AI freelance pay in 2026: what Upwork and Fiverr actually report
Upwork and Fiverr both shrank in buyers and grew in spend per buyer in Q2 2026. Read the side-by-side numbers, and what neither company discloses.
The Editors · 9 min read · Updated
Both platforms are getting smaller in people and larger in money per person. That is the Q2 2026 story on Upwork and Fiverr, and it holds on both sides of the market.
Upwork reported second quarter 2026 results on August 10, 2026: gross services volume of $966.4 million, down 4% year over year, across 763,000 active clients, with GSV per active client at $5,230, up 5% and the eighth consecutive quarter of sequential growth. Take rate came in at 19.8%.
Fiverr reported second quarter 2026 results on July 29, 2026: revenue of $97.8 million, down 10.0%, with annual active buyers at 2.7 million against 3.4 million a year earlier, and annual spend per buyer at $368 against $318. Marketplace take rate for the twelve months to June 30, 2026 was 28.0%.
Read together, the shape is clear. Fiverr shed roughly 700,000 buyers in twelve months while raising what each remaining buyer spends by 15.6%. Upwork gave up volume while raising what each remaining client spends, eight quarters running. Both are consolidating around fewer, larger relationships, and Fiverr keeps 8.2 percentage points more of each transaction.
The Q2 2026 numbers, side by side
The two companies disclose different metrics, so this table reports each one as its own filing states it rather than forcing a common unit.
| Upwork | Fiverr | |
|---|---|---|
| Results reported | August 10, 2026 | July 29, 2026 |
| Revenue | $191.7M | $97.8M, down 10.0% |
| Marketplace volume | $966.4M GSV, down 4% | not disclosed as GSV |
| Buyer-side count | 763,000 active clients | 2.7M annual active buyers, from 3.4M |
| Spend per buyer-side account | $5,230 GSV per client, up 5% | $368 per buyer, from $318 |
| Take rate | 19.8% | 28.0%, twelve months to June 30 |
| Bottom line | $64.1M adjusted EBITDA | $4.5M GAAP net income |
Neither company reports how many freelancers it has
Both platforms count the buying side only. Upwork discloses active clients. Fiverr discloses annual active buyers. Neither figure counts sellers, and neither company puts a seller headcount in its quarterly results.
Figures that circulate for "freelancers on Upwork" trace back to marketing pages and old registration totals rather than to a filing, and registration totals count accounts that have never earned a dollar. If seller headcount matters to a decision you are making, treat any number you find as unsourced until a filing carries it.
What the filings do support is the direction of the seller market: buyer-side accounts are falling on both platforms while spend per account rises, so the money per remaining relationship is going up even as the number of relationships goes down.
What Upwork's two reports actually report
Upwork publishes two different things and they get treated as one. They are not the same.
The February 2026 In-Demand Skills report (February 4, 2026) measures skill-level demand growth on the marketplace. It is a year-end summary built from freelancer earnings across six work categories during calendar 2025, with demand originating in the United States. It said top AI-related skills more than doubled year over year.
The quarterly 8-K is a financial disclosure. The Q1 2026 release (May 7, 2026) reported gross services volume from AI-related work up more than 40% year over year. On the Q2 2026 call, management put growth in high-value AI strategy and consulting work at 51% year over year while describing low-complexity, highly automatable work as burning off.
The skills report describes a slice of demand in one country over one calendar year. The 8-K describes company revenue and volume. Quoting one as if it were the other is the most common error in coverage of this market.
What Fiverr's numbers actually say
Fiverr's Q2 2026 disclosure is a story about mix. Revenue fell 10.0% while spend per buyer rose 15.6% and take rate rose four tenths of a point. A marketplace losing a fifth of its buyers and still growing revenue per buyer is trading breadth for depth.
That is consistent with the pivot the company described earlier in 2026. In Q1 2026 (April 30, 2026), services revenue hit $38.4 million, up 30% year over year, and the company put AI consulting and AI development at +118% year over year on the earnings call. Projects over $1,000 grew at a strong double-digit rate, with 18% year-over-year growth in clients completing $1,000+ projects.
For a seller, the practical read is unchanged and now better evidenced: an offering priced like the legacy Fiverr stereotype is swimming against the platform's own economics.
The durable lane vs the viral lane
The viral percentages, AI video at +329% and AI data annotation at +154% in the February 2026 skills report, come from small absolute bases. A category goes from 100 jobs to 430 jobs and posts a +329%. That can still be a real opportunity, and it does not mean the base is large.
