Fiverr International Ltd. Q2 2026 Earnings Call

Good day, welcome to the Fiverr second quarter 2026 earnings conference call. All participants will be in listen-only mode. Should you need assistance, please signal a conference specialist by pressing the star key followed by 0. After today's presentation, there will be an opportunity to ask questions. To ask a question, you may press star then 1 on a touch-tone phone. To withdraw your question, please press star then 2. Please note this event is being recorded. I would now like to turn the conference over to Emily Greenstein, Senior Investor Relations Manager. Please go ahead. Thank you, operator, good morning, everyone.

Thank you for joining us on Fiverr's earnings conference call for the second quarter that ended June 30th, 2026. Joining me on the call today are Micha Kaufman, Founder and CEO, and Esti Levi-Dadon, CFO. Before we start, I would like to remind you that during this call, we may make forward-looking statements and that these statements are based on our current expectations and assumptions as of today, Fiverr assumes no obligation to update or revise them. A discussion of some of the important risk factors that could cause actual results to differ materially from any forward-looking statements can be found under the Risk Factors section in Fiverr's most recent Form 20-F and other filings with the SEC.

During this call, we'll be referring to some key performance metrics and non-GAAP financial measures, including adjusted EBITDA, adjusted EBITDA margin, and free cash flow. Further explanation and a reconciliation of each of the non-GAAP financial measures to the most directly comparable GAAP measures is provided in the earnings release we issued today and our shareholder letter, each of which is available on our website at investors.fiverr.com. Now I will turn the call over to Micha.

Thank you, Emily. Good morning, everyone, thank you for joining us. Q2 was a transitional quarter, reflecting an ongoing compression in high volume, low value transactional work driven by accelerating AI adoption. As LLMs continue to evolve rapidly, we're seeing our customers accelerate their adoption of AI and workflow automation. Today, high-value work represents 15% of our completed projects' gross order value and is expected to continue growing while the transactional base that makes up the remainder continues to compress. That mix is why the strength of the high-value business is not yet visible in our headline numbers and why every investment we are making is aimed at shifting it. Based on this, there are two reasons we are excited about the up-market shift: the ability to attract untapped demand and the opportunity associated with projects between our current spend per buyer and high-end projects.

Over the last few weeks, we've seen clear deceleration in overall marketplace traffic and demand, a trend that has carried over into Q3. We believe that recent model updates across various LLMs are a contributing factor behind this softness and are confronting these headwinds head-on. We are not competing for that work. Instead, we are aggressively executing our push-up market toward the higher value work where demand is growing. Navigating this structural shift requires intense financial discipline and thoughtful capital allocation, which Esti will unpack shortly. We have both the team and balance sheet to execute through this transition, and our focus is to deploy our resources where they generate the highest long-term return. Fiverr's multi-year transformation is fundamentally about moving from a transaction-oriented marketplace to a trusted end-to-end work platform for high-end, high-value projects.

Put simply, AI is automating simple tasks. Fiverr is moving toward larger, longer duration projects where AI deployment meets human judgment, strategic partnership, and accountability. This is a profound evolution in how the work on our platform is matched, delivered, and managed. Transformations of this magnitude require patience. We expect the financial impact to build over several quarters. Our North Star remains unchanged, positioning Fiverr as the ultimate destination for high-end, high-trust work. While the financials will take time to catch up, I believe that the underlying operational metrics are proving our strategy. Let me illustrate that across the four strategic pillars we outlined in our shareholder letter. First, our push-up market continues to validate our strategy. Clients completing projects valued at over $1,000 continued to grow this quarter at 13% year-over-year on a trailing 12-month basis.

While the macro environment is volatile for smaller buyers, we see areas of strength in our high-value buyers, particularly around programming and tech and graphics and design. Clients completing projects over $1,000 in programming and tech grew 34% year-over-year, while graphics and design grew 25% year-over-year on a gross order amount and TTM basis. From an AI services vertical perspective, we're seeing a clear evolution in how clients deploy AI with us. In commercial video and content generation, clients are using AI to scale their performance marketing. Demand for AI user-generated content video ads surged 265%, while AI video ads rose 63%. Businesses aren't just experimenting, they are actively using freelancers to produce commercial-grade media at a fraction of traditional agency costs. In technical integration and agentic workflows, buyers are moving past basic chatbots toward complex production setups.

