Dad's research method fit in a single line. "Always eliminate the phone book first," then "ask, ask, ask". The phone book is gone and the asking now starts in a search box, but the instinct underneath it, trust nobody until you have checked, matters more today than it ever did, not less.
Google did not replace the method, it sped it up
Dad's ring-around, phone every club, record company and publisher until somebody gives you the next name, is mostly a search now. That is a real gain. What took an afternoon of calls can take ten minutes. But look at what he wrote about the internet back when it was the new arrival. He called it "the most fabulous source of information" in the same breath as warning that it carries plenty of material that is "hoax, ill-informed and WRONG". He was already teaching you to distrust the screen. The search box did not change that lesson. It just handed you more screen to distrust.
AI is the newest "most fabulous source", and it lies with a straight face
Dad never typed a word into an AI model. He died in January 2025, before most newsrooms had even written a policy on them, so everything in this section is mine, not his. But he handed us the rule that handles them, and it is the same rule as above. A large language model gives you a fluent, confident, well-organised answer, and a certain share of the time that answer is simply invented. It will make up a quote. It will hang a real-sounding source on a fact that no source ever carried. It will give you a date that is close and wrong. The danger is not that it sounds like a machine. The danger is that it sounds like a competent colleague who never admits to being unsure.
How to use it without being used by it
Treat anything an AI tells you the way Dad treated a tip from a stranger: as a lead, never as a fact. Use it to draft your questions, to summarise a long document you will then read yourself, to suggest who might know. Then verify every name, number, date and quote against a primary source, exactly as you would once have rung to confirm. Never quote a model in your copy, because there is no person behind the words and nobody to ring back. And when it hands you a citation, open the citation, because the most convincing AI errors are the ones wearing a footnote.
The places to do the real checking still exist. There is the institution's own paper trail through freedom of information, the primary record in Trove, and the plain scepticism you should bring to Wikipedia, which is a fine place to start and never the place to stop.
The phone still beats the inbox
Here is the part of Dad's factsheet that no model can touch. His fastest research tool was never a database. It was a friendly phone manner and the willingness to "ask, ask, ask". He made the point that emailing an expert is slow, because busy people let email sit, while a polite phone call gets you a fresh quote in minutes. That is even more true now that every inbox is drowning and half of it is automated. A real person, on the phone, telling you something true, is the one source a model can never hallucinate. So make the call. Then make the next one.
What stays the same
Everything that matters. You still eliminate the phone book first, you still ask around, ask around, ask around, and you still check before you trust. The tools got faster and louder and more sure of themselves, and not one of them removed the job of verifying. If anything, the more fabulous a source claims to be, the harder you check it. Read Dad's original here: Journalistic Research.