Consumer AI Can Free Brands From the Middlemen. Or Lock Them In for Good.

A guest opinion from Kleio’s CEO on whether AI shopping assistants will free brands from aggregators or create new gatekeepers.

By Philippe Wellens, co-founder and CEO of Kleio

With the emergence of agentic AI, for the first time in two decades, brands can reclaim direct control over how they interact with their customers online. But they have 6 to 9 months at best to get it right.

AI assistants from Meta’s Muse to OpenAI’s ChatGPT and independent agents like Instinct are rapidly becoming the primary interface where consumers research, compare, and prepare to make purchases. For brands selling complex, high-consideration products, this shift offers a rare window to bypass the aggregators that have stood between them and their customers for twenty years.

The window is open today, but the clock is already ticking. Within 6 to 9 months, the major booking platforms, listing portals and marketplaces will be wired into every leading assistant, capturing more traffic and more leverage over brands. Every month a brand waits is a month of traffic lost, with no quick way back in.

Three Weeks That Changed Everything

On September 8, 2026, Meta launched Muse, an autonomous AI assistant capable of searching, comparing, and purchasing directly on a user’s behalf. Within three weeks, Muse surged to the No. 1 spot on the US App Store. Third-party estimates underscore the speed of adoption: Sensor Tower estimates 3.4 million downloads, with Apptopia placing it as high as 4.3 million. More tellingly, Muse surpassed 642,000 daily active users in its initial weeks – nearly triple ChatGPT’s daily active user base (231,000) at the same post-launch stage.

Concurrently, AI shopping assistant Instinct claimed over $1 billion in annual transaction volume – half of it in travel. And on September 29, OpenAI answered with Dots, its own always-on consumer agents.

Tech incumbents responded immediately: Amazon blocked Muse from shopping on its site, protecting a $68 billion advertising business that only works when a human scrolls through sponsored listings.

In Europe, the Shift Is Already Here

Across Europe, AI-written answers have sat at the top of Google for more than a year: in the UK since August 2024, in Germany, Spain, Italy, Belgium and others since March 2025. France, held back by its publishers’ rights law, was the last major market to switch on, on July 22, 2026. That gave a rare before-and-after: within nine days, the most exposed French sites had lost nearly a quarter of their Google clicks, and one in five had lost more than 30%.

This matches broader global consumer shifts. According to IBM data, 45% of consumers now use AI at some stage of their shopping journey, while Adobe reported that AI-driven retail referral traffic grew 393% year-over-year in Q1 2026. McKinsey projects the broader agentic commerce market will reach $3 to $5 trillion by 2030.

As AI assistants move from recommending to buying on our behalf, brands have only a short window before consumer habits harden around new default intermediaries.

The Challenge of the “Undecided” Buyer

For simple, transactional purchases, AI discovery is largely solved. When a customer knows precisely what they want – “This T-shirt, size M” – platforms handle it effortlessly. Shopify now connects its merchant catalogues directly to ChatGPT and Muse with minimal friction.

However, high-consideration, complex purchases such as “a sunny getaway in February for a family of four under €3,000,” or “a reliable used SUV for €20,000, big enough for a large dog,” remain a challenge:

The right product exists, but it sits buried inside complex catalogues, spec sheets, and real-time inventory databases. The challenge isn’t keyword matching; it’s understanding intent and guiding an undecided buyer toward the right decision – the precise role traditionally played by an expert salesperson.

Big Tech historically struggled with this level of domain specificity. Google attempted real estate and abandoned it; in travel, it focused primarily on standard hotel inventories. Third-party aggregators – booking platforms, automotive classifieds, and real estate portals – filled this gap. They did the complex catalogue work, built the customer interface, and charged brands handsomely for it through high commissions, listing fees, and lost direct customer relationships.

The Intermediaries Are Already at the Table. Now Brands Can Take a Seat.

Consumer AI is reshaping this dynamic. With AI assistants, the purchasing decision happens directly inside the conversation, before the consumer ever visits a merchant website or aggregator portal.

A brand present inside that initial conversation no longer requires a third-party aggregator. However, major OTAs and listing platforms recognised this early: Booking.com and Expedia have been in ChatGPT for over a year, and Expedia was live on Muse from day one. If brands fail to adapt their infrastructure, they will simply find the same legacy intermediaries taking tollbooth positions in conversational AI.

Trust is building fast among early adopters: 40% of Instinct users hand the agent their credit card within three weeks, according to its founder. Most consumers are not there yet: only 14% of US consumers would let an AI agent complete a purchase end-to-end without oversight.

The industry’s answer is arriving. On October 6, Meta and partners including Walmart, Stripe and Shopify unveiled the Personal Agent Protocol, an open standard that lets a consumer’s agent talk directly to a brand’s own agent. The brand decides how it is represented and what the agent can do, and the customer logs in with the brand, not with the assistant.

For the first time, brands can join the conversation on their own terms instead of being summarised, ranked or resold by an intermediary. The question is no longer whether consumers will buy through assistants, but whether brands will be there to answer when those assistants call.

Three Paths Forward for Brands

To navigate this shift, enterprise brands face three strategic choices:

• Doing nothing: Relying on legacy SEO and standard performance marketing leads to declining organic search traffic, escalating customer acquisition costs, and increasing reliance on third-party aggregators.
• Doing everything themselves: Building proprietary AI integration layers requires significant capital, dedicated engineering resources, and continuous maintenance as underlying LLM models and platform protocols update constantly.
• Deploying Specialised Infrastructure: Integrating enterprise-grade AI infrastructure enables brands to structure their catalogue data, maintain persistent customer memory, and connect directly to emerging AI assistant networks in weeks rather than years.

What Brands Need to Be Ready

To sell directly through consumer AI without losing control of the customer relationship, brands with complex product offerings require three core capabilities:

• Intent & memory retention: The ability to translate natural language requests into precise product criteria while maintaining customer context across sessions. Crucially, this first-party customer data and context must remain owned by the brand, not handed over to the third-party AI assistant.
• Real-time catalogue orchestration: Dynamic inventory parsing that queries complex catalogues (pricing, availability, localised options) in real-time, surfacing exact matches rather than generic search lists.
• Multi-platform interoperability: A unified API integration layer that connects brand data seamlessly across evolving assistant ecosystems – whether ChatGPT, Meta Muse, or future domain-specific agents, and emerging open standards such as the Personal Agent Protocol – without requiring custom re-engineering for every platform update.

The arrival of consumer AI represents the most significant shift in digital commerce since the rise of search engines. Every month of waiting is traffic handed to the intermediaries, and assistants learn fast where their users like to buy. Getting in now, brands set their own terms. Getting in later, they will pay someone else’s.