White House, while reading one of the recent tell-all books. To find, surprising to me, the Oval Office is not in the middle of the traditional original building as you might thing from the visible shape. Nor is the “West Wing” the west half of the original building. I hope my enlightenment did not set off flashing red lights with the FBI.
They can ask for lists of who did certain searches. They might even use cross-matching and AI to narrow the list down. But ultimately, if the number of hits is too big, the authorities will not have the manpower to look at all the suspect hits.
There’s also the "flash crowd’ effect. If some simple term is mentioned by the news, or a popular web article, then likely hundreds of thousands of people will be googling that for weeks. Did you look up the house and street where that couple threatened BLM protesters with their guns? I did. (Months later, thanks to another news article - they were arguing over the back yard property line) Doesn’t mean I’m going to do anything to them. .