2026 年 AI 網絡生意趨勢：工具服務增長，答案型網站萎縮
Making Money on the Web with AI in 2026: What's Growing, What's Dying
作者分析指出，AI 正令 2026 年網絡生意呈現分化：提供可直接使用成果的工具服務增長，依賴搜尋流量提供答案的網站則收縮。
If your page ranks first on Google and an AI answer appears above it, you get 58% fewer clicks. That's from Ahrefs. Meanwhile Cursor, a code editor, went from $2B to over $4B in annualized revenue in about four months, according to press reports. Those two numbers sum up what AI did to making money on the web: services that do the work grow, sites that sell answers shrink.
Short version: I wouldn't start an ad-funded content site in 2026, and I wouldn't build a Cursor competitor either. For a solo developer or a small team, the realistic options are a narrow paid tool for one profession, or doing AI implementation work for companies.
This is part 2 of a short series on how people make money with AI right now. Part 1 covered mobile apps; this one covers websites and web services.
What's growing: tools that do the work
The fastest money on the web this year is in tools that write code. Press reports put Cursor above $1B in annualized revenue by late 2025, $2B by February 2026 and over $4B by June. These are leaks, not filings, since the company is private. Even with a wide margin of error, it's one of the fastest ramps in software history.
Next come the builders, where you describe an app or site in plain language. Lovable reached $400M ARR by February 2026, per TechCrunch, and later reports put it near $600M by September. Replit's CEO cited $240M in 2025 revenue, with a stated goal of a $1B run rate by the end of 2026.
What they have in common: you pay for a result you can use, working code or a working site, not an answer to read. It's the same pattern as photo and video apps in part 1. People pay for things they can take away and use at work.
The second growing group is narrow B2B tools in the browser: decks, meeting transcription, product photos, documents. Productivity takes 15 of 50 spots in a16z's top consumer AI products by spend, and many of those live on the web rather than in the App Store. Companies pay for them, so prices are higher and churn is lower.
What's shrinking: sites that answered questions
On the other side are sites that lived on search traffic. The Reuters Institute, using Chartbeat data, found organic Google traffic to 2,500+ publisher sites fell 33% worldwide and 38% in the US between November 2024 and November 2025. Reference content was hit hardest: weather, horoscopes, TV listings, simple how-tos. Chartbeat data reported by Axios shows small publishers lost 60% of search referrals over two years, versus 22% for large ones.
The main driver is the AI answer above the results. Your page keeps the top spot but loses the visitors.
Two examples of what this looks like in money:
- Chegg, a homework-help service, got most of its subscribers from Google searches for study questions. Q4 2024 revenue fell 24%, it hired Goldman Sachs to explore a sale, and in February 2025 it sued Google over AI Overviews.
- Stack Overflow's monthly question volume fell back to 2009 levels, when the site had just launched. The decline started before ChatGPT, but sped up sharply after it shipped in late 2022.
Business models: what grows, what shrinks
My estimate of the entry bar is for one developer or a small team. Figures come from different sources and are measured differently, so don't compare them directly.
| Model | Examples | Trend and numbers | Entry bar |
|---|---|---|---|
| Growing | |||
| Dev tools and builders | Cursor, Lovable, Replit, Bolt | Cursor: $4B+ annualized by June 2026 (press) | Closed: needs model budgets and investors |
| Narrow B2B tool in the browser | Gamma, Photoroom, Otter, Fireflies | Productivity: 15 of 50 in a16z top-50 | Medium: best shot for a solo dev |
| AI implementation for clients | Automation, chatbots, integrations | Demand growing with company interest in AI (my estimate) | Low: needs experience, not capital |
| Flat | |||
| Thin wrapper over someone else's model | "One prompt plus a checkout" | Repeating the fate of mobile wrappers (my estimate) | Low, but ChatGPT and Gemini absorb the feature |
| Storefront for your own product or services | Landing pages, portfolios, app pages | Less from search, more from social, channels, newsletters (my estimate) | Low: you bring the traffic yourself |
| Shrinking | |||
| Ad-funded content site | Reference, how-tos, disposable news | −33% Google traffic worldwide, −60% for small publishers (Reuters Institute, Chartbeat) | Low, but the visitors are leaving |
| Affiliate reviews | "Top 10 vacuums", comparison pages | −58% clicks on position one with an AI answer (Ahrefs) | Low, but the AI answer takes the click |
| Q&A sites | Chegg, Stack Overflow | Chegg: −24% revenue in Q4 2024, lawsuit against Google | Closed |
Why some grow and others shrink
The line comes down to one question: can AI deliver the result right inside the search box or chat window?
If a site's product is an answer, a text or a list, AI hands it out on its own and nobody clicks through. That's what happened to horoscopes, recipes, reference pages, reviews and homework answers. The site wrote a good answer, the search engine summarized it, and the ads on the site stopped getting impressions.
If the product is an action, chat AI doesn't replace it. It makes it cheaper and faster. Code has to run, a deck has to be shown to a client, a meeting has to be recorded and summarized, a business process has to be wired into a CRM. That happens in a separate service, and people pay for it.
The second takeaway: a website stops being a traffic source and becomes a storefront. People arrive from a channel, a newsletter, social media or a referral, not from a random search. You have to bring them yourself.
What could break this
Model costs. Builders like Lovable and Bolt pay for every model call. Users complain about credits burning fast, and margins are thin. If the provider raises prices or ships the same product, a business built on someone else's model takes the hit.
The platform ships it. ChatGPT already makes decks, spreadsheets and simple sites. What protects a narrow tool is depth: the customer's data, integrations, a workflow tuned to one profession.
Search will keep changing. The traffic numbers cover 2024–2025. AI answers keep expanding, and a model that looks "flat" today may be shrinking a year from now.
What I'd pick
I wouldn't start an ad-funded content site now. Even a great page at position one gets fewer and fewer clicks, and better SEO won't fix that. I wouldn't build a Cursor competitor either: that race is between companies with billions in funding.
A solo developer has two realistic paths. One is a narrow paid tool for a specific profession where AI removes the routine: marketplace sellers, accountants, lawyers, real-estate agents. It's the same logic as Cal AI in part 1, just in the browser and for businesses. The other is AI implementation for companies. That's what I do myself: automation and AI-powered services for businesses. Right now there's more demand for that than for "just a website".
You still need a site. Treat it as a storefront you bring people to, not as a machine that collects traffic from Google.
Originally published at klukyanov.ru.
Shorter weekly write-ups (in Russian) — on Telegram.
來源：DEV Community · dev.to