AnswersPublished22 sources
What is former Twitter CEO Parag Agrawal’s argument that the internet’s advertising and click based business model cannot survive if AI agents become its primary users, what is his vision for a “parallel web” built for agents—including replacing human click data with agent feedba
Agrawal’s central claim is that a web funded by human attention breaks when AI agents, rather than people, do most of the reading and searching. Agents do not view ads, develop brand preferences, or produce meaningful clicks; yet they could query the web orders of magnitude more often than humans, removing publisher...
Published byImages generated with GPT Image 1.5
Research answer
Key takeaways
- Agrawal’s central claim is that a web funded by human attention breaks when AI agents, rather than people, do most of the reading and searching.
- Agents do not view ads, develop brand preferences, or produce meaningful clicks; yet they could query the web orders of magnitude more often than humans, removing publishers’ traffic and revenue incentive.
- [2][14] Why click economics fail: Today’s search and publishing loop uses clicks as both an economic unit—ad impressions and referrals—and a ranking signal.
- Agrawal argues that agents often consume many sources silently, synthesize an answer, and may never send a human visitor to the originating page.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What is former Twitter CEO Parag Agrawal’s argument that the internet’s advertising and click based business model cannot survive if AI agen. Article summary: Agrawal’s central claim is that a web funded by human attention breaks when AI agents, rather than people, do most of the reading and searching.. Topic tags: general web, llm, agents, ai, workflow. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers, clickbait thumbnails, icons, and tiny thumbnail layouts. Make
Agrawal’s central claim is that a web funded by human attention breaks when AI agents, rather than people, do most of the reading and searching. Agents do not view ads, develop brand preferences, or produce meaningful clicks; yet they could query the web orders of magnitude more often than humans, removing publishers’ traffic and revenue incentive. 2
14
Why click economics fail: Today’s search-and-publishing loop uses clicks as both an economic unit—ad impressions and referrals—and a ranking signal. Agrawal argues that agents often consume many sources silently, synthesize an answer, and may never send a human visitor to the originating page. Thus, measuring clicks would neither reveal a source’s actual contribution nor pay the publisher for it.
1
The “parallel web”: His proposed alternative is an agent-native layer of web infrastructure, rather than a wholesale replacement for human-facing websites. Search, ranking, retrieval, and payment would be optimized around agents’ task outcomes: did a source help the agent correctly research, reason, complete a workflow, or improve the final answer? Parallel’s APIs are positioned as the retrieval layer for that agentic use—search, extraction, deep-research tasks, entity discovery, chat, and monitoring.
Replacing clicks with agent feedback: Instead of inferring relevance from aggregate human clicks, Parallel wants agents to provide direct signals about whether retrieved material was useful for a particular task. That produces feedback tailored to an agent’s objective and potentially supports ranking for factual utility rather than attention capture. This is still a product thesis, not an established web-wide standard.
13
Paying publishers with Shapley values: The intended compensation mechanism assigns each source a Shapley value: in essence, the incremental contribution that source made to the agent’s successful output, accounting for the other sources available. A source that is indispensable or highly additive would receive more than one that is merely redundant or tangential. The attraction is that compensation is based on marginal task value, not on a click or a flat licensing fee; the hard problem is reliably calculating that contribution at scale and preventing manipulation.
16
What Index does: Parallel’s Index is the publisher-side piece of this model. It gives publishers, data providers, and independent creators visibility into how agents discover, cite, and rely on their content, while offering compensation tied to the value their content contributes. Parallel publicly frames this as an incentive-aligned model for content owners, and its documentation describes Index in those terms.
14
Funding and valuation: Parallel raised a $30 million seed round in 2024, then a $100 million Series A in November 2025 at a reported $740 million valuation.
15
Customers and developer adoption: Public reporting available here does not provide a reliable, comprehensive customer list or a verified developer-count figure. The clearest public evidence is that Parallel sells APIs for agent and developer use and has launch partners for Index among publishers and content distributors, including The Atlantic.
Google Cloud: Parallel has become a search/grounding option for Google Cloud customers building enterprise agents: reported integrations let developers using Google Cloud’s agent tooling choose Parallel Search alongside Google Search for web grounding.
12 Strategically, that gives Parallel distribution inside a major enterprise-agent platform, while Google remains both a powerful partner and a potential competitor.
Why he puts the shift at 12–24 months: Agrawal’s timeline is a forecast, not a measured consensus. His reasoning is that capable models, tool use, and agent frameworks are improving quickly; agents can launch many searches and sub-tasks per human request; and enterprises can adopt those agents through existing cloud and API channels. If that combination crosses a usability and cost threshold, web-query volume could become agent-dominated well before consumer web habits fully change. The premise is captured in Parallel’s stated belief that agents may use the web “1000x more” than humans.
The key caveat is that Index and Shapley-based payment describe a proposed market design. It has not yet been demonstrated as a universal replacement for advertising, nor is there sufficient evidence that publishers, competing agent platforms, and the broader web will adopt one shared attribution system.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "What is former Twitter CEO Parag Agrawal’s argument that the internet’s advertising and click based business model cannot survive if AI agents become its primary users, what is his vision for a “parallel web” built for agents—including replacing human click data with agent feedba"?
Agrawal’s central claim is that a web funded by human attention breaks when AI agents, rather than people, do most of the reading and searching.
What are the key points to validate first?
Agrawal’s central claim is that a web funded by human attention breaks when AI agents, rather than people, do most of the reading and searching. Agents do not view ads, develop brand preferences, or produce meaningful clicks; yet they could query the web orders of magnitude more often than humans, removing publishers’ traffic and revenue incentive.
What should I do next in practice?
[2][14] Why click economics fail: Today’s search and publishing loop uses clicks as both an economic unit—ad impressions and referrals—and a ranking signal.