Skip to main content
What the OpenAI-Upwork Freelance Report Actually Shows

What the OpenAI-Upwork Freelance Report Actually Shows

Josef Holm8 min read

Key Takeaways

  • The OpenAI and Upwork freelance report shows a shifting mix of work inside AI-related postings and bigger projects, but the headline numbers are easy to misread if you don't check what each one measures.
  • Within AI-related posts, video production, social campaign management, marketing automation, and AI API integration gained share between 2023 and 2025, while AI-generated written content, training programs, and some cloud and data integration work lost share.
  • The widely cited 45% figure describes how spread out task groups have become across AI-related postings, not a 45% increase in work per brief. At the post level, AI-related jobs averaged 3.35 distinct task groups versus 3.19 for other posts.
  • After adjusting for work category and posting month, AI-related fixed-price budgets were about 52% higher than non-AI posts and first-90-day hours about 41% higher, while adjusted hourly rates were only about 1.5% higher.
  • These are observed differences on one marketplace with a broad definition of AI-related, not proof that AI caused larger projects or better returns; the value for freelancers looks more like carrying longer, broader engagements than charging more per hour.

The November 2026 OpenAI and Upwork report, The Rebundling of Freelance Work in the AI Age, looked at what companies are actually posting when they hire freelancers for work that involves AI. The headline findings are interesting, but they get misread quickly if you don't pay attention to what each number measures. So let me walk through what the report shows, where the numbers come from, and what I'd take from it if I were hiring freelancers or trying to earn a living as one.

What counts as an "AI-related" post

The report defines AI-related posts as those that request AI skills, require AI tools, or involve improving AI-generated output. [3] That's a broad definition. It covers someone hiring an engineer to integrate an API, and it covers someone hiring a video editor to polish a clip generated by a model. Keep that in mind, because the differences between AI and non-AI posts are averages across that whole mix.

The mix of work is shifting

Inside AI-related posts, the kinds of tasks companies ask for have moved. Video production, social media campaign management, marketing automation, and AI API integration gained share between 2023 and 2025. AI-generated written content, training programs, and some cloud and data integration tasks lost share.

Top gaining and losing task clusters within AI-related Upwork posts, 2023 to 2025

Within AI-related Upwork postings, video production, social-media campaign management, marketing automation, and AI API integration gained share, while AI-generated written content, training programs, and cloud/data tasks lost share.

Scroll the chart sideways to see every point.

Create and manage social media campaigns+1.68 ppProduce & edit motion graphics for videos+1.50 ppCreate and edit AI-generated video+1.46 ppImplement and manage marketing automation+1.40 ppCreate and produce videos+1.22 ppIntegrate AI APIs into applications+0.42 ppCreate and edit AI-generated written content-1.07 ppDesign and deliver training programs-1.54 ppEngineer and optimize cloud infrastructure-0.77 ppDesign and implement data and system integrations-0.71 ppTutor and coach writing and language skills-0.66 ppBuild chatbots and machine-learning models-4.0 pp (21.8% → 17.8%)
Top gaining and losing task clusters within AI-related Upwork posts, 2023 to 2025, exact data
LabelValueUnitPeriod
Create and manage social media campaigns+1.68 pppercentage points change2023 to 2025
Produce & edit motion graphics for videos+1.50 pppercentage points change2023 to 2025
Create and edit AI-generated video+1.46 pppercentage points change2023 to 2025
Implement and manage marketing automation+1.40 pppercentage points change2023 to 2025
Create and produce videos+1.22 pppercentage points change2023 to 2025
Integrate AI APIs into applications+0.42 pppercentage points change2023 to 2025
Create and edit AI-generated written content-1.07 pppercentage points change2023 to 2025
Design and deliver training programs-1.54 pppercentage points change2023 to 2025
Engineer and optimize cloud infrastructure-0.77 pppercentage points change2023 to 2025
Design and implement data and system integrations-0.71 pppercentage points change2023 to 2025
Tutor and coach writing and language skills-0.66 pppercentage points change2023 to 2025
Build chatbots and machine-learning models-4.0 pp (21.8% → 17.8%)percentage points change2023 to 2025
OpenAI analysis of Upwork Marketplace data. Shares are percentage of all task-cluster assignments in AI-related job posts; task clusters are not mutually exclusive. Calculation: Values copied directly from Exhibit 2.3, which shows 15 largest share gains and 15 largest share declines within AI-related work between 2023 and 2025; chatbots/ML models cluster share reported separately in note. Sources: cdn.openai.com.

