Fiverr Lost 26% of Buyers, Upwork AI Work Grew 22%: What the Filings Show

Fiverr ended September 2026 with 2.4 million annual active buyers, down 25.8% in a year. Over a similar window, Upwork reported that GSV from AI-related work grew more than 22%, while its total gross services volume fell 3.6%. The freelance market after AI is not one market shrinking. It is one kind of work collapsing while another, smaller kind grows, and neither company will tell you exactly where the line between them sits.

What the Fiverr and Upwork filings actually measure

Fiverr and Upwork are the two listed pure-play freelance marketplaces, so their filings are the closest thing to a public ledger of freelance work after AI. They do not measure the same thing, and comparing them without adjusting for that produces nonsense.

Both companies report a count of paying customers and a spend figure per customer. Both report a total volume. The words differ, the windows differ, and the comparisons in this post only work if you hold those differences in view.

Fiverr counts buyers on a trailing 12 months

Fiverr defines annual active buyers as buyers who have ordered a Gig on its marketplace "within the last 12-month period, irrespective of cancellations". Annual spend per buyer is gross merchandise value over the same 12 months divided by that buyer count.

Both are trailing figures. A buyer who placed one order 11 months ago still counts today. That makes Fiverr's buyer number slow to move, which matters for how you read the decline. A 25.8% drop in a trailing count means the damage started well before the quarter it shows up in.

Since Q4 2024, these metrics cover the marketplace only. Fiverr's services revenue, which includes seller tools and acquired businesses, sits outside them.

Upwork counts clients and gross services volume

Upwork reports gross services volume, or GSV, which its Q2 2026 10-Q describes as "the total dollar value transacted through all Upwork platforms and other workforce solutions". That includes enterprise programmes and value-added services, not only freelance contracts on the open marketplace.

Upwork's release points to an earnings presentation for its active client definition. That presentation blocked automated retrieval during research for this post, so the definition is not quoted here. Treat the Upwork client count as comparable in direction to Fiverr's, not in level.

The quarterly table: buyers, spend and GSV by company

Here are six quarters of Fiverr and five quarters of Upwork, taken from each company's own earnings releases. Upwork's Q3 2026 results had not been published when this post was written. Fiverr's Q3 2026 results came out on 29 October 2026.

Buyers, spend per buyer and marketplace revenue are Fiverr's reported figures. Implied trailing GMV is our calculation: buyers multiplied by spend per buyer, which reverses Fiverr's own definition. Spend per buyer is reported to the dollar, so the result carries a rounding error of a few million.

All figures from Upwork's quarterly earnings releases filed with the SEC. The Q2 2025 AI growth rate is stated in the Q3 2025 release as the prior-quarter comparison. Q2 2025 GSV per client was not in the releases opened for this post, so the cell is left empty rather than estimated.

Fiverr's AI-first cut: 762 staff to 528, and a smaller marketplace

In September 2025, Fiverr cut about 250 jobs, roughly 30% of its workforce. CEO Micha Kaufman told staff the goal was to "turn Fiverr into an AI-first company that's leaner, faster," according to Reuters' report on the letter. In an essay quoted by The Register, he called it "a painful reset."

Fiverr's annual report makes the result exact. The company had 528 employees at the end of 2025, against 762 a year earlier. The filing attributes the decrease to "restructuring initiatives undertaken in September 2025 as part of the Company's focus on AI initiatives".

What the marketplace did after the reset

At the time, Fiverr said it did not expect the cuts to materially affect marketplace activity in the near term. The marketplace numbers since then say the opposite story was already underway. Buyer decline accelerated in every quarter after the cut, from 11.7% in September 2025 to 25.8% in September 2026.

The tone of the releases moved with it. In July 2025, Fiverr said "AI-related services are booming". By July 2026, CEO Kaufman said "AI absorbs high-volume, low-value, transactional tasks," and the release cited "persistent weakness across categories most exposed to AI automation". Marketplace revenue fell 24.7% year on year in Q3 2026, to $55.4 million.