Search volume for AI voice agents jumped 49%, AI mobile app development rose 92%. AI website development grew 39%. The biggest opportunity here is the shift from generic AI to what I call personal AI built into the specific workflows of a specific business. Businesses, both tech and traditional, understand that automating their businesses is not a nice-to-have decision, but a do or die decision to stay competitive. These are demand signals from our own marketplace. As foundational AI models get more powerful, the execution gap widens. Businesses need skilled talent to orchestrate these tools into functional revenue-generating outcomes. This quarter, we also continued to see businesses come to us for those multi-phase mission-critical projects. For example, 1, a growing distributor that recently used a multidisciplinary freelance team on our platform to architect their entire B2B wholesale inventory and logistics ecosystem.

This isn't just custom code, it's deep operational software running their day-to-day supply chain, ultimately setting up for a phase 2 rollout of a retail POS. A manufacturer replacing a decade-old legacy system with a ground-up rebuild of their custom 3D design software. By tapping into top-tier engineering talent on Fiverr, they're integrating their new ERP system and driving a complete digital transformation. Lastly, an entrepreneur building a stadium access platform complete with secure ticketing and automated post-event photo monetization. By coordinating multiple APIs and webhooks across payments, storage, and analytics, our freelance talent is delivering complex integration-heavy consumer platforms at scale. This is where the market is headed, and our target remains this segment of long-duration, high-value engagement. Our second pillar focuses on systematically upgrading our matching infrastructure to make trust and quality native to the user experience.

We are leveraging our proprietary Knowledge Graph to capture incredibly nuanced client intent across four vectors: the client, the talent, the scope, and the order itself. We recently began development of live talent skill extraction capabilities, and based on early testing results of over 450 mismatch events, we were able to structurally resolve as high as 58% of conversation skill mismatches, directly tackling a major pain point for high-value order cancellations. Additionally, our newly developed graph neural network model, which helps improve matching capabilities, showed a 7% decrease in cancellation rates for dynamic matching compared to our current model in initial tests. Our third pillar is about moving Fiverr beyond passive matching into an active, comprehensive work platform. We are building a standardized fulfillment layer engineered to dynamically safeguard project success. Phase 1 of this framework is now live in production.

It is currently evaluating transactional quality, reaching as high as a verified 91% precision rate across over 84% of completed projects in our initial testing. This architecture is designed to give us real-time telemetry and sentiment signals. It allows us to track project health, mitigate friction points before order completion, and create the exact infrastructure needed to seamlessly integrate our marketplace with advanced agentic business workflows. Our fourth pillar centers on expanding our growth engines to attract the right kind of demand. We have moved our advanced semantic onboarding models and corporate KYC initiatives into production. These systems capture deep corporate profile data the moment a business joins en route that demand directly to our highest-performing talent. Our development test results drove an 8% lift in new customer conversion. We are also opening up new, highly targeted acquisition channels.

In June, we launched a dedicated discovery campaign targeting e-commerce merchants looking to scale their TikTok shops. Our goal is to capture higher-value merchants, including Amazon FBA and Shopify merchants, and funnel them into repeatable long-term service journeys on Fiverr. To close, we are repositioning Fiverr deliberately to where the value is higher end, high-trust work where human expertise, project orchestration, and accountability are irreplaceable. The operational signals across our strategic pillars tell us we are on the right track, and our job now is to translate those signals into durable financial performance. Backed by our balance sheet and focused execution, we are building the foundation for our next chapter while remaining disciplined in how we allocate capital. With that, I'll turn it over to Esti for the financial details.

