I scraped 17,255 open AI jobs. Only 43.5% are engineering roles.

I've been pulling AI job listings straight from company career pages rather than reposting aggregators, and the function split surprised me enough that I wanted to put it somewhere people would argue with it.

Out of 17,255 live roles today: Engineering 43.5%, Sales & GTM 17.8%, Operations 11.6%, Research 5.6%, Marketing 4.7%, Product 3.6%, Policy & Safety 1.0%, Other 12.3%.

The honest caveat first, because it changes the headline: "Other" is titles my classifier could not confidently label, and spot-checking it, it contains real engineering roles. So "56% is not engineering" would be overclaiming. What I can defend is 44.3% sitting in a named non-engineering function. Sales & GTM alone is bigger than Research, Marketing and Product combined.

Two other numbers that stuck with me. Only 19.3% of these roles publish a salary. And 23% are remote, which is lower than the discourse suggests.

The reason I think it matters: nearly every AI job board is engineering-only. If you are a marketer, a recruiter, an ops lead or a policy person trying to move into AI, you are searching on boards that structurally do not contain your job, and concluding the roles do not exist. They do, they are just invisible to you.

Genuinely curious what people here think. If you have hired at an AI company recently, does that split match what you saw internally, or is my sourcing skewed toward companies that use modern ATS software? And if you are hunting right now, does a missing salary stop you applying or do you apply and ask later?

Happy to run any cut of the data people want. I have the raw numbers on every role, company, stage and location.