My reading is that the AI-first restructuring did not cause the marketplace decline and did not prevent it either. It was a cost decision taken ahead of a revenue problem that management could already see forming.

The upmarket bet, in Fiverr's own numbers

Fiverr's answer is to move toward larger projects. In Q4 2025 it reported that GMV from transactions over $1,000 grew 22.8%. In Q1 2026, clients completing $1,000-plus projects grew 18%. In Q2 2026, that growth slowed to 13%.

The Q3 2026 release changed the metric. It reported that projects at or above $1,000 made up 16% of completed projects' gross order amount on a trailing basis, with no growth rate attached. When a company swaps a growth metric for a share metric in the quarter the growth slows, read that as information.

Upwork's AI-related GSV: growing, slowing, and still small

Upwork took a different line in public. It has reported growth in AI-related work every quarter, and that growth is real. The direction of the rate is the part its releases do not put in a headline.

AI-related GSV grew 30% in Q2 2025, 53% in Q3 2025, more than 50% in Q4 2025, more than 40% in Q1 2026 and more than 22% in Q2 2026. Over the same stretch, total GSV went from growing 2.8% to shrinking 3.6%.

How big AI-related work is inside Upwork

Upwork gave one size disclosure. Its Q4 2025 release said GSV from AI-related work "surpassed $300 million on an annualized basis." Q4 2025 total GSV was $1,020.3 million, or about $4,081 million annualised. That makes AI-related work a little over 7% of Upwork's GSV on our calculation.

Coverage of the Q2 2026 earnings call, reported by Nasdaq, put the annualised AI-related figure near $330 million. That number comes from secondary call coverage, not the release, so the share it implies (about 8.5% of annualised Q2 GSV) is directional.

What the rest of Upwork's marketplace did

Here is the figure neither release states. If AI-related work ran at about $330 million annualised in Q2 2026, that is roughly $82.5 million for the quarter. Growth of 22% means a year-earlier base of no more than about $67.6 million.

Subtract those from total GSV. Non-AI GSV comes out near $884 million in Q2 2026 against about $935 million in Q2 2025, a fall of roughly 5.5%. Because the release says "more than 22%," the true decline is that size or larger. Upwork's total figure of -3.6% understates how fast the non-AI part of the marketplace is contracting.

Upwork's own 10-Q says GSV and active clients fell because of "the evolving impact of AI on certain categories of freelance work and on new client acquisition and retention, as well as macroeconomic uncertainty". It does not name the categories.

How each company defines AI work (neither one does)

The brief for this post was to quote each company's AI category definitions. There are none to quote. That absence is a finding in itself, and it matters more than any single growth rate in the table above.

Upwork's Q2 2026 release and 10-Q do not explain how a contract gets labelled "AI-related," whether by job category, required skill tag, or something else. Fiverr's releases speak of "AI-related categories" in 2025 and "categories most exposed to AI automation" in 2026, without a list or a share of GMV for either.

Quotations are from each company's quarterly earnings releases. Upwork's earnings presentation may carry fuller definitions. It could not be retrieved for this post.

Notice what Upwork does. It highlights a different AI sub-category almost every quarter, usually the one with the best growth rate that period. That is legal and common. It also means no outside reader can track one consistently defined AI category across five quarters, which is the one series that would settle the substitution question.

If you want to judge a company's AI claims, this is the test. A metric that changes name every quarter is a marketing line. This blog has argued the same point about which AI revenue disclosures are worth trusting, and freelance marketplaces fail it for the same reason software companies do.

Which freelance work is shrinking and which is growing

With the definitions missing, the filings still support a narrower conclusion. The shape is consistent across both companies and six quarters, even though the boundaries are blurred.

Shrinking: small, one-off, transactional orders. Fiverr's buyer count is falling much faster than its spend per buyer is rising. The buyers who leave are the ones who placed small orders, because the average spend of those who remain keeps climbing. Fiverr's own executives named this segment in July 2026: "low-value, transactional work."