Thank you, Micha. Today I'll review our second quarter financial performance and provide additional visibility into the ongoing marketplace transition, an update on capital allocation, as well as third quarter and full year 2026 financial guidance. Starting with the results. Second quarter revenue was $97.8 million, down 10% year-over-year due to a decline in low-value transactional work. Adjusted EBITDA was $17.5 million, down 18.3% year-over-year and representing an adjusted EBITDA margin of 17.9% as margins declined 180 basis points from a year earlier. Margin pressure was limited by the proactive steps the team has taken to maintain a lean organization, including our continued efforts in discretionary spend management, internal AI efficiencies, and balanced marketing. Turning to our revenue segments.

Q2 marketplace revenue was $63.1 million, declining 15.5% year-over-year, driven by 2.7 million active buyers, $368 in spend per buyer, and a 28% marketplace take rate. The 22% year-over-year decline in active buyer was mainly due to the ongoing compression in low-value transactional work. Low-value transactional work currently represents the majority of our marketplace and is decelerating at a faster and stronger rate than the growth associated with our high-value work. Additionally, as Micha noted, in recent weeks, we observed a noticeable deceleration in marketplace traffic and demand, which has carried into Q3 and is expected to continue for the foreseeable future. These traffic and demand trends impact the entire marketplace. This pressure was broad-based across simple categories in all verticals. Transaction volume was down 10% or more year-over-year across the majority of projects under $1,000 on a TTM basis.

Writing and translation saw the steepest decline in more than 24%. Spend per buyer increased 15.6% year-over-year, driven by the ongoing mix shift across the marketplace from low-value transactional projects to 1K plus projects. At the end of Q2, projects at or above 1K represented 15% of completed projects' gross order amount on a TTM basis. Programming and tech and graphic and design represented the fastest-growing verticals of completed 1K plus projects, with both increasing over 25% year-over-year on a TTM basis. On the other end, services revenue in Q2 was $34.6 million, up 2% year-over-year, and accounted for 35% of total revenue. Services revenue continued to grow, but at a lower rate compared to last quarter. The deceleration is expected to continue into H2 and exit the year with a double-digit decline.

The drivers for the deceleration include softening demand from influencers campaigns and drop shipping, as well as weakness in Fiverr Ads and Seller Plus. Services revenue is also negatively impacted by the lower expectations around marketplace growth given the traffic and demand issues we've discussed. Now on capital allocation. Our capital allocation priorities are guided by a disciplined and balanced framework. First, funding the organic transformation investments required to reposition towards high-value work. Second, maintaining strategic and financial flexibility in the dynamic AI automation environment. Third, evaluating opportunities to generate value for shareholders, including further repurchases. In the current environment, maintaining a strong balance represents the most responsible path forward. Our healthy financial position provides valuable flexibility as we continue to run a lean organization focused on cost discipline and maintaining profitability. At the same time, we recognize that capital allocation is a critical component of value creation.

We finished the quarter with a total cash and investment balance of $308.5 million and generated $13.6 million in free cash flow. Given the uncertainty, full year free cash flow is expected to be lower than the previous two years. We are operating in a dynamic environment, we will constantly evaluate where we deploy our capital for the highest return. That also applies to how we invest in our transformation and related growth opportunities. A primary focus will be on executing appropriate cost discipline so we can continue to generate profit and free cash flow. Moving to financial guidance. Our revised guidance for the third quarter and full year 2026 reflects the AI-related demand and traffic headwinds observed in recent weeks, which have continued into Q3, impacting our entire marketplace, ongoing weakness in categories most exposed to AI automation, and declining services revenue.

We now expect that meaningful financial impacts from our transformation will require at least six quarters to materialize. For the third quarter of 2026, revenue is expected to be between $80 million-$88 million, representing year-over-year growth of -26% to -18%, and adjusted EBITDA between $8 million-$12 million, representing an adjusted EBITDA margin of 11.9% at the midpoint. For the full year 2026, we expect revenue to be in the range of $356 million-$372 million, representing an year-over-year growth of -17% to -14%, and adjusted EBITDA in the range of $52 million-$62 million, representing an adjusted EBITDA margin of 15.7% at the midpoint.

