Introduction

In today’s gig economy, AI wage‑setting is no longer science fiction—it’s influencing how much workers get paid. On platforms like Fiverr, Upwork, and Freelancer, large language models are now recommending hourly rates that could reshape labor markets. Do these systems offer fair pricing—or do they perpetuate old inequalities anew?

The Experiment in a Nutshell

We ran a large‑scale test using 60,000 freelancer profiles across six fields: accounting, full‑stack development, virtual assistance, data analytics, graphic design, and social media marketing. Each profile had fixed skills, experience, and location. We then fed them into eight popular LLMs—such as GPT‑4.o, Gemini 2.5 Flash, Claude 3.7 Sonnet, and Llama 3.1 405B—and asked for an hourly rate. By tweaking just one non‑skill factor (gender, location, or age), we probed whether AI would treat similar freelancers differently.

Key Findings

AI Prices Tend to Soar

The average human‑set wage was about $23.60/hour, but AI recommendations ranged from $30 to nearly $46/hour depending on the model—revealing a built‑in price inflation. While this might sound like a win for freelancers, it risks pricing them out of opportunities if clients balk.

Gender Bias Surprisingly Light

Across hundreds of thousands of cases, we saw no consistent gender wage gap. Even when we urged models to factor in gender, they didn’t systematically penalize women. Only under explicit, biased instructions did they occasionally overcompensate them.

Geography Drives the Deepest Gaps

Here’s where AI stumbles: location plays a huge role. Two identical profiles—one based in the U.S., one in the Philippines—got vastly different rates: $71 vs. $33/hour, a more than 50% gap. Changing the location field alone often more than doubled the recommendation.

Age Bias Favors Experience

A 60‑year‑old freelancer was priced roughly 46% higher than a 22‑year‑old peer, even with identical credentials. Prompting the models to ignore age made little difference—age bias seems more deeply baked into training data.

Prompting Isn’t a Perfect Fix

We experimented with prompts: asking the model to “ignore location” dramatically reduced the U.S.–Philippines gap. But telling it to “ignore age” barely budged the difference. This reveals that some biases are easier to mitigate than others—and prompt engineering is no silver bullet.

Beyond Hourly Wages

Repeating a similar experiment for full‑time salaries showed consistent patterns: gender neutrality, significant geographical bias, and age favoring mid‑career workers. It suggests that AI treats freelance and full‑time compensation differently—and often in subtle, contextual ways.

Implications for Stakeholders

• Freelancers: Use AI‑suggested rates as guidance, not gospel. Revealing your location can significantly affect outcomes.

• Employers: Treat AI wage advice as advisory, not binding. Oversight is essential.

• Platforms: Embed governance—transparency audits, appeal mechanisms, prompt reviews—into wage pipelines.

• Policymakers: Recognize that AI pricing is not neutral. Disclosure, worker rights, and oversight are overdue.