The durable percentage is Upwork's AI Integration & Automation: largest AI sub-category by GSV, up 50% in Q1 2026, up 178% in skills demand for the full prior year. It is the work clients pay for when they want a working AI pipeline, with data flowing in, a model called, and output landing where it needs to land. This work scopes to project pricing. Fiverr is moving the same direction with its $1,000+ tier.
Absolute size sets the floor on how much work is reachable. A +329% category holding 0.2% of platform volume gives you fewer realistic opportunities than a +50% category at 5%. The Upwork and Fiverr disclosures agree on which bucket is larger.
The platform economics are tilting upmarket
Both platforms are deliberately moving toward bigger contracts, and Q2 2026 is the clearest evidence so far.
Upwork's GSV per active client has now risen for eight consecutive quarters while total volume fell, which is what a deliberate mix shift looks like in a filing. Its Business Plus offering for SMB clients grew GSV 34% quarter over quarter in Q1 2026, with Business Plus active clients up 35% quarter over quarter (Q1 2026 8-K).
Fiverr has stopped describing itself as a transactional marketplace, and its buyer count is falling faster than its revenue.
Two consequences follow for anyone weighing this work. The economic incentive on both platforms is to deliver fewer, larger engagements. And rate compression at the bottom is real before oversupply even enters the picture, because AI tools do entry-level annotation and integration cheaply enough that humans cannot price below them.
We covered the cost side of running AI work in the agent-loop math piece: token prices fell while bills tripled because the loops got longer. That dynamic is one reason clients want senior operators on $1,000+ projects. Someone has to keep the agent stack from running up a bill.
Who actually gets paid
The releases do not disclose median freelancer earnings, and the structure of the data still points at a distribution.
- Operators who scope $1,000-10,000 AI integration projects. They map to Upwork's AI Integration & Automation sub-category and Fiverr's AI development and consulting categories. Both platforms route demand here. The work is concrete: connect a CRM to a model, run a structured-output pipeline that lands rows in a sheet, build an agent loop that runs on a schedule and writes its own logs.
- Specialists in narrow viral sub-categories such as AI video editing, RAG evaluation tooling, or niche annotation. Real demand growth on a small base, with more competition per dollar and category-fashion risk.
- Generalists running cheap small gigs. Both platforms squeeze this lane by design. Fiverr's buyer count fell 700,000 in a year while spend per buyer rose, which is this group leaving.
Group 1 is where the data points, on either platform. The discipline that gets you there is the one that compounds elsewhere: pick one durable lane, learn the platform's pricing tier, deliver one project well, then raise the rate. The boring money moves that compound apply here too.
What the platform numbers don't show
A teardown of platform reports says nothing about off-platform freelancing. Direct client relationships, retainer work and repeat consulting from a personal network never appear in Upwork's 8-K. That market is larger than the platform market and structurally different: higher rates, longer payment cycles, no platform take rate and no top-of-funnel either.
Upwork's skills report measures demand originating in the United States; Fiverr's reporting is closer to global. Conflating them hides the geography.
The supply side is the largest blind spot. Every figure above measures demand or buyer behaviour. If the number of sellers offering AI services doubled alongside demand, rate compression is the result, and no disclosure in either filing would show it.
What to watch next
Three things will tell you whether the trend holds into 2027.
Whether Upwork's GSV per active client extends past eight quarters. If the streak breaks while total GSV keeps falling, the upmarket mix shift has run out of room rather than settled in.
Whether Fiverr's buyer decline slows. Losing a fifth of buyers in a year is a pivot; losing a fifth again is a different story, and the twelve months to June 2027 will separate the two.
Whether either company starts disclosing seller-side counts. Both have reasons to stay quiet while the seller market compresses, so a voluntary disclosure would be a signal in itself.
What changed
- August 19, 2026. Refreshed on Q2 2026 results from both companies (Fiverr July 29, Upwork August 10). Replaced the Q1-only framing, added the side-by-side table, added a section on the seller headcount neither company reports, and answered the three questions the previous revision left open.
- June 13, 2026. First published, built on Q1 2026 results and Upwork's February 2026 In-Demand Skills report.
Sources
- Upwork, Q2 2026 financial results (August 10, 2026)
- Fiverr, Q2 2026 results (July 29, 2026)
- Upwork, Q1 2026 8-K press release (SEC) (May 7, 2026)
- Upwork, In-Demand Skills 2026 (February 4, 2026)
- Fiverr, Q1 2026 results (April 30, 2026)
- Fiverr, Q1 2026 earnings call transcript (Motley Fool) (April 30, 2026)
This is not financial advice.