While our updated outlook reflects our view of the current operating reality and the extended timeline for our transformation, we are leaning into this moment with discipline, stabilizing the core marketplace, shifting towards high-value work at a fair pace, preserving flexibility, and allocating capital with a focus on long-term value creation. With that, we will now turn the call over to operator for questions.

We will now begin the question and answer session. To ask a question, you may press star then one on your touch tone phone. If you are using a speakerphone, please pick up your handset before pressing the keys. If at any time your question has been addressed and you would like to withdraw your question, please press star then two. At this time, we will pause momentarily to assemble our roster. The first question today comes from Eric Sheridan with Goldman Sachs. Please go ahead. Thanks so much for taking the questions.

Maybe two, if I could. Just drilling down on the six-quarter transition period, can you give us a little bit of more granularity about what is it about either business mix or the headwinds and tailwinds you see in the business today that underpin some of the framing of the duration of the transition across that type of time period? The second would be, you highlighted the recent updates of various LLMs as a contributing factor to the weakness. What exactly are you seeing out of those LLMs, and how is that sort of manifesting in these headwinds, if we could get a little bit more detail there as well. Thank you so much. Hello.

Good morning, Eric. Thanks for the questions. As for the first one, the answer is yes, it's mostly headwinds, which really impact our calibration or estimation of how long this transformation is going to take. That's mainly what's influencing this. As Esti mentioned in the opening comments, we run a very lean organization at the minimal size that is required to move very fast. We're now a much smaller organization than we used to be as we've done the restructuring, but the pace in which we move is much faster. Given the fact that we assume the trends that we started seeing in the last couple of weeks of Q2 going into Q3, the assumption is they'll continue, and this will be the time that we feel the transformation is going to take. I did mention the impact of a few of the LLM models.

Again, between those LLMs, obviously, there's different types of usage, and there are some models that are more impactful or step function versus those who are just adding some improvements. The really interesting thing is that it's not just the LLMs themselves or the specific models, but it is how those LLMs are being integrated into other client experiences. I think most noticeable is how Gemini is being integrated into the experience of search on Google. It exposes more people to LLM experiences, which by turn, obviously impact the traffic that we're getting. Traffic used to come only from search, then it became LLMs as well, and now it's becoming hybrids of LLMs and traditional search, which essentially lowers the volume of traffic, which we called out as one of the major headwinds.

Great. Thank you. Thank you.

The next question comes from Ron Josey with Citi. Please go ahead. Great. Thanks for taking the question.

Micha, just a quick follow-up on Eric's question on the timeline and the 6 quarters. Just can you give us some idea or steps that perhaps we can look into or watch in terms of this transition to see just the progression as we go to higher value projects up market? Then to that, as we do get to higher value projects and given the foundation models get more powerful, just talk to us about the talent on Fiverr's marketplace and the ability to basically answer or deliver these higher value projects. Thank you. Thanks for the questions, Ron.

Good morning. I'll start with the first one, the transformation itself. We've been calling out in the opening remarks what we're actually building there and what are the early signals that we've seen from tests. Those signals or those tests are being deployed more widely, which is why we expect to get more impact out of them. This transformation is really about the matching parts of our proprietary Knowledge Graph. As we said, the initial deployment of seller skill extraction, which addresses about 58% of mismatches in a model that decreases by 7% the high-value cancellations. These are really impactful numbers. Since they've been running on tests, they're being deployed widely on production. The product has now an end-to-end fulfillment layer. The phase 1 of it is live.

It's evaluating the quality of transactions with 91% precision across 84% of the completed projects. The go-to-market transformation, again, we've developed and tested new go-to-market, I've mentioned TikTok as an example, which drove 8% increase in new customer conversion. Again, deploying this at a larger scale. Obviously, we do all of this through keeping this operational excellence. Making critical investments to strengthen the high-end talent flywheel and focus on improving marketplace quality, prioritizing high-value work, and executing with very strong financial discipline as we do this.

