In all cases, the most common (modal) age for women is in their 20s, whereas in images from IMDb and Google, the most common ages for men are 40 years and 50 years, respectively

One of those research papers which raise more questions than it answers but is interesting none-the-less.  From Age and gender distortion in online media and large language models by Douglas Guilbeault, Solène Delecourt & Bhargav Srinivasa Desikan.  All sorts of reasons to dismiss this (essentially related to the inherent possible quality issures of the input data rather than the methodology per se.)  But it is a good college try.  From the Abstract.

Are widespread stereotypes accurate or socially distorted? This continuing debate is limited by the lack of large-scale multimodal data on stereotypical associations and the inability to compare these to ground truth indicators. Here we overcame these challenges in the analysis of age-related gender bias, for which age provides an objective anchor for evaluating stereotype accuracy. Despite there being no systematic age differences between women and men in the workforce according to the US Census, we found that women are represented as younger than men across occupations and social roles in nearly 1.4 million images and videos from Google, Wikipedia, IMDb, Flickr and YouTube, as well as in nine language models trained on billions of words from the internet. This age gap is the starkest for content depicting occupations with higher status and earnings. We demonstrate how mainstream algorithms amplify this bias. A nationally representative pre-registered experiment (n = 459) found that Googling images of occupations amplifies age-related gender bias in participants’ beliefs and hiring preferences. Furthermore, when generating and evaluating resumes, ChatGPT assumes that women are younger and less experienced, rating older male applicants as of higher quality. Our study shows how gender and age are jointly distorted throughout the internet and its mediating algorithms, thereby revealing critical challenges and opportunities in the fight against inequality.

As usual, I get very suspicious when research does not report absolute measures.  How many years younger?  Why isn't that in the abstract?  

The researchers get distracted by the ideological goal of stamping out inequality.  But the research findings stand, independent of the ideological interpretation of those findings.  So what are the actual findings which aren't in the abstract?  

The Abstract does not mention by how much younger are women depicted than men which seems to me a crucial omission.  If it's six months then that is probably within margin or error.  If it is six years, then there might be something interesting.

The report itself gives the answer for celebrities:

Next, we analysed the 2018 IMDb–Wiki dataset43 and the 2014 Cross-Age Celebrity Dataset (CACD)44 consisting of Google Images, each of which provides the true gender and age of the celebrities depicted using their public bio pages and time-stamped photographs. Figure 1 shows that female celebrities are, on average, 6.5 years younger than men on IMDb (t = −169.9; P = 2.2 × 10−16; n = 451,562 images; Fig. 1d), 3.27 years younger on Wikipedia (t = 10.64; P = 2.2 × 10−16; n = 57,972 images; Fig. 1e) and 5.35 years younger in Google Images (t = −90.92; P = 2.2 × 10−16; n = 149,889 images; Fig. 1f). In all cases, the most common (modal) age for women is in their 20s, whereas in images from IMDb and Google, the most common ages for men are 40 years and 50 years, respectively. These analyses show that age-related gender bias online is not an artefact of human perceptions of gender and age, because it is replicated using verified objective information about the age and gender of those depicted. That age-based gender bias replicates strongly in the context of celebrities is concerning, given the salient role that celebrities play in reinforcing stereotypes.

6.5 years is in the interesting category.  

However, the media industry is by repute notorious for favoring young starlets while providing more roles for older men.  So maybe 6 years gap isn't surprising.  What about other occupations and categories.

From the bowels of the report it appears that their other approaches (aside from IMDb) don't have nearly the gap. One or two years seem the modal gap depending on the different methods used.  Certainly within the margin of error given questions about data integrity and methodological sensitivity.  

From potentially interesting research, this now begins to look like cognitive pollution.  They wanted to find a discriminatory bias and they were able to play with the data to get one.  But mostly by building a house of cards, likely to collapse at the slightest breath of inquisition.  

Christian Rudder had far better evidence some fifteen years ago, see One data set can have multiple truths.