The researchers took 52 college students and got them to complete a standard NEO personality questionnaire. They also had to state the date, time and place of their birth.
Three weeks later, the participants were then given two personality summaries - one based on the personality tests, and one on their astrological chart generated with a computer program.
The trick was that everyone also got a pair of bogus summaries, one of each kind. These were simply someone else's results, picked at random from the other 51 volunteers. They weren't told which were the fakes and which were real - they had to work it out, based on which one matched them best.
The results showed that the subjects were no better than guessing when trying to tell which of the two astrology charts was theirs. They were able to pick their own personality scores better than chance, although only 80% of them got it right, and guesswork gets you to 50% - so this is not all that impressive. . . .
This study is a modern update of Shawn Carlson's classic 1985 Nature paper, A double-blind test of astrology. In Carlson's experiment, though, people weren't even able to accurately pick out their own personality scores.
When asked to say which of the four reports was the best match overall match to their personality, 55% of the participants picked their own real personality one - but no fewer than 35% preferred one of the astrology charts, and 10% went for someone else's personality scores.
I note that the sample size is ridiculously small for a study where the WEIRD sample is so easily gathered. Also, query if the results would be different if people were picking a personality test based on the assessment of someone who knew them well, rather than one based on their own answers.