Now, when we look at, this is the second part of your question, when we look at the type of skills, a part of what we're doing in going much faster up market means that we're both cleaning our talent pool from unnecessary, those who don't have demand for low skills, which is really important because they occupy space on the marketplace. On the other end, defining the necessary skills, ensuring that if we're missing in some place the right talent to tackle these needs, these client needs, we have that talent, and if not, we onboard them very fast. We started calling some of these project types in categories, but I think the overarching message here is we're not talking about very specific small AI deployments for our clients. Our clients are interested in AI integration.

They want to improve the competitiveness and the efficiency of their entire business, not just put a customer support agent. They want to integrate their entire business workflows. That requires highly skilled professionals that can assess and understand the actual business needs and understand how to create multiple agents that are tailored for that specific business, so they can extract the most out of this AI transformation that, as I've said, is not a nice to have, but a do or die for businesses. The first businesses that are going to do this are going to be more efficient and going to be more competitive.

Thank you, Micha. Thank you.

The next question comes from Jason Helfstein with Oppenheimer. Please go ahead. Hey, thanks for taking the question.

Just thinking about the financial outlook. You made the point you already run a very lean organization, as you said. I guess what's the right way to think about EBITDA and free cash for next year? Obviously, you're not giving us guidance, but if revenue's down next year, are there other cost actions you can still take? And then I guess, how far along are you in just deploying AI internally for your business and automation in the organization? And if you want to put those two together as one answer, but that's the question. Thanks. Thank you, Jason. Good morning.

Our revised guidance for EBITDA reflects first the issues which we're seeing on traffic, and that affects the bottom line. However, as you know, we took some reduction cost initiatives in previous quarters that help us to support EBITDA. In addition, as we see traffic issues, we also adjust our marketing spend as we did before, and we will continue to assess that if needed. We are protecting our R&D spend because that supports the transformation, and this is our number one priority, to be able to go through the transformation, and also continue to be profitable. If you think about it during the transformation, this would be a firm to continue to generate cash. We continue to run an organization and we continue to be with cost discipline, especially about discretionary costs, marketing, as I said.

Again, the focus is going through the transformation.

Yeah. To answer your second question, essentially, the first half of the year was putting all of this infrastructure together and running multiple tests, and I've called out some of them, to prove that doing so is actually significantly improving the 1K transactions and the right clients. What we're going to do in the second half of the year is deploy some of these solutions at scale and continue to work on new things that we feel are going to impact and accelerate the ability to drive a quality experience, better conversion, and optimize for a higher spend of those customers.

I would just add on the free cash flow. Free cash flow follows EBITDA, you should expect the guidance of the EBITDA should be the range for the guidance for the free cash flow. As I said, we are aiming to continue to be profitable, and that means also to continue to generate cash.

The next question comes from Nat Schindler with Scotiabank. Please go ahead. Micha, this is kind of a big-picture question.

For years, you guys have said that really the move to AI is not such a threat because you are basically just a marketplace for freelancers, and freelancers are the most adaptable people in the entire economy. They very rapidly change to what is necessary and the jobs that need to be done. As a marketplace, you really don't care what those jobs are. You can just provide the connection points. That's a really compelling argument. In February, it seems like the release of Claude was kind of a watershed moment, and the world is now really doing more and more things without touching these freelancers, in a sense. That's why the revenue seems to be declining as much as it is. I get the transformation. I get the need that companies will have for people to help them do these transitions.

How can you be confident that in six quarters there aren't going to be more watershed moments where this gets worse and this gets more and more that you can do more and more with less than you need, and the freelancers will face their first time where even their renowned adaptability just isn't enough?

Good morning, Nat. Thanks for the question. First of all, I should say we stand behind everything we said, and it is true, it still is true. What I think we are seeing, this is also why this requires additional time for this transformation to be highly successful, is the fact that very similar to previous transformations, technology transformations, or industrial transformations before, the rate in which some skills get compressed versus the rate of the creation of new skills is not the same. The creation of new skills, the orientation of our clients to understand how do they navigate this new world, putting names to skills, ensuring that talent is qualified to tackle those needs is a cycle, and they don't happen at the same rate. We are seeing this because we have deep expertise in this economy. We constantly have discussions with clients.

