They found that while their software could take a decent stab at guessing a tweeter's age if they were between 17 and 30, it was much harder to age older users correctly.
They found that language use in Twitter's 140-character missives barely changes with their advancing years, rendering age guesses over 30 pretty useless.
The team, led by Dong Nguyen at the University of Twente, examined 3000 Dutch Twitter accounts to try to discover the age of users.
Their software was better and faster than humans at guessing exact age, but the mean errors were an unfeasibly large four years.
"We find strong changes in the younger ages, however after an age of about 30 most variables show little change," researchers said in their paper.
The language switch at around age 30 seems related to a change in moving from youthful exuberance - using many capitalised words and acronyms - to more complex language as what the researchers dub 'impression management' becomes more important to the thirty something professional.
These traits include writing longer tweets, with longer words, links and hashtags.
Privacy advocates will doubtless welcome the findings, but there are other ways of guessing factors people might lie about on a social network, such as their age, gender and job.
For instance, Microsoft's lab in Beijing, China, revealed in 2007 how browser histories can tell all. And networks like Facebook, where much longer messages are allowed, give away far more about you than just your age.