Two things are worth noting. First, these are shares within AI-related posts, not absolute volume. A task can lose share while the total number of posts grows. Second, the report is describing posted demand on one marketplace. It isn't a map of the whole freelance economy, and it doesn't settle whether AI caused the shift or whether other things (ad budgets, platform trends, who happens to use Upwork) played a role.

If you're a freelancer whose work sits in a shrinking share, it doesn't mean the work is gone. It does mean the buyer's expectations are probably changing. Writing a blog post from scratch is a different brief than taking a model's draft and making it good.

The "45%" number is about variety, not brief size

This is the one I've seen misread most. The report's 45% figure refers to how spread out the mix of task groups has become across AI-related postings, not a 45% increase in the amount of work per brief. [3] It's a diversity measure across the market, not a per-post measure.

Distinct task groups per Upwork posting: AI-related vs non-AI

AI-related postings bundle about 5% more distinct task groups per brief than non-AI postings.

Scroll the chart sideways to see every point.

3.35AI-related posts3.19Non-AI posts
Distinct task groups per Upwork posting: AI-related vs non-AI, exact data
LabelValueUnitPeriod
AI-related posts3.35distinct task groups per posting2023–2025
Non-AI posts3.19distinct task groups per posting2023–2025
OpenAI/Upwork, 'The Rebundling of Freelance Work in the AI Age' (Nov 10, 2026), full report p. 11. Task-group counts from Upwork Marketplace postings classified via GPT-powered classifier; novelty regressions use a one-in-20 sample of 104,030 posts. Calculation: Direct values from report. Sources: AY Automate.

Effective number of tasks in AI-related Upwork postings, 2023–2025

The mix of task groups in AI-related postings broadened by about 45% from early 2023 to late 2025.

Scroll the chart sideways to see every point.

Jan–Jun 202314.6Jul–Dec 202521.2
Effective number of tasks in AI-related Upwork postings, 2023–2025, exact data
LabelValueUnitPeriod
Jan–Jun 202314.6effective number of tasksH1 2023
Jul–Dec 202521.2effective number of tasksH2 2025
OpenAI/Upwork, 'The Rebundling of Freelance Work in the AI Age' (Nov 10, 2026), full report section 3.1, p. 10. Effective number of tasks across AI-related Upwork postings. Calculation: Change from 14.6 to 21.2 corresponds to about +45%, as reported. Sources: AY Automate.

Effective number of task clusters in AI-related Upwork posts

The effective number of task clusters rose from 14.6 to 21.2 (about 45%) between 2023 and 2025, indicating a more spread-out task mix.

Scroll the chart sideways to see every point.

14.6H1 2023 (avg Q1–Q2)21.2H2 2025 (avg Q3–Q4)
Effective number of task clusters in AI-related Upwork posts, exact data
LabelValueUnitPeriod
H1 2023 (avg Q1–Q2)14.6effective clusters (1/HHI)Q1–Q2 2023
H2 2025 (avg Q3–Q4)21.2effective clusters (1/HHI)Q3–Q4 2025
Task-cluster assignments in AI-related Upwork posts, averaging Q1–Q2 2023 against Q3–Q4 2025. Calculation: Effective number of clusters = 1/HHI (concentration index); not the per-post task count. Sources: OpenTools.

At the individual-post level, the difference is smaller. AI-related posts averaged 3.35 distinct task groups versus 3.19 for other posts, and task groups can overlap within a single post. [3]

Distribution of distinct task clusters per Upwork post: AI vs non-AI

AI-related Upwork posts are more concentrated in three- and four-task bundles than non-AI posts, averaging 3.35 vs 3.19 distinct tasks per post.

Scroll the chart sideways to see every point.

