Prompt engineer jobs: why the title is rare and where the skill is hired

Prompt engineer jobs are searched for far more often than they are posted. On September 17, 2026, AI Jobs Map listed 139 roles with prompt engineer or prompt engineering in the title, out of 28,283 AI roles posted in the last 30 days. That is about one AI posting in 203. This guide is honest about that number and then shows where the same skill is actually hired, because it is hired, just under other names.

Prompt engineer roles: prompt design, evaluation and LLM behaviour tuning. The count covers titles containing "prompt engineer", "prompt engineering". Postings that ask for prompt engineering as one requirement among many, under a different title, are not in it, and those are the majority.

How many prompt engineer jobs there are

The title is small, but it has two features worth knowing. Its entry level share is unusually high: 34 of the 139 postings on September 17, 2026 were classified as entry level, a 24% share against 8% across all AI roles. And its remote share was 14%, close to the site average of 12%.

The high junior share reflects what the title has come to mean. When a posting is titled prompt engineer today it is often an operations, content or support role: writing and maintaining the prompts behind a product, testing outputs, building examples. It is real work and a reasonable way in. The title had a moment in 2023; the postings on the page today describe a narrower job, and their requirements say so.

Where prompt engineering is actually hired

The skill has been absorbed into larger titles. The table compares the prompt engineer count with the titles that most often list prompt design, evaluation and LLM behaviour tuning among their requirements.

LLM engineer jobs are the closest match. LLM engineer roles: fine-tuning, inference, evaluation, prompt systems and large language model applications. Prompt systems are named explicitly, and the role adds the parts a prompt engineer title leaves out: fine-tuning, inference and evaluation infrastructure.

Generative AI jobs are the broadest match. Generative AI roles across engineering, research and product. Many of these postings are the same job as a prompt engineer posting with a wider remit and a higher grade.

AI engineer jobs are the largest pool by far, with 10,474 roles on September 17, 2026. AI engineer roles building LLM-powered products, agents, retrieval systems and AI platform infrastructure. Prompt design appears in these postings as one line among ten; the other nine are software engineering.

The tool pages give a second angle. Postings that mention Claude or Anthropic numbered 43,447 on September 17, 2026 and postings that mention OpenAI numbered 17,015. Those counts cover every posting in the index that names the vendor, in any role, so they are much larger than any AI title; what they show is how many employers are building on a model API, which is exactly the population that needs someone who can write and evaluate prompts.

What an AI prompt engineer job looks like in 2026

Read the postings on the prompt engineer jobs page next to those on the LLM engineer page and the difference is not the prompts. It is everything around them:

• Evaluation. Every serious posting wants someone who can say whether a change made the output better, with a test set and a number, not a feeling.
• Structured outputs and tool calls. Prompts that return JSON for a program to consume, and fail cleanly when the model does not comply.
• Retrieval. Deciding what context goes into the prompt is most of the work in a RAG system, and postings treat it as part of the prompt job.
• Cost and latency. Shorter prompts, cached prefixes, smaller models where they suffice.
• Code. Python at minimum. The prompt lives in a repository, is versioned and is tested like anything else.

If you can do the first and the last of those, you are competitive for the larger titles, and you should apply to them. If you can only write prompts, the small title is the honest fit, and it is a way to learn the rest.

How to search for it

Use the title page for the title, and the text for the skill. Free-text search accepts phrases, so “prompt engineering” on its own returns every posting that asks for it under any title, which is a far larger list than the title page. Combine it with a role page when the list is too long. Check the last three days for the newest of them; small titles close fast.

Every figure here is measured over roles posted in the last 30 days and still open on September 17, 2026. The pages linked are rebuilt daily and will differ from these figures.