Is freelancing dying? What the platforms' own numbers show in 2026

Is freelancing dying? What the platforms' own numbers show in 2026

On 16 September 2026 TrustList decided to build a freelancer marketplace: public profiles for individual freelancers, pages by country and skill, and later paid engagements with the payment held until the work is accepted. Today that part of the site does not exist yet. We checked on 25 September 2026 and the planned /freelancers address still returns "not found". We have no freelancer profiles, no freelancer reviews and no count of freelancers to report, and we will not invent one.

What we do have is the business side of the same market. When we built our offshore rankings in September we found 770 firms whose own websites show evidence of staff augmentation: 545 in India, 165 in Ukraine and 23 in Pakistan. Many of those firms fill their benches with contractors, and several describe working with freelancers directly. So before we build a product for freelancers, we had to answer a plain question our owner put to us: are freelance services dying, failing to keep up as demand falls away? This article tests that claim against the platforms' own filings, official labour statistics and two peer-reviewed studies. The short answer is that one kind of freelancing is shrinking fast, another is growing, and the total is being squeezed rather than wiped out.

What we are building, and why the question matters to us

Our plan has three phases. First, guides for freelancers on getting paid, contracts and intellectual property, and how to build a track record someone can check. Second, profiles built on our existing public person pages, with country, city, skills, day-rate range and availability, ranked only on verified facts and client reviews. Third, engagements where a buyer pays, TrustList holds the money, and the freelancer is paid when the buyer accepts the work.

If freelancing really were collapsing, phase three would be a poor use of our time. If it is shifting towards larger, more specialised work, then a marketplace that checks people and holds payments might be more useful than another place to post small fixed-price jobs. The evidence below is what we used to decide which of those is true.

The two big platforms: fewer buyers, larger jobs

Upwork and Fiverr are the two largest listed freelance marketplaces. Both publish their numbers every quarter to American regulators, so their figures are the closest thing to hard data on platform freelancing. We read the releases themselves, not summaries of them.

Measure	Upwork	Fiverr
Buyers a year earlier	796,000 active clients (Q2 2025)	3.4 million annual active buyers (30 June 2025)
Buyers now	763,000 active clients (Q2 2026), down 4%	2.7 million annual active buyers (30 June 2026), down 21.9%
Money per buyer	$5,230 GSV per active client, up 5%	$368 annual spend per buyer, up 15.6% from $318
Volume or revenue	Q2 2026 gross services volume $966.4 million, down 4%	Q2 2026 marketplace revenue $63.1 million, down 15.5% from $74.7 million

Upwork. Upwork counts a client as active if it has spent money on any Upwork platform in the previous 12 months. By that measure it had 832,000 active clients at the end of 2024, 785,000 at the end of 2025 and 763,000 at the end of June 2026. In its annual report for 2025 the company said the 6% fall over 2025 was "driven by slower growth in acquisition of new clients as well as lower retention of existing clients". Yet gross services volume, the total that clients spent, rose 1% over 2025 to $4.03 billion, and revenue rose 2% to $787.8 million. Fewer clients spent more each.

By the second quarter of 2026 the total had started to fall too: gross services volume was $966.4 million, down 4% on a year earlier, and revenue was $191.7 million, down 2%. Upwork's full-year revenue guidance for 2026 is $730 million to $750 million, below the $787.8 million it reported for 2025.

Inside those totals, one category is growing quickly. Upwork reported that spending on AI-related work passed $300 million a year on an annualised basis in the fourth quarter of 2025, up more than 50% from a year earlier, and that it grew by more than 22% year on year in the second quarter of 2026. Spending on AI strategy and consulting work grew by more than 50% in the same quarter.

Upwork's annual report also names the competition directly: "AI tools are emerging as an additional area of competition as technology becomes capable of generating work outputs that could otherwise be produced by human talent."

Fiverr. Fiverr's fall is sharper. Its annual active buyers went from 3.6 million at the end of 2024 to 3.1 million at the end of 2025, then to 2.7 million at the end of June 2026. That is 21.9% fewer buyers in a year. Each remaining buyer spent more, $368 a year against $318, and Fiverr says the number of clients completing projects worth $1,000 or more grew 13% over the trailing twelve months.

The overall picture is still one of decline. Fiverr's second-quarter revenue was $97.8 million, down 10.0%, and its marketplace revenue, the fees from buyers and sellers on the main site, fell 15.5%. Its separate services revenue grew 2% to $34.6 million. In its revised outlook, published in July, Fiverr expects full-year 2026 revenue of $356 million to $372 million, a fall of 14% to 17%, and third-quarter revenue 18% to 26% below a year earlier. Its chief executive, Micha Kaufman, put the cause plainly in the results release: AI "absorbs high-volume, low-value, transactional tasks".

Fiverr's model has always leaned towards small, fixed-price jobs, the kind of "logo for a start-up" or "500-word blog post" order that people once bought in minutes. Upwork's leans towards longer contracts. The gap between a 4% fall in clients and a 22% fall in buyers is, in our reading, the clearest single sign of where the damage is.