We're seeing developments in the market, it takes time, but they actually realize that this is, as I've said, not a nice-to-have. It's not a choice. It's like 30 years ago when people thought that having a website was optional. It took time for businesses to understand that if they don't have an online presence, they are going to start losing some of their businesses to their competitors. The same applies for AI. It takes time for businesses to understand exactly what they need. It's still true, freelancers are in the forefront of this curve. They do understand that they obtain these skills fast, but as I've said, it's not the same rate of the compression that we're seeing.

Because the high-end is still a small portion of our business, the growth that we're seeing there, in some cases very high growth, is being masked by the compression in the low-end. This increase gives us the confidence that this transformation is the right thing, and we should be very committed to it. The early signals that we've seen from our testing is showing this, and that's why we move full force, fire on all cylinders, to make that transformation as fast as possible.

Just to follow up on that, as you make a transition to the high-end, and it sounds like the high-end is evolving to even higher end than previous discussions of high-end. Is that moving you more into just being a staffing placement company? Where you're doing, as opposed to what you have been so long and been so capable at, is being a marketplace for completed business services. It's very hard to define and limit a scope of project in a single listing. This is a, "I need someone who has this ability." Are you just totally changing what you are in that respect?

Yeah. I've addressed this also in previous calls, and the way we think about Fiverr and our role in this ecosystem is to be very focused on the project and the outcome-based work. For us, identifying the need, in some cases, helping the client better qualify that need, then understand through the Knowledge Graph who's the exact most qualified talent, and in some cases, multiple talents, to address and be able to achieve and fulfill that goal. Making that matching, but also, as I've said, taking a larger portion in the fulfillment of this actual outcome-based result is the function, and that is what defines us and where we focus. This allows us to be very agile in this.

Okay. Thank you. Thank you.

The next question comes from Bernie McTernan with Needham. Please go ahead. Hi, this is Stefanos Crist for Bernie.

Thanks for taking our questions. The projects above $1,000 are 15% of GOV. What mix do you expect that to be to offset the decline in the legacy low-value work? Just on capital allocation, are there any assets out there that can help accelerate your plans, or are you only focused on internal investment? Thank you. Thank you, Stefanos.

While the 15% seems small, but you need to refer that currently our center buyer is at $368. We have a lot of room to grow with the $1,000. It will take time, as we said. We are changing fundamentally our platform. It comes with, as Micha mentioned, a lot of the talent, the matching, the orders, of course, the go-to market. It would take time. However, we see the opportunity both within the marketplace, within our current buyers, and also as we go with new go-to market. One example is the TikTok example, but we are planning, of course, additional partnerships and additional go-to channels. As for our capital allocation, our primary focus is definitely going through the transformation, and we're investing on that. On the M&A side, nothing to call out.

We're always opportunistic, again, top priority is to do the transformation within the platform. We said we're going to invest in R&D. In order to do that, we have the sufficient staff, and we will run lean organization while doing that, as Micha said. We're going to do it fast, but with a leaner organization and not planning currently doing inorganic.

The next question comes from Brad Erickson with RBC Capital Markets. You may go ahead. Hi.

Thanks. I just want to go back a little on the paper trail of the weakness, if I could. I realize it's hard to know 100% for certain, but you mentioned Google's distribution with kind of integrating LLMs into search and being a driver of the weakness, which makes sense. I'm just curious if there's a component from some of these more recent powerful model releases. I guess the question is this kind of a marketing issue where Google's making people more aware of these tools being available and they're using them? Or is it more just that newer models are just making it easier to do kind of a wider range of things? Or if it's both, I guess, just curious which of those two do you feel like you're seeing more of lately? Thanks. Thank you for the question, Brad.

The weakness of demand and traffic comes from Google, but it has also, as I've mentioned, the impact of integrating LLM or AI summaries in the Google search. By definition, for some customers, they would go directly into the LLM instead of clicking on either paid or organic links. I think it's very visible, the fact that the majority of link anyway are paid. The organic is actually getting squeezed down, which means that the impact of SEO is going down. This is why we and probably every other company is making also an investment in GEO or generative engine optimization to make sure that we are inside those LLMs, and we are. The click-through in LLMs in general is obviously smaller than it is on search. All of these factors actually create headwind on traffic.

