Invisible but inevitable: The shadow internet will rewrite marketing
Opinion
In the shadow internet, brands without endpoints don’t exist, warns UM London’s planning partner.
For three decades, marketers have fixated on the same web of pages, ads, and clicks. However, beneath the surface, something more profound is unfolding. A shadow internet is emerging that isn’t the dark web or criminal. It’s lighter, faster, and crafted not for humans but for machines.
This new layer isn’t about search engines or social feeds. It’s about AI agents making decisions on our behalf: booking restaurants, purchasing tickets, ordering groceries, and resolving queries. Here, the rules change altogether. Endpoints replace homepages. Data feeds and APIs replace banner ads. Trust, speed, and clarity determine who thrives.
The shadow internet will be an additional layer of machine-to-machine communication. Instead of a human browsing a site, an AI assistant queries a clean endpoint, such as an API, a structured feed, or a simple llms.txt file that directs to authoritative data.
It looks nothing like the web we recognise. Just structured information delivered swiftly. Think of it as the service corridors behind a shopping centre, hidden from view, but where the real action takes place. Delivery trucks don’t queue at the main entrance, and AI agents don’t need to trawl through pages. They just want the data.
The signs are already here. Why would an agent navigate a cumbersome restaurant website with pop-ups and PDF menus when it could simply use a booking API? Why would it click through CAPTCHA and upsells when it can access schedules and fares directly from a feed? Instead of searching through FAQ pages, an agent can ping a returns endpoint and get the answer instantly. Even Glastonbury tickets could be transacted immediately via a verified ticketing API.
For marketers, the implications are profound. In the shadow internet, visibility involves structuring content so AI agents can extract it directly, exposing product information and policies in machine-friendly formats, and demonstrating provenance so agents consider you credible. This isn’t SEO 2.0. It’s a complete shift from attention economics, where the goal was to buy impressions, to access economics, where being the trusted endpoint that an agent calls is key.
Which agents should you trust?
Endpoints not only reshape access but also transform the way money is handled. The most impartial information owners will charge AI platforms for structured access to their data. Affiliate 2.0 will emerge, where commerce occurs via transactional APIs with commissions built in. Verification services will establish new layers of trust, where brands pay to be recognised as verified sources.
Allowing machines in carries risks. For decades, brands have spent millions blocking bots. But in the next decade, survival depends on permitting the right ones in. Which agents do you trust? What happens when different agents filter differently, and consumers see varying versions of the internet? What about undisclosed advertising? If an agent recommends a product, was it truly the best choice, or a sponsored endpoint? The open web was built for humans. The shadow internet needs regulations for machines.
There is also a behavioural aspect. Unlike humans, AI agents won’t be fooled by pop-ups or emotional framing. However, they can still be influenced through structured defaults and how options are presented. If a feed highlights one option as the default or includes metadata about scarcity or sustainability, agents may be more inclined to choose it. Behavioural economics doesn’t vanish in the shadow internet; it adapts. An airline could set its booking endpoint so that the lowest-emission fare or a flexible ticket is the default. Subtle differences become a competitive edge.
The environmental dimension matters too. The average web page is over 6MB in size, packed with trackers and scripts. AI agents often retry heavy sites before timing out. That’s wasted bandwidth, compute, and carbon. The shadow internet, by contrast, is designed to be lighter. Endpoints deliver only what’s needed. In a world where AI inference already carries a heavy energy cost, the shadow internet could be the sustainable internet.
How to prepare
Marketers cannot afford to sit this out. To prepare, brands must expose agent-ready endpoints for products, bookings, policies, and FAQs. They must implement structured data such as Schema.org markup, JSON LD, and llms.txt files. They need to prioritise trust by providing consistent facts and digital signatures. They should also offer more advanced functionality to trusted agents, allowing them to transact on behalf of consumers.
Additionally, they must partner with platforms to negotiate whitelisting into OpenAI, Google, Anthropic, and Perplexity ecosystems. Your website remains for people, but your endpoints are now for machines.
The shadow internet is invisible and poised to have just as significant an impact as the regular internet we all use daily. It’s where decisions will increasingly be made before a human ever sees a page.
Two consumers asking the same question could already be seeing two answers. Their agents will fetch data from different endpoints, filter data from various sources, and deliver different results. The idea of a single shared internet is disappearing.
The future of marketing isn’t another platform or algorithm tweak. It’s a hidden layer where AI agents decide what your customers see, buy, and trust. If SEO made your brand searchable, endpoints will unlock a new environment.
Brands that stand against blandness in this new era will thrive.
Lawrence Dodds is planning partner at UM London.