What the research found about which work went

Company numbers tell you what happened but not why. Two academic studies, both built on data from large freelance platforms, looked at the months after ChatGPT and image generators were released at the end of 2022.

Demirci, Hannane and Zhu, in a paper first issued as a CESifo working paper in 2024 and since published in the journal Management Science, found a 21% fall in job posts for automation-prone work related to writing and coding, compared with jobs that need manual-intensive skills, within eight months of ChatGPT's release. After image generators arrived, job posts for image creation fell by 17%. Two further findings matter for anyone working in those categories. Fewer posts meant more competition among freelancers for each one. And the jobs that remained were more complex and paid more.

Hui, Reshef and Zhou, in a paper published in Organization Science in November 2024, followed individual freelancers rather than job posts. They found that freelancers in the occupations most exposed to generative AI lost both work and earnings. They found similar effects after image models were released. Most uncomfortable of all, they found no evidence that a strong track record protected people; if anything, their evidence suggested top freelancers were hit harder.

Put those two studies next to the company filings and the pattern is consistent. The first work to go was short, well-defined writing, simple coding and image jobs, where a buyer could get a passable result from a tool in minutes. The work that grew was the work of deciding what AI should do inside a business and making it work, which is exactly the category Upwork reports as growing fastest.

One caution. These studies measured the first year or so after the new tools arrived. We have not found a peer-reviewed study of the same platforms covering 2025 and 2026, so we cannot say from research whether the fall in writing and image work has levelled off or deepened. The Fiverr buyer numbers suggest it has not stopped.

Outside the platforms: the official statistics

Platform marketplaces are only part of freelancing. Most self-employed people find work through their own networks, agencies or repeat clients, and never touch Upwork or Fiverr. Official statistics give a wider view, though a slower and blunter one.

United Kingdom. The Office for National Statistics estimates self-employment from its Labour Force Survey. In the release of 15 September 2026, it put the number of self-employed people at 4.53 million for May to July 2026. A year earlier, for May to July 2025, the estimate was 4.43 million, so the count rose by roughly 100,000 over the year. It is still well below the 5.02 million recorded for December 2019 to February 2020, just before the pandemic. The ONS itself warns that "some volatility remains in the LFS estimates", particularly for mid-2023 and 2024, so treat small changes with care.

Self-employment is a much broader group than freelancers in digital services. It includes builders, drivers, hairdressers and many others. The ONS figure shows no general collapse in people working for themselves in the UK, but it cannot tell you how UK copywriters or designers are doing.

United States. The Bureau of Labor Statistics measured independent contractors in a special survey in July 2023 and published the results in November 2024. It counted 11.9 million independent contractors, 7.4% of all employment, up from 6.9% in May 2017. Contingent workers, those who do not expect their job to last, were 4.3% of workers against 3.8% in 2017. That survey came eight months after ChatGPT's release, so it is too early to show much of the AI effect, and the BLS has not yet published a newer one.

Outside Europe and North America. Some countries report freelance earnings through their central banks. Profit, a Pakistani business publication, reported in July 2026 that State Bank of Pakistan figures showed freelancers earned $1.76 billion in foreign exchange in the year to June 2026, up 78% from $984 million a year earlier, with $1.16 billion of it from IT work. A World Bank report in September 2023 estimated that online gig work accounts for up to 12% of the global labour market. Neither figure is about the same buyers as Upwork's or Fiverr's, but both argue against a story of worldwide collapse.

Dying, shrinking or shifting: our reading

The owner's framing was that freelance services are dying and failing to keep up with falling demand. Here is what we think the evidence supports, and what it does not.

Supported: demand for some freelance work has fallen hard. Short, repeatable writing, simple design and small coding jobs have lost buyers. (Translation is often named too; the studies we read did not measure it separately, so we cannot confirm it.) Two peer-reviewed studies found job posts and earnings falling in exactly these categories, and Fiverr, the platform built on such jobs, has lost more than a fifth of its buyers in a year and expects revenue to fall by up to 17% in 2026. For people whose whole business was that kind of work, "dying" is not an exaggeration.

Not supported: freelancing as a whole is dying. Upwork's client count is down only 4% in a year and its spend per client is at a record. Spending on AI-related work on Upwork is growing by more than a fifth a year. UK self-employment rose over the past year. Pakistan's central bank reports record freelance earnings. None of that looks like a sector in terminal decline.

Partly supported: the platforms are not keeping up. Both companies are smaller businesses than they were, and both now depend on fewer, larger buyers. That is a risk for the platforms and for freelancers who rely on them for all their work. It is not the same as demand for skilled independent work disappearing.

So our answer is "shifting and shrinking at the low end", not "dying". The work that a tool can do in minutes is going. The work of deciding, integrating, checking and taking responsibility is not, and some of it is growing. That is why we are still building the marketplace, and why it will rank people on checked facts rather than on how cheap they are.

What a freelancer should do now

If you work for yourself in a digital service, these are the decisions we would make with the evidence above.