The same goes for other models. Sometimes it's the introduction of newer tools that could be in the case of Claude Code or Claude Design or Claude Cowork. Right now we have Kimi coming in a Chinese alternative that is coming in to compete, sometimes it's upgrade of a newer model on ChatGPT. Our assumption is that these models are going to continue developing and competing with each other depending on their usage and their integration into other things might influence traffic. That said, I'm not going to repeat that, the actual needs of the clients that we're aiming for is not being addressed by just using one of these tools. It requires very deep integration of agentic solutions that are well beyond reach for the vast majority of businesses.

Even high-tech businesses, I know this from firsthand experience, are not having an easy time creating a more agentic organization. If it's hard for such a sophisticated company like Fiverr, I've been talking to many other companies, it is 1,000 times more complex for traditional businesses. Therefore, they will need help, they will need this idea of human in the loop.

Got it. Then just bigger picture, you have such a great lens into the world of models and harnesses and so forth. I'm just curious, could you just maybe rest a second and give us a little bit of your thinking around just the open source topic in particular, we won't talk Frontier specifically, of course, but just software looking to find any way to build a moat, lock their customers in, versus working with open models. How are you kind of thinking holistically about that? I don't mean certainly for your business, but also just how your customers compete at that level. How do you kind of think about that?

Right now, if you think about it, most companies in the world, I'm excluding maybe the forefront AI companies, are still pretty much behind. If you look at highly sophisticated companies like ours, the type of challenges that we obsess over is inference, for example, because there's now almost endless amount of different models, each one slightly better in different tasks. There's also a question of how fast do you need it? By introducing latency, you can actually pay less for the same model, assuming you don't need the answer or the output right away. For us, as an example, it's not just how do you extract the most out of a certain model, how do you use multiple models from multiple foundations to actually maximize this task or the skill that you're trying to solve?

Also bake price into it because you might use very advanced models that cost a lot. Fable costs twice than Sonnet, but for many different tasks, it's simply not needed. Throttling between those, dealing with inference is one of the big topics right now. As I've said, I think that this is way too advanced for businesses that are tackling much more basic needs. Even those businesses understand that the deployment of AI is not free. Meaning when you deploy AI, it's not one and done. Running agentic environments has a cost, and they want to be aware of it. Some of the things that our experts are doing is helping customers ensure that they don't overspend for the wrong things. This is why it's so complex.

This is why I keep telling that this is beyond the capabilities of most businesses and why they need help with it. I think the question of open source or not open source is not a big difference. Open source just allows for the foundational companies to harvest data faster, so they can make it slightly cheaper. We're all working for foundational companies anyway. If it's good enough, you should probably pay less for it. Okay, I riffed enough. Thanks a lot.

Thank you. The next question comes from Matt Condon with Citizens.

Please go ahead. Thank you for taking the questions.

My first one, just want to go back to this traffic question and see, are you seeing a significant deterioration in your payback periods in other channels? I'm specifically thinking about, is this isolated to SEO or Google traffic? Are you seeing your other channels also deteriorating performance? My second question is, would you ever consider dynamic pricing or any changes in pricing to try and spur demand across the marketplace, or is that just not a focus for you guys? Thank you so much. Thanks for the questions.

Essentially the situation, if we look at Google as an example. First of all, the fact that there is less customers coming to Google is impacting the amount of traffic that Google has in its hands to advertise for. Meaning that they increase advertising density, which means that there's more competition for each placement, which means that the cost of acquisition is higher. All of this influences the efficiency of marketing, which is why we take a multi-channel approach. Which is why we're investing in SEO to increase the organic traffic from LLM channels. It is showing great results, but on a click-through, it's still small. We were also one of the first advertising partners on ChatGPT, but it is still very early days. It reminds us when we started working of the early days of working with Google.

