Which businesses are being disrupted by AI?
In April, we argued that transactional marketplaces were proving more resilient to AI than subscription and lead-generation models. In July we looked at why public markets were rewarding some marketplaces while punishing others.

Public markets have now had four years since the launch of ChatGPT-3.5 to decide which online business models AI is most likely to disrupt. We have followed 10 company market caps from the end of 2021, the last full year before generative AI reached consumers, until November 2026. These 10 listed businesses in sectors most exposed to AI, including price comparison and travel search, creative and content tools, freelance marketplaces and learning, have lost $206bn of market value between them, falling from $305bn at the end of 2021 to $99bn in November 2026.

Interestingly, revenue has until recently held up; nine of the ten grew revenue between fiscal 2021 and fiscal 2025 with their combined revenue rising 47% from $22.2bn to $32.7bn, with Chegg the only one to shrink. Investors are pricing what these businesses will earn in future, and the first signs are now coming through: in the third quarter of 2026 revenue fell 23% at Tripadvisor's hotels business, 25% on Fiverr's marketplace, 33% at Shutterstock and 40% at Chegg.
The businesses that appear most exposed share a few characteristics. Many of them gather non-proprietary information, organise it, and earn a subscription, an advertising fee or a referral when the customer finds what they were looking for. Others are built around a single task that is fairly standardised, such as producing a slide deck, a logo or a paragraph of marketing copy. These businesses tend to lose the customer once a model provider, or a platform with existing distribution, offers the same or similar inside its own product set. Freelance marketplaces appear to be affected in a similar way, even though they take the payment, because a buyer can now generate much of that work directly without hiring anyone.
Below we go through the sectors where the loss of value is clearest, what the companies involved are doing about it, and where we think the next wave of value creation will come from.
Price comparison and travel search
Price comparison was one of the pioneering internet business models of the 2000s: Yahoo paid €475m for Kelkoo in 2004, and in 2008 Microsoft paid $486m for Greenfield Online in order to own Ciao.
The model had a structural weakness, because a comparison site owns neither the supply nor the customer; it acquires traffic from a search engine and resells it to a merchant as a lead. Once Google placed its own shopping results above theirs, the traffic and the margin went with it; Ciao, then one of Europe's largest comparison shopping sites with 26.5m monthly visitors, shut down in 2018.
LLMs are now doing to metasearch what Google did to comparison shopping. A traveller who asks an assistant for the best-value hotel in Lisbon gets the comparison inside the answer, and never sees the comparison site.

Tripadvisor is down 77%, from $3.8bn in December 2021 to $0.88bn in November 2026. Hotel revenue fell 23% in the third quarter of 2026 and the company wrote down $370m of goodwill, citing accelerating traffic losses from changes in search. Its response is to move towards the booking: it has sold TheFork, its European restaurant-booking platform, for roughly $741m and is reorganising around Viator, its travel experiences marketplace, which now generates 65% of group revenue.
Trivago has seen a similar loss of value over the period, falling by 67%, from $0.78bn in December 2021 to $0.26bn in November 2026. Revenue is growing again, and its fastest-growing product is Book & Go, which takes the booking on Trivago itself and more than doubled its share of referral revenue over the past year.
LendingTree, the US loan and insurance comparison site, is down 68%, from $1.6bn in December 2021 to $0.52bn in November 2026, even though revenue grew 22% in the fourth quarter of 2025, 25% in the second quarter of 2026 and 12% in the third. Investors are discounting that growth because LendingTree is paid for introducing a borrower to a lender, and they expect an AI agent that can compare loan offers on the borrower's behalf to take over that introduction.
Booking Holdings' market value grew by 20%, from $98.5bn in December 2021 to $118bn in November 2026. It wrote down Kayak, its metasearch subsidiary, by $457m in 2025 after Google's AI Overviews cut Kayak's free search traffic, while room nights booked across the group roughly doubled from 591m in 2021 to 1.2bn in 2025. We think Booking has proved resilient because an AI assistant can suggest where to stay, but the reservation and the payment still take place on Booking, which earns its commission on a completed transaction.
Creative and web software