1 task, AI jobs5.3%1 task, Non-AI jobs8.6%2 tasks, AI jobs14.3%2 tasks, Non-AI jobs20.4%3 tasks, AI jobs35.1%3 tasks, Non-AI jobs30.2%4 tasks, AI jobs33.9%4 tasks, Non-AI jobs27.8%5+ tasks, AI jobs11.7%5+ tasks, Non-AI jobs13.4%
Distribution of distinct task clusters per Upwork post: AI vs non-AI, exact data
LabelValueUnitPeriod
1 task, AI jobs5.3%percent of poststhrough 2025
1 task, Non-AI jobs8.6%percent of poststhrough 2025
2 tasks, AI jobs14.3%percent of poststhrough 2025
2 tasks, Non-AI jobs20.4%percent of poststhrough 2025
3 tasks, AI jobs35.1%percent of poststhrough 2025
3 tasks, Non-AI jobs30.2%percent of poststhrough 2025
4 tasks, AI jobs33.9%percent of poststhrough 2025
4 tasks, Non-AI jobs27.8%percent of poststhrough 2025
5+ tasks, AI jobs11.7%percent of poststhrough 2025
5+ tasks, Non-AI jobs13.4%percent of poststhrough 2025
Upwork Marketplace job postings through 2025. Means per post: 3.35 AI, 3.19 non-AI. Calculation: Values taken directly from Exhibit 3.1 Panel B showing shares of AI and non-AI posts by distinct task-cluster count. Sources: cdn.openai.com.

So the average AI-related brief isn't dramatically bigger in task count. What's changed more noticeably is the range of combinations that show up across the market. Someone is asking for video editing plus automation plus a bit of prompt work in the same post. That combination might have been rare two years ago.

For an operator, the practical read is this: if you're scoping a project that touches AI, assume you're hiring someone who will be stitching tools and skills together rather than doing one thing in isolation. The brief should say so.

Budgets, rates, and hours

Now to the money. After adjusting for work category and posting month, AI-related fixed-price budgets were about 52% higher than non-AI posts. [3] Adjusted hourly rates were only about 1.5% higher. [3] Among hired hourly contracts with positive recorded hours, first-90-day hours were about 41% higher for AI-related contracts. [3]

AI-related vs non-AI Upwork posts: budget, rate, and hours differences

AI-related Upwork posts carry much larger project budgets and generate substantially more hours per hired contract, with only a small posted hourly-rate premium.

Scroll the chart sideways to see every point.

70.3%Fixed-price budget (raw)52.4%Fixed-price budget (adjusted)9.2%Posted hourly rate (raw)1.5%Posted hourly rate (adjusted)20.1%90-day hours, hired (raw)40.6%90-day hours, hired (adjusted)
AI-related vs non-AI Upwork posts: budget, rate, and hours differences, exact data
LabelValueUnitPeriod
Fixed-price budget (raw)70.3%percent higher vs non-AIthrough 2025
Fixed-price budget (adjusted)52.4%percent higher vs non-AIthrough 2025
Posted hourly rate (raw)9.2%percent higher vs non-AIthrough 2025
Posted hourly rate (adjusted)1.5%percent higher vs non-AIthrough 2025
90-day hours, hired (raw)20.1%percent higher vs non-AIthrough 2025
90-day hours, hired (adjusted)40.6%percent higher vs non-AIthrough 2025
OpenAI/Upwork analysis of Upwork Marketplace postings through 2025. Adjusted estimates control for work subcategory and posting month; 90-day hours restricted to hourly contracts with positive recorded hours; top 1% of posted prices trimmed. Calculation: Values taken directly from Exhibit 2.4 raw and adjusted percentage differences between AI-related and non-AI job posts. Sources: cdn.openai.com.

Adjusted gaps between AI-related and non-AI Upwork posts

AI-related posts carry much larger fixed-price budgets and more first-90-day hours, but nearly the same posted hourly rate after adjustment.

Scroll the chart sideways to see every point.

+52%Fixed-price budget (adjusted)+1.5%Posted hourly rate (adjusted)+41%First-90-day hours (adjusted)
Adjusted gaps between AI-related and non-AI Upwork posts, exact data
LabelValueUnitPeriod
Fixed-price budget (adjusted)+52%% higher vs non-AI (adjusted)through April 2026
Posted hourly rate (adjusted)+1.5%% higher vs non-AI (adjusted)through April 2026
First-90-day hours (adjusted)+41%% higher vs non-AI (adjusted)through April 2026
Positive posted budgets (top 1% trimmed) and posted hourly-range midpoints through April 2026; hours cover hired hourly contracts with positive recorded hours in their first 90 days. Calculation: OLS regressions controlling for work subcategory and posting month; associations, not causal. Sources: OpenTools.

