What Percentage of Upwork Job Posts Are AI-Written?
We analyzed over 80,000 posts across 21 days. The number is higher than most freelancers expect.
The short answer
9 in 10 Upwork job posts are AI-written.
Across 21 days of job post data, 90.3% of posts were classified as AI-written. Of the remainder, 9.4% were human-written and 0.3% were inconclusive.
Which categories are most (and least) affected
AI adoption varies significantly by niche. Marketing, recruiting, and admin roles are almost entirely AI-written. Technical and creative work lags, but not by as much as you'd expect.
Most AI-written categories (categories with 100+ posts in the dataset):
| Category | AI-written |
|---|---|
| Lead Generation & Telemarketing | 95.6% |
| Sales & Marketing Copywriting | 95.0% |
| Recruiting & HR | 94.6% |
| Digital Marketing | 93.9% |
| Virtual Assistance | 92.5% |
| AI Apps & Integration | 92.1% |
Least AI-written categories (most human-written):
| Category | AI-written |
|---|---|
| 3D Modeling & CAD | 72.1% |
| Electrical & Electronic Engineering | 73.3% |
| Audio & Music Production | 73.5% |
| Energy & Mechanical Engineering | 74.9% |
| Photography | 77.8% |
| Art & Illustration | 80.1% |
Even in the "least AI-written" categories, roughly 3 in 4 posts are AI-generated. There is no Upwork niche where human-written posts are the norm anymore.
Does it actually matter?
Here's where the data gets interesting. Counterintuitive, too.
AI-written posts have higher budgets
For fixed-price jobs, the average budget breakdown:
| Post type | Average fixed-price budget | Posts analyzed |
|---|---|---|
| AI-written | $993 | 30,798 |
| Human-written | $397 | 4,074 |
AI-written posts carry a 2.5x higher average budget. The clients using AI to write job posts aren't cutting corners on pay. If anything, the opposite.
Client seriousness is nearly identical
We scored every post on client seriousness based on post quality signals, client history, and other factors:
| High seriousness | Medium | Low | |
|---|---|---|---|
| AI-written | 28.6% | 53.6% | 17.8% |
| Human-written | 33.7% | 50.0% | 16.3% |
A small gap, but not a meaningful one. An AI-written post doesn't predict a low-quality client.
AI posts are slightly more likely to be suspicious, but the effect is small
Post quality classification:
- AI-written: 96.0% genuine, 3.9% suspicious
- Human-written: 98.8% genuine, 1.1% suspicious
AI-written posts are about 3.5x more likely to be flagged as suspicious. That sounds alarming until you see the base rate: 3.9% vs 1.1%. The vast majority of AI-written posts are legitimate job posts from real clients.
This matches external research: a MIT Sloan study found AI-written job posts are 15% less likely to result in a hire. A real signal, but not a reason to skip 9 in 10 jobs.
What this means in practice
An AI-written post is not a red flag. It's a stylistic choice that's become the default. Filtering based on it would mean skipping 9 in 10 jobs, including the higher-budget ones.
The signals that actually separate good from bad posts are the same ones that always mattered:
- Payment verified: non-negotiable
- Client hire rate: has this client followed through before?
- Client total spent: do they have a track record of paying?
- Budget vs scope: does the ask match the money?
- Specificity within the post: does the post mention a real situation?