In short
Of 1,737 live agentic AI listings we track, 225 are the same role advertised in another city by the same employer. That is 13.0% of the board, spread across 129 groups. The market holds 1,512 distinct roles, not 1,737. One Databricks backend role is posted in nine cities, and a Wells Fargo product role in seven.
We track every agentic AI listing we can find. As of today the corpus holds 1,737 live roles across 455 companies. We recently counted how many of those listings share both an employer and an exact job title with another listing.
The answer is 225, in 129 groups. That is 13.0% of the board.
What a duplicate actually looks like
These are not spam and they are not scraper errors. Every one has its own source URL on the employer's own careers system, and every one is a real application you could submit. They are one job requisition advertised in several places at once.
The largest groups in the corpus right now:
- Databricks, Staff Software Engineer, Backend: 9 listings, including Amsterdam, Seattle, Mountain View, Bellevue, Belgrade, Berlin and Aarhus
- Databricks, Sr. Solutions Architect: 8 listings, including Auckland, Queensland and Milan
- Wells Fargo, Senior Lead Product Manager, Home Lending Agentic AI Transformation: 7 listings, including Charlotte, Minneapolis and San Antonio
- AssemblyAI, Senior Design Engineer: 7 listings, including San Francisco, New York City and Seattle
- Databricks, Senior Software Engineer, Backend: 6 listings
Five groups account for 37 listings between them. Strip the repeats and the market holds 1,512 distinct roles rather than 1,737.
Why this matters if you are counting the market
Listing counts are the most quoted number in any "state of AI hiring" claim, including ours. A board that reports 1,737 openings and a board that reports 1,512 are describing the same market. The difference is entirely a question of whether you deduplicate, and almost nobody says which they did.
The distortion is not evenly spread, which is the part that matters. Duplication concentrates in large enterprises with many offices and a centralised requisition system. Databricks alone holds 228 listings in our corpus, the largest single employer on the board, and a meaningful share of those are one role in several cities. A startup with four openings has four requisitions. So any comparison between "how much is the enterprise hiring" and "how much are startups hiring" is inflated on the enterprise side before you start.
It also inflates growth. If a company opens one role and posts it in nine cities, a listing-count metric records nine new jobs.
Why this matters if you are looking for work
For a job seeker the effect is smaller but more annoying. Search a role and the same title fills the first screen, which reads as either a hiring spree or a broken search. It is usually neither.
Two practical notes. First, when a role appears in several cities, the employer has usually not decided where it sits, which is more negotiating room on location than a single-city posting implies. Second, apply through the listing for the location you actually want. The requisition may be shared, but the recruiter routing frequently is not.
What we changed
We now treat each group as one role for indexing purposes. A single listing carries the canonical version, declares every city it is open in, and links the others so you can still reach the posting for your own location. Every listing stays live, because the per-city application links are the useful part. What changes is that search engines see one role in nine cities instead of nine competing pages.
Our own headline count stays honest about which number it is quoting. When we say the board holds 1,737 live roles, that is listings. When we say 1,512, that is distinct roles. Both are true and they are not interchangeable.
The number to ask for
If you read a claim about the size or growth of the AI job market, the question worth asking is whether the count is listings or requisitions. Most published figures are listings, because listings are what a scraper returns. The gap between the two is about 13% on our board, and there is no reason to expect it to be smaller on a board that aggregates more enterprise sources than we do.
You can browse the deduplicated market on the full job index or by specialisation on the role hubs. The underlying counts, including the employer concentration figures, are published as an open dataset in the State of Agentic Hiring report. Our piece on the long tail of single-opening employers is here.
Frequently asked
How many agentic AI job listings are duplicates?
In our corpus of 1,737 live listings, 225 share both an employer and an exact job title with another listing, spread across 129 groups. That is 13.0% of the board, leaving 1,512 distinct roles.
Are duplicate job listings a sign of a fake or padded job board?
Not usually. Each duplicate in our corpus has its own URL on the employer's own careers system and is a real application. They are one requisition advertised in several cities, which is standard practice at large employers with many offices.
Which company posts the most duplicate listings?
Databricks. Its Staff Software Engineer, Backend role appears nine times across cities including Amsterdam, Seattle, Mountain View, Bellevue, Belgrade, Berlin and Aarhus, and its Sr. Solutions Architect role appears eight times.
Does duplication inflate reported AI job market growth?
Yes. A single requisition posted in nine cities registers as nine new jobs in any listing-count metric. Because duplication concentrates in large multi-office employers, it also inflates enterprise hiring relative to startup hiring.
Should I apply to every city version of the same job?
No. Apply through the listing for the location you want, since recruiter routing is often per-location even when the requisition is shared. A role open in several cities usually means the employer has not fixed the location, which gives you more room to negotiate it.