Adobe is the largest casualty of the 10 in absolute terms. It was worth $269bn in December 2021 and $92bn in November 2026, a fall of 66%, while revenue grew from $15.8bn in fiscal 2021 to a guided $26.6bn in fiscal 2026. The market is questioning how many creative seats will be needed when images, video and layouts can be generated from a prompt. Adobe's response has been to build generative AI into the suite: it reports AI-first ARR of over $650m, up more than 150% year on year, although that remains a small part of $27.5bn of total ARR.
Wix has lost 68% of its value, falling from $9.0bn in December 2021 to $2.9bn in November 2026, despite growing revenue from $1.3bn in 2021 to $2.0bn in 2025. Investors are pricing the risk that a small business owner can now describe a website to an LLM and have it built; Wix has responded by buying another disruptor: it acquired Base44, an AI app builder, for $80m in June 2025, and by the third quarter of 2026 Base44 was nearing $250m of ARR.
In both cases the direction of the pivot is the same: adopt the technology that threatens the core product before a challenger does. The market remains unconvinced, and both companies trade at around a third of their 2021 value.
Shutterstock has lost 98% of its value, falling from $4.1bn in December 2021 to $0.10bn in November 2026. By 2023, the first full year after AI image generators reached the public, paid downloads were 15% below their 2021 level and content revenue had fallen 7%. The decline has since accelerated: in Q3 2026, content revenue fell 19%, total revenue fell 33% and paying subscribers declined 16% year on year, as marketers increasingly generate the generic images they once licensed. Licensing its archive to AI developers has not offset the decline. The segment containing its data licensing business grew 256% to $137m in 2023 and reached $203m in 2025, before contracting 47%, 16% and 76% year on year in the first three quarters of 2026, respectively, leaving quarterly revenue at $15.8m.
Freelance marketplaces
Freelance marketplaces have been repriced as sharply as the lead-generation sites, because the work they intermediate is digital and an LLM can now produce much of it directly.

Fiverr has fallen 93%, from $4.2bn in December 2021 to $0.29bn in November 2026, because buyers need fewer of the small digital jobs the platform was built on: a logo, a translation, a paragraph of copy. Annual active buyers fell from 3.3m to 2.4m in a year and marketplace revenue fell 25% in the third quarter of 2026, while the buyers who remain are spending more, with annual spend per buyer up 13% to $373.
Upwork's market cap is down 72%, from $4.4bn in December 2021 to $1.2bn in November 2026; its revenue fell 8% in the third quarter. One part of the business is growing quickly. Spend on freelancers with AI skills rose 55% to an annualised run rate of nearly $450m, revenue from Business Plus, its premium tier for business clients, rose 120%, and spend per active client reached a record $5,418. Lifted, the subsidiary that manages and pays contingent workers on behalf of enterprises, is guided to grow volumes by over 30% in 2026.
Transactional marketplaces are more defensible where fulfilment requires physical supply that AI cannot reproduce, such as a room, a meal or a ride. Freelance marketplaces face a different risk: AI can replace the work being traded, reducing demand for both the freelancer and the platform. Their opportunity is to manage complex projects, combining AI and specialist talent to deliver the finished work and charging for delivery.
Content creation
Presentation software: Pitch, the Berlin-based maker of collaborative presentation software for teams, raised at a reported $600m valuation in 2021; by January 2024 it had laid off two-thirds of its staff and recapitalised. Tome, a start-up that generated slide decks from a text prompt, raised $81m and reached 25m users, yet lost its edge when Microsoft and Google built AI presentation tools directly into PowerPoint and Google Slides, and shut the product down in April 2025 with ARR of only around $3.5m. Nevertheless, a16z-backed Gamma still appears to be growing, citing 100m users in Q4 2026 compared to 70m in Q1.
AI copywriting shows that being AI-native is no protection when the product is a thin layer on someone else's model. Jasper, a tool that writes marketing copy for businesses using third-party language models, raised at a $1.5bn valuation in October 2022, one month before ChatGPT launched. Within a year it had cut its internal valuation by 20% and laid off staff.
The same has happened in AI audio, where Huxe, an app that turned a user's inbox and calendar into a personal audio briefing, raised $4.6m in September 2025 and shut down eight months later, in the week Spotify launched a similar feature.
These businesses charged for an output, a deck, a paragraph of copy, an audio briefing or a logo, that the underlying models could increasingly produce themselves. Once model providers and established platforms bundled those capabilities into products customers already used, the reason to pay for a separate tool weakened. What had attracted funding as a standalone business was becoming a feature in someone else's product.
We are seeing this across other sectors too

Chegg has lost 98% of its value, falling from $4.4bn in December 2021 to $73m in November 2026, because it sold step-by-step homework answers on a monthly subscription, which is precisely what a general-purpose LLM now provides for free. Its answers were written by paid subject experts, a task a model now performs, and the answer itself has become a feature of LLMs like ChatGPT. Revenue has halved, from $776m in 2021 to $377m in 2025, third quarter 2026 revenue fell a further 40% to $46.5m, and the company is now repositioning towards skilling and employability.
Coursera, which sells online courses and professional certificates, is down 73%, from $3.4bn in December 2021 to $0.94bn in November 2026, despite merging with Udemy in May 2026, and management expects consumer revenue to decline at a low-double-digit rate in 2027. Its response is a $100m investment in LearnVector, an AI-native learning company founded by Andrew Ng, and a new AI-native skills platform due in 2027.
Stack Overflow, the question-and-answer site for programmers that Prosus bought for $1.8bn in 2021, has lost 99% of its question volume, falling from 109,119 new questions in November 2022, the month ChatGPT launched, to 965 in September 2026. The site depended on a developer posting a problem in public and waiting for another developer to answer it, and coding assistants such as ChatGPT, GitHub Copilot and Claude now give an immediate answer inside the developer's own editor, so the question is never posted. Stack Overflow's response has been to license its archive of answers to OpenAI and Google for model training.
Where does the value go next?
Across the sectors examined, AI is putting pressure on businesses that charge for information, discovery or tasks that a general-purpose model can increasingly perform. Being AI-native offers no automatic protection either. The early AI boom drew capital into software businesses built around single features, some of which have since been absorbed into general-purpose LLMs with far greater distribution. We draw two conclusions about where value is likely to endure and where new businesses will emerge.

