CompTIA's Two AI Hiring Numbers Are Twenty Times Apart

CompTIA reported in January that more than 275,000 active job postings referenced a need for some level of AI skills. In July it reported that nearly 14,000 job postings sought AI and machine learning expertise.

Both figures come from the same organisation, analysing the same underlying source, job posting data from Lightcast. They are more than twenty times apart.

The gap is not a collapse in hiring between January and July. It is a difference in what was counted, and the January release is unusually explicit about it. The 275,000 figure, it says, "included dedicated AI positions, such as AI engineers or AI architects, and jobs that required AI fluency and the ability to utilize and optimize AI tools to perform tasks and boost productivity." That second category is the large one. A marketing manager expected to use AI tools counts toward it; so does an AI architect.

The July figure asks a narrower question, counting postings that seek AI and machine learning expertise rather than postings that mention AI anywhere in their requirements.

There is a second difference, in what is being counted rather than how it is defined. The January number counts active postings, the stock of jobs open at a given moment. The April report counts 55,726 new postings for AI roles or roles requiring AI related skills, added during that month, which is a flow. A stock and a flow are not comparable even when the definition behind them matches.

The broad figure is the one still in circulation. In its August analysis, issued on 4 September, CompTIA said active postings requiring AI related capabilities surpassed 320,000, a 4.5% increase from July, again drawn from Lightcast. No narrow figure appears anywhere in that release. We reported the 320,000 ourselves on 5 September without noting the definition behind it, which is the same thing this piece is describing.

None of this is hidden. Each release states its own wording plainly, and CompTIA footnotes Lightcast as the data source in each. What travels, once a figure is quoted elsewhere, is the number and not the sentence around it. The two shapes then circulate as though they were the same measurement of the same thing.

One limitation applies to all of these counts. Job postings are collected from employer sites and job boards, where the same role frequently appears more than once, and CompTIA does not publish how duplicate postings are handled. The data source is named rather than the method shown, which is the difference between knowing where a number came from and being able to check it.

The practical consequence for anyone reading an AI hiring statistic: ask whether it counts roles that require AI expertise or postings that mention AI at all, and whether it counts jobs currently open or jobs newly posted. Those two questions separate figures that differ by more than an order of magnitude, and they are almost never answered in the sentence that carries the number.