Just in: What Google Killing Assistant for Gemini Means for Brands Selling Through AI Agents

What does Google retiring Assistant actually change for brands?

Google began shutting down Google Assistant on Android and Wear OS on 4 September, the first day of a rollout that reaches every device over the coming weeks. Gemini becomes the default AI layer on phones, tablets, watches, earbuds and Android Auto, with no path back to Assistant once a device migrates. This is not a feature update, it is Google forcing the largest single handover of consumer AI attention the platform has seen, and it lands the same week brands are being told, again, that their product data is not ready for it (source: 9to5Google).

For brands, the practical effect is that a decade of Search optimisation now has a second, less understood surface sitting on top of it: Gemini's conversational answers, and increasingly its shopping and agent actions, reaching an enormous captive audience by default rather than by opt-in. The most plausible near-term path is that Google folds commerce features into Gemini progressively over the next several months rather than flipping a single switch, which means there is a genuine window right now to test how a brand actually shows up in a Gemini answer before that behaviour settles into habit.

Why can most merchants not see AI agents shopping on their own sites?

PYMNTS Intelligence, commissioned by Visa Acceptance Solutions, published its Global Digital Shopping Index: Merchant Edition on 2 September, surveying more than a thousand retailers across the US, Brazil and the UAE. The findings are stark: only 23 per cent of merchants can clearly identify both AI-driven traffic and AI-driven purchases today, just 24 per cent can trace a sale back to an AI platform, and a mere 15 per cent have product data an agent can actually read (source: PYMNTS).

This is the gap that should worry every c-suite tracking agentic commerce more than any adoption forecast: the measurement stack has not caught up with the behaviour it is meant to be measuring. Budget is already being allocated to agent-readiness on instinct rather than evidence, and the brands that close their own visibility gap first are working from a genuine information advantage, not just a marketing one. The fix starts with an unglamorous audit: whether your product feeds and structured data would actually resolve cleanly if an agent queried them today.

Who is legally responsible when an AI agent shops on your behalf?

On 4 August, the Ninth Circuit vacated the injunction that had blocked Perplexity's Comet browser from shopping on Amazon for users, ruling that when a person directs an agent to complete a task on a website, it is the person, not the AI company, accessing that site under US computer fraud law. Legal teams across the industry have spent the weeks since working through what it means for every platform currently relying on its terms of service to keep unwanted agents out (source: Cooley).

If that reading holds as more courts weigh in, and the likely trajectory looks like convergence rather than divergence, platforms lose a key legal lever for blocking agent traffic and the contest shifts toward technical controls, contract terms and rate limiting instead. Retailers currently counting on terms of service as their main line of defence should have legal reviewing that posture now, and product teams should treat building a genuinely good branded agent experience as the more durable answer, since customers have less reason to route through a third-party agent if yours is better.

What does it mean that AI agents can now hold a wallet?

Cloudflare began rolling out Cloudflare Wallets this week, giving AI agents a stablecoin balance and a human-readable handle so they can pay for APIs, content and services without a person approving each transaction, built on the x402 protocol it developed with Coinbase and now under Linux Foundation stewardship. More than twenty companies are already routing agent-initiated payments through it (source: The Block).

This is the plumbing layer of agentic commerce, and plumbing tends to be where the durable power sits. Once agents can autonomously pay for data and inventory, the economics of media, content and retail all start to shift toward whoever controls the payment rail and the identity layer beneath it, and a genuinely new question opens up: whether your own content or catalogue should be payable by a machine, not just a human.

So what should a c-suite actually do with all this?

None of these four stories is really about technology for its own sake. They are about who sees the customer, who controls the transaction, and who is liable when something goes wrong, and all three questions just moved. The practical takeaway is narrower than the headlines suggest: get your product and content data into a state a model can parse cleanly, get legal reviewing your bot and terms-of-service posture against the Ninth Circuit's reasoning, and treat Gemini and agent visibility as a distinct workstream from classic SEO rather than an extension of it. The brands doing that this quarter are working from evidence. Most others are still working from instinct, and that gap is closing faster than the trade press coverage suggests.