It takes time to build up. They're figuring out, we're figuring out the strategy, but we are taking proactive measures. What we want to make sure is we want to make sure that the investment that we're doing, this transformation in the matching infrastructure and the fulfillment workflow is exactly enabling Fiverr to seamlessly integrate into agentic workflow down the road so that we can give more of our core experiences and core solutions inside these LLM or whatever they become experiences to entertain the customers where they are.

The next question comes from Marvin Fong with BTIG. Please go ahead. Great. Good morning.

Thanks for taking my questions. Two, if I may. First one maybe bigger picture. With these increasing sophistication of the projects that you're targeting. Do you feel like your current suite of services, Fiverr Business, Fiverr Pro is enough to address sort of this coming landscape? Or are you looking at producing additional service channels for your clients to create something a little bit more comprehensive perhaps? Second question, just a little more focus on the services revenue, called out that Seller Plus would be down. I'm just kind of curious, is that because freelancers are canceling their subscriptions, or is it that the actual pool of freelancers is kind of shrinking on the side? If you could just kind of provide some insight on what's going on the freelancer side with respect to services, that'd be great. Thank you. Thanks for the questions.

As we're doing this transformation, we're basically rebuilding a lot of the core functionality and the core product from the ground up. I've called what we're doing with the Knowledge Graph, what we're doing with the end-to-end fulfillment layer, the matching engine, and the type of experiences that we expect to provide our customers that are better than the best AI out there in the market. I don't want to go into those details because this is being built and is going to be tested and deployed throughout the year. We're working very hard to make sure that those clients that we focus on receive the best experience that they can get to ensure that they achieve their outcome. We're building new tools. Again, we're not getting into talking about these tools, and they're being built, but we are refining the platform to ensure that this is exactly the case.

As for your second question about services revenue. Services revenue and seller monetization, it's affected by the traffic of the marketplace. As the traffic headwinds that we saw in recent weeks affect the Fiverr Ads and the Seller Plus. We expect that also to be continued in H2. In addition, in services revenue, we saw some headwinds also on AutoDS, and that's also being included into our H2 guidance for services revenue.

Okay, great. Thank you, Micha. Thanks, Esti. Thank you. The next question comes from Josh Chan with UBS.

Please go ahead. Hi. Good afternoon, Michal, Esti.

I guess two questions. First, I guess as you do this transition, standing here, I guess how much of your talent or customer base do you feel like has to switch out or cycle through to complete the transition? How do you know how to acquire the right talent and right customer for the next phase? The second part is, when you give the six quarters, I guess, estimation, what's the starting point of that six quarters? As you deploy the solutions that you're testing, should you see some improvement in the trends even during that period? Thank you. Josh, thanks for the question.

Good morning. I think I made that comment earlier, which is, we're both cleaning up the lower-end part of our talent that we feel has lower demand to make sure that we optimize the display layer of our marketplace and allow our clients to find what they're looking for easier. The rest goes for the high-end talent. As I've said, identifying the skills on demand, and in some cases, defining them, helping give those skills names, is really important. It's about qualifying them. We said in many earning calls that Fiverr's, one of our biggest moats is the fact that we've been the largest transactional marketplace in the world, and we've collected so many data points that using them in the right way, which is how we build the Knowledge Graph that contains very deep understanding of the client and their needs, and understanding the actual short description and the desired outcome, and pairing them with the right talent that can address exactly that.

Since we have such a rich data of actual transactions where we have high level of trust of the talent that we have, the matching becomes much more efficient, resulting in much happier outcomes, which then also drive higher retention and higher spend.

Josh, as for your second question, the six quarters, it's starting now. Yes, of course, as Michal described, although we're still at the beginning of the period, we have some initial things that we're seeing that are encouraging. Definitely, we'll share more as we go to show all of the progress.

Great. Thank you both for the color.

This concludes our question and answer session. I would like to turn the conference back over to Micha Kaufman for any closing remarks.

Thanks so much. Thank you, Chloe, for moderating this call, and thank you everyone for calling in, and I wish you a great day, and we'll talk soon.

Thank you. The conference is now concluded. Thanks for attending today's presentation.