That pattern is worth sitting with. The hourly rate premium is basically a rounding error. The budget and hours gaps are large. Put plainly: companies aren't paying freelancers much more per hour for AI work, but the projects are bigger and run for longer.

A natural question is whether AI caused this. The report doesn't claim that, and I wouldn't either. [5] The adjustments control for category and month, not for buyer sophistication, project ambition, or which companies happen to post AI work on Upwork. It could be that firms willing to describe a job as "AI-related" are also firms with bigger budgets and longer runways. The data is consistent with that, and it's consistent with AI genuinely changing project scope. The report shows the association. It doesn't settle the mechanism.

What I'd take from this

A few practical observations, nothing more.

If you're hiring: write the brief to match the shape of the work. AI-related projects, in the data, tend to combine more things and run longer. If you expect a freelancer to use models, integrate APIs, and clean up AI output, say that in the post. Budget for the full span, not just the first deliverable.

If you're freelancing: the rate premium for calling your work "AI-related" is small. What's bigger is the project size and the hours. That suggests the value is in being the person who can carry a longer, broader engagement, not in charging more per hour for the same deliverable.

And for everyone: this is one marketplace, one platform, and a specific definition of "AI-related." Upwork postings don't speak for the whole labor market, and the report is careful not to claim they do. The shifts are real in the data. What they mean for any specific business depends on the business.

If you want more writing like this, you can subscribe to The Operator's AI Brief or find more posts on the blog.

Research notes

Research checked 12 November 2026.

Sources

  1. Work at the Frontier: How AI is changing the shape of freelance work, cdn.openai.com Retrieved 12 November 2026.
  2. OpenAI and Upwork study finds AI projects get bigger budgets and more hours on freelance markets | MLQ.ai, MLQ.ai Published 10 November 2026. Retrieved 12 November 2026.
  3. The Rebundling of Freelance Work in the AI Age, cdn.openai.com Retrieved 12 November 2026.
  4. Work at the Frontier: AI’s effect on freelance work | OpenAI, OpenAI Retrieved 12 November 2026.
  5. OpenAI–Upwork study: AI-related freelance projects bundle more work and carry larger fixed budgets, but the data don't show job replacement | OpenTools, OpenTools Retrieved 12 November 2026.
  6. OpenAI and Upwork: How AI Is Changing Freelance Work | CellCog, CellCog Retrieved 12 November 2026.
  7. OpenAI and Upwork Study: AI Projects Post 52% Bigger Budgets, AY Automate Published 11 November 2026. Retrieved 12 November 2026.

Infographic

Infographic summary of: What the OpenAI-Upwork Freelance Report Actually Shows

Frequently Asked Questions

What counts as an AI-related post in the OpenAI and Upwork report?
The report defines AI-related posts as those that request AI skills, require AI tools, or involve improving AI-generated output. That covers a wide range, from hiring an engineer to integrate an API to hiring a video editor to polish a clip generated by a model.
What does the 45% figure in the report actually measure?
It describes how spread out the mix of task groups has become across AI-related postings over time. It is a diversity measure across the market, not a 45% increase in the amount of work inside an individual brief.
How many distinct task groups does an average AI-related post have?
AI-related posts averaged 3.35 distinct task groups versus 3.19 for other posts, and task groups can overlap within a single post. So the average AI-related brief is not dramatically bigger in task count.
How do budgets, hourly rates, and hours compare between AI and non-AI posts?
After adjusting for work category and posting month, AI-related fixed-price budgets were about 52% higher than non-AI posts and first-90-day hours were about 41% higher for hired hourly contracts, while adjusted hourly rates were only about 1.5% higher.
Does the report show that AI caused these differences?
No. The adjustments control for category and month, not for buyer sophistication, project ambition, or which companies choose to post AI work on Upwork. The report shows an association between AI-related posts and larger budgets and hours, not proof of cause.
What should freelancers and people hiring take from the findings?
For freelancers, the rate premium for AI-related work is small, but projects tend to be bigger and run longer, so the value is in carrying a broader engagement. For buyers, if a project will combine tools and skills, the brief and budget should reflect the full span of work, not just the first deliverable.