he first is that transactional marketplaces have proved more resilient. We have seen substantial value destruction among businesses paid for discovery, comparison and leads, while Booking, which owns the transaction, has grown in value. Our hypothesis is that as AI agents take on more of discovery and comparison, they will increasingly work with marketplaces that aggregate supply, facilitate transactions and manage fulfilment, rather than replace them. These marketplaces could become important transaction partners for AI agents, even if customers increasingly discover and select products elsewhere. Tripadvisor and Trivago are moving closer to the booking, while Upwork is taking greater responsibility for delivering the work. For businesses built on passing customers to someone else, moving further into the transaction is becoming increasingly important.
The second, and in our view the more significant opportunity over the next decade, is doing the work itself. For much of the past 20 years, internet businesses have been paid to introduce a customer to a service provider, or to sell the software with which the customer performs a task. AI allows platforms to take on more of that work and capture more of the value created. Wix's Base44 illustrates this shift: the customer describes the application they want, and the platform builds it. The same opportunity is emerging in customer service. Businesses have traditionally paid outsourcing providers for human agents or software companies for the tools those agents use. Sierra and Intercom's Fin use AI agents to handle conversations, answer questions and resolve customer issues, with charges linked to successful resolution.
Mercor recruits and vets specialists such as doctors, lawyers and engineers, then organises their work to deliver training datasets and evaluations that clients use to improve AI models. The client receives the completed work, rather than a shortlist of CVs to recruit and manage themselves. Founded in 2023, Mercor reached a $20bn valuation in November 2026, less than four years later.
At EIV, we believe the next wave of valuable marketplace and services businesses will include both transactional platforms that work alongside AI agents and AI-native companies that take responsibility for delivering the outcome themselves. The opportunity extends into accounting, legal services and many other industries where AI can change both how work is delivered and the economics of delivering it.
Next week, we will examine which professional services are most likely to be disrupted first, and the AI-native companies already taking them on.
At EIV, we advise marketplace and services platform businesses on fundraises and M&A. If you are building a business that completes the transaction or delivers the outcome, we would love to hear from you.
EIV data: 10 listed businesses in AI-exposed sectors lost $206bn of market value between December 2021 and November 2026, falling from $305bn to $99bn, while their combined revenue rose 47% from $22.2bn in fiscal 2021 to $32.7bn in fiscal 2025. Booking Holdings is shown for comparison.
Sector | Company | Market cap Dec 2021 | Market cap Nov 2026 | Change | Revenue FY2021 | Revenue FY2025 | Change |
|---|---|---|---|---|---|---|---|
Price comparison and travel | Tripadvisor | $3.8bn | $0.88bn | -77% | $0.90bn | $1.89bn | +110% |
Price comparison and travel | Trivago | $0.78bn | $0.26bn | -67% | $0.41bn | $0.62bn | +52% |
Price comparison and travel | LendingTree | $1.6bn | $0.52bn | -68% | $1.10bn | $1.12bn | +2% |
Creative and content tools | Adobe | $269bn | $92bn | -66% | $15.8bn | $23.8bn | +51% |
Creative and content tools | Wix | $9.0bn | $2.9bn | -68% | $1.27bn | $1.99bn | +57% |
Creative and content tools | Shutterstock | $4.1bn | $0.10bn | -98% | $0.77bn | $0.99bn | +28% |
Freelance marketplaces | Upwork | $4.4bn | $1.2bn | -72% | $0.50bn | $0.79bn | +57% |
Freelance marketplaces | Fiverr | $4.2bn | $0.29bn | -93% | $0.30bn | $0.43bn | +45% |
Learning | Coursera | $3.4bn | $0.94bn | -73% | $0.41bn | $0.76bn | +82% |
Learning | Chegg | $4.4bn | $0.07bn | -98% | $0.78bn | $0.38bn | -51% |
Total (10 companies) | $305bn | $99bn | -$206bn | $22.2bn | $32.7bn | +47% | |
Comparison | Booking Holdings | $98.5bn | $118bn | +20% | $11.0bn | $26.9bn | +146% |
Source: Yahoo Finance, CompaniesMarketCap, company filings, Stockanalysis.
How we built this. We selected 10 listed businesses in sectors most exposed to generative AI: price comparison and travel search, creative and content tools, freelance marketplaces and learning. Market capitalisation is tracked monthly from December 2021, the last full year before generative AI reached consumers, to November 2026, estimated from month-end share prices scaled to year-end market capitalisation. Revenue covers fiscal years 2021 to 2025; Adobe's fiscal year ends in November, and Trivago's euro revenue is converted at €1 = $1.13. Published November 2026.
