4 fact-checking tips for the age of misinformation and AI hallucinations
Savvy PR pros can follow these four best practices to ensure their content is accurate and keep their org’s repuation intact.
“Research is formalized curiosity. It is poking and prying with a purpose. It is a seeking that he who wishes may know the cosmic secrets of the world and they that dwell therein.” — Zora Neal Hurston, Dust Tracks on a Road (1942)
AI can summarize a 50-page whitepaper in seconds. It can also confidently tell you that Aristotle invented Wi-Fi. Misinformation has always been a problem, but now it’s faster, shinier and harder to catch — especially when AI-generated content looks just plausible enough to slip past a busy comms team.
For comms and PR pros, a single bad fact in a blog post, press release, executive statement or even LinkedIn caption can snowball into a credibility crisis. And it’s not just AI—outdated statistics, misquoted sources and PR-driven “research” make it easy to spread bad information, even with the best of intentions.
Here’s how to research and fact-check like your reputation depends on it (because it does).
1. Study your studies.
“A recent study shows…” is not a get-out-of-jail-free card.
Studies are only as good as their methodology. Some are rigorous and peer-reviewed. Others are based on a survey of 100 people and sponsored by a company with a financial interest in the results. If you’re going to cite data, make sure it holds up under scrutiny.
Best practice: Always dig to find the original study, not just a press release or a summary that Yahoo covered. Look at the sample size, methodology and who funded it. Might those backers have hidden motives or an interest in a certain outcome?
In practice: Let’s say you and an executive are about to submit a thought leadership piece to Forbes saying “90% of customers prefer brands that take a stand on social issues,” based on a widely cited report. You track it down and realize the survey was conducted by a PR firm… on a sample of 150 people who were already brand loyalists. Now you don’t have to explain that one away later.
2. Check statistics for a pulse.
Narrowing down into the weeds of ill-gotten research, make sure you’re not accidentally giving new life to a zombie statistic when you scrape it from Google’s AI summary. But even if you’re looking beyond a quick Google, approach conveniently applicable stats with skepticism and scrutiny. Some stats refuse to die, even after they’ve been debunked. The notion that human attention spans have plunged to lower than that of a goldfish? False, but discoverable in some usually reliable sources. The idea that we only use 10% of our brains? Also false. Just because a number sounds compelling doesn’t mean it’s real.
Best practice: Use a fact-checking tool to verify — Penn State has a collection of tools here for different kinds of media — or even search for the stat plus the words “debunked” or “fact-check.” If you can’t trace it to a credible source (such as research from reliable nonpartisan think tanks like Pew Research, academic journals, peer-reviewed scientific studies from prestigious institutions, research from government agencies published before 2025), don’t use it.
In practice: The CEO wants to say, “80% of adults under 30 get their news from TikTok.” A quick fact-check shows the real number is closer to 40%. Good thing you checked before a reporter did, or your CEO may have risked undermining their own point.
Or perhaps you find a fintech study that claims “97% of consumers want blockchain-powered banking.” Sounds impressive — until you check and find out the study was commissioned by a blockchain startup and surveyed its own users. You cut the stat before it makes it into your messaging.
3. Beware visibility bias and the weight AI tools place on information.
Speaking of Google’s AI summaries, please, for the love of all that is holy, do not rely on your search engine’s AI summary without clicking through to the original source. It has an unfortunate habit of scraping from artfully SEO’ed blog posts and the like.
And if you’re working directly in a tool like Perplexity or ChatGPT, ask it for sources, and click through to read them.
AI can summarize, but it can also improvise. It has no problem making up sources, misattributing quotes, or filling in missing details with pure fiction.
Another common misstep: If you ask an AI tool to generate a case study using information rounded up from your organization, it might fabricate numbers, data points and quotes that you didn’t prompt it with because it thinks case studies need to include data and quotes. Make sure all of the info in its output is your info.
Best practice: If AI gives you a statistic, claim or quote, don’t trust it until you verify it. Track down original sources and read them yourself.
In practice: Suppose AI-generated media monitoring says a journalist reported your company is “under federal investigation.” You check the actual article — it says regulators are reviewing the entire industry, not your company specifically. That’s the difference between an informed response and an unnecessary PR crisis.
4. Learn to discern deceptive data viz.
A well-designed chart can make data look authoritative — even when it’s misleading. Watch for tricks like truncated Y-axes (which exaggerate differences), cherry-picked timeframes and graphs that imply causation where there’s only correlation.
Best practice: Always check the full dataset and original source before using any chart. If a graph looks too extreme, zoom out and see what’s missing.
In summary, to build a quick-and-dirty fact-checking habit, before hitting “publish,” ask:
- Where does this come from? Primary source or someone else’s summary?
- Who benefits from this claim? Unbiased research or marketing spin?
- Has this been debunked before? Google “[claim] + debunked.” (And then double check the source doing the debunking too.)
- Is this the full picture? Or just a selectively framed part of it?
If you can’t confidently answer those, don’t put it in writing.
In comms, credibility is currency, and misinformation (and AI hallucinations) can dismantle that credibility instantaneously. The only way to stay ahead is to fact-check like a journalist—challenge everything, verify before you publish, and never assume a number is real just because it sounds good.
A few additional sources to boost your mis- and disinformation detection skills:
- Columbia Journalism Review on AI and misinformation in news today
- University of Washington’s consumer-facing research guides to media literacy
- Oklahoma State University’s guides to “pre-bunking” and detecting mis- and disinformation
- SPJ’s Code of Ethics
For several years now, I’ve been advising (cautioning and warning) clients and others in PR to not entirely rely on AI tools to do their work. I will recommend they read this important, excellent, informative, and quite useful PR Daily story.