1. Move up the size of job. The work that remains on the platforms pays more and is more complex. Stop competing on small fixed-price tasks that a buyer can now do with a tool. Offer the larger job around them: the content plan and not just the post, the design system and not just the logo, the working integration and not just the script.
2. Sell judgement and accountability. Buyers still pay for someone who is responsible for the result. Say in your proposals what you check, what you guarantee and what you will fix at no cost.
3. Use the tools openly. Say which AI tools you use and where a person checks the output. Hiding it helps no one when a buyer can tell anyway.
4. Do not rely on one platform. Fiverr's buyer count fell by more than a fifth in a year. Keep your own list of past clients, your own website and at least one channel where buyers can find you without a platform in the middle.
5. Build a record someone can verify. Named clients who agree to be named, links to live work, dates, and a short description of what you did and what changed. The Hui, Reshef and Zhou study found past ratings did not protect people; evidence of results you can prove is harder to replace.
6. Learn the growing categories. Upwork's fastest-growing work is AI integration, automation and strategy. You do not need to become a machine-learning engineer, but knowing how to put these tools into a client's real processes is now a sellable skill in most fields.
7. Watch your own numbers monthly. Count enquiries, win rate, average job value and repeat clients. If enquiries fall but job value rises, you are following the market. If both fall, change what you sell.

What a business hiring a freelancer should do

1. Decide what you are buying: output or judgement. If a tool can produce the output and someone on your staff can check it, you may not need a freelancer. If you need someone to decide what good looks like and take responsibility for it, you do, and you should expect to pay for experience.
2. Ask how they use AI. Ask which tools, for which steps, and who checks the result. A good answer is specific. Put the answer in the contract if it matters to you, especially for confidential material.
3. Check the evidence, not just the rating. Ask for two past clients you can speak to and links to live work with dates. Platform ratings were not a reliable guide to who kept their work in the research above.
4. Write the contract properly. Scope, deliverables, dates, who owns the intellectual property, confidentiality, how and when payment is released, and what happens if the work is late or wrong. Pay against accepted milestones rather than all in advance.
5. Consider a firm for anything long or large. For an ongoing team rather than a single job, a staff augmentation firm takes on employment, replacement and continuity. Our offshore staff augmentation rankings and buyer's guide explain the checks. For how this fits your wider technology plans, read our guide on how CEOs and CTOs should structure technology teams, and for the wider shift in delivery, offshore development and changing IT delivery.
6. Keep the relationship if it works. The platforms are losing buyers partly because repeat work moves off them. A freelancer you trust, on a clear contract, is worth more to you than a new search every month.

Sources

• Upwork Reports Second Quarter 2026 Financial Results: Q2 2026 revenue, gross services volume, active clients, GSV per active client, AI-related work growth and 2026 guidance. Upwork, 10 August 2026.
• Upwork fourth quarter and full year 2025 results: 2025 revenue, GSV, year-end active clients and AI-related work. Upwork, 9 February 2026.
• Upwork Form 10-K for 2025: definition of active client, reasons for the fall in active clients, and the statement on AI tools as competition. Upwork, 12 February 2026.
• Fiverr Announces Second Quarter 2026 Results: Q2 2026 revenue, marketplace and services revenue, active buyers, spend per buyer, $1,000+ clients, chief executive's statement and revised 2026 outlook. Fiverr, 29 July 2026.
• Fiverr fourth quarter and full year 2025 results: active buyers and spend per buyer at 31 December 2025 and 2024. Fiverr, 18 February 2026.
• Who Is AI Replacing? The Impact of Generative AI on Online Freelancing Platforms: 21% fall in writing and coding job posts, 17% fall in image creation posts, remaining jobs more complex and better paid. Demirci, Hannane and Zhu, CESifo Working Paper 11276, 2024; later published in Management Science.
• The Short-Term Effects of Generative Artificial Intelligence on Employment: Evidence from an Online Labor Market: falls in employment and earnings in exposed occupations, and no protection from a strong track record. Hui, Reshef and Zhou, CESifo Working Paper 10601, 2023.
• Publication record for Hui, Reshef and Zhou: published in Organization Science, volume 35, issue 6. Washington University in St. Louis, November 2024.
• LFS: Self-employed, UK, aged 16 and over (MGRQ): self-employment for May to July 2026, May to July 2025 and December 2019 to February 2020. Office for National Statistics, 15 September 2026.
• Employment in the UK: September 2026: the caveat on Labour Force Survey volatility. Office for National Statistics, 15 September 2026.
• Contingent and Alternative Employment Arrangements, July 2023: independent contractors and contingent workers, compared with May 2017. US Bureau of Labor Statistics, 8 November 2024.
• Pakistan's freelancer earnings hit record $1.76 billion in FY26: State Bank of Pakistan figures for freelance export earnings. Profit, 21 July 2026.
• Demand for online gig work rapidly rising in developing countries: online gig work as up to 12% of the global labour market. World Bank, 7 September 2023.
• TrustList offshore research: 770 firms with staff augmentation evidence on their own websites, counted 16 September 2026; the /freelancers address checked 25 September 2026.