Growing: larger projects and AI build work. Upwork's GSV per active client reached a record $5,230 in Q2 2026, its eighth straight quarter of sequential growth. Its fastest-growing named categories are integration, automation and consulting, which are projects where a client hires someone to build AI into a workflow rather than to produce a single deliverable.

The gap in ticket size between the two platforms shows how differently they started. Upwork's GSV per client of $5,230 is about 14 times Fiverr's annual spend per buyer of $373. Fiverr was built for the small, fixed-price order. That is exactly the kind of order a general-purpose AI model now completes for the price of a subscription.

Both freelance platforms cut staff harder than their marketplaces shrank

Fiverr's 30% cut drew the headlines in 2025. Upwork's cut drew far fewer, and it is proportionally almost as large. On 7 May 2026, Upwork announced a restructuring "that includes a reduction of the Company's total workforce by approximately 24%".

By 30 June 2026 it had reduced its workforce by about 20%, with the rest due by year end. The quarter carried $13.8 million of restructuring charges and $12.4 million in retention awards for staff who stayed.

The comparison in the figure above is the uncomfortable one. Fiverr's workforce fell 30.7% while its trailing marketplace volume fell about 16%. Upwork's planned workforce reduction of 24% compares with a 3.6% decline in GSV. Neither company is cutting staff in proportion to lost work. Both are cutting in anticipation of it, and in the name of their own AI tooling.

I would not take either company's staff cut as evidence of AI productivity. A company that cuts 24% of staff while revenue is roughly flat is making a margin decision. That may be right. It is not the same as showing AI did the work, a distinction covered in detail in the test applied to Cloudflare's AI layoff claim.

Where this argument is weakest

The neat story is that AI replaced the small freelance gig and created a smaller market for AI build work. The filings support that story. They do not prove it, and there are two competing explanations you should weigh honestly.

Some of the decline is search traffic, not substituted work

Fiverr's Q3 2026 outlook blames "AI-related disruption to traditional SEO and SEM". It says the "shift from search to AI-driven discovery" will drive a steeper decline in marketplace traffic through 2027. Upwork's management made a similar point about search on its Q2 call.

That is a different mechanism. If buyers still want a logo or a translation but no longer find Fiverr through Google, the work has not been replaced by AI. It has been rerouted by AI search. The filings cannot separate the two, and the traffic effect is closely tied to the pattern described in the analysis of zero-click search and what it does to referral traffic.

Two companies is a thin sample, and the metrics do not match

Fiverr and Upwork are the listed freelance marketplaces, not the freelance market. Work may be moving to direct relationships, agencies, or talent networks that file nothing. Fiverr's buyer count is trailing and marketplace-only. Upwork's GSV includes enterprise and value-added services. The Upwork AI share and non-AI decline above rely on a call figure reported secondhand.

Macroeconomic weakness also plays a part. Upwork names it alongside AI in its 10-Q. Small businesses cut discretionary freelance spend in any slowdown, and nothing in either filing allows you to split the effect. Broader labour data has the same problem, as the reading of AI job losses in payroll data shows.

My position, after weighing both: the buyer collapse at Fiverr is too steep and too steady to be macro alone, and too concentrated in small orders to be search alone. AI substitution is the most likely main driver. I hold that at moderate confidence, not high.

Frequently asked questions

Where to start if you buy or sell freelance work

If you buy freelance work, sort last year's freelance spend into two piles this week. One is orders under $1,000 with a single deliverable, the other is longer engagements. Run the first pile through the AI tools you already pay for before you reorder. The filings say most buyers in that pile have already left the marketplaces, and for good reason.

If you sell freelance work, or run an agency that does, price and package for the second pile. The growth both companies report sits in projects where a person is accountable for integrating AI into a working process. That shift mirrors the move described in the case for selling outcomes instead of hours. Then put a reminder on Upwork's Q3 2026 release in early November and check one number: whether AI-related GSV growth fell again.

Related reading