The Platform Ate the Funnel

Google's AI shopping tools, conversational ad formats, and agentic commerce features are pulling discovery, comparison, and checkout inside the platform itself. The marketing funnel that once ran through your website is being absorbed. Brands that survive this shift will be the ones whose identity travels clearly through machine-mediated surfaces.

This was the week Google made the funnel feel obsolete. Between I/O and Marketing Live, the announcements point in one direction: the platform wants to handle more of the journey, from product discovery to cart to checkout, without the customer ever leaving. For marketers, the question is no longer how to drive traffic. It is whether your brand can hold its shape when the platform does the talking.

The funnel moves indoors

Google's Universal Cart is rolling out across Search and Gemini in the US this summer, with YouTube and Gmail to follow.¹² The Shopping Graph now holds more than 60 billion product listings.³ At the same time, AI agents are being embedded directly into Search, capable of comparing products, evaluating trust signals, and helping shoppers complete purchases without visiting a retailer's website. That is not a feature update. It is an architectural change in where buying happens, and it touches every business that depends on search-driven discovery.

For years, the marketing funnel ran through your website. Search delivered the visitor. The landing page did the persuading. The product page closed the sale. Analytics told the team what happened. Marketers optimised each step because they controlled each step. AI-mediated commerce bends that sequence entirely. If a shopper can compare, choose, add to cart, and check out across Google surfaces, the website becomes less like a shop floor and more like a structured data source feeding the agent. Business Insider reported on the growing AI fight inside the shopping cart, with Amazon, Walmart, Google, Meta, Target, Sephora, and Ulta all pushing AI-assisted shopping experiences.⁴ Vogue framed Google's Universal Cart as a fundamental change in how people will shop across Search, YouTube, Gmail, and Gemini.

The implication for marketers is uncomfortable. Product data, availability, pricing, delivery terms, return policies, and review quality are no longer supporting details on a product page. They are the primary interface through which an AI agent evaluates your business. If the agent cannot read them, the business does not make the shortlist. PYMNTS argued that merchants have prepared for many kinds of customers, but not the one arriving now: AI agents that query, evaluate, and transact in seconds.⁵ For small businesses especially, this means less glamorous work than campaign creative: updating location data, cleaning product pages, collecting reviews, improving image descriptions, and making policies machine-readable. In an AI-mediated buying journey, messy information stops being a minor annoyance and becomes a marketing liability.

Ads that answer questions

Google is testing Gemini-built ad formats inside AI Mode and Search: Conversational Discovery ads, Highlighted Answers, AI-powered Shopping ads, and Business Agent for Leads.⁶ In the old model, you wrote the ad, chose the landing page, controlled the message, and tried to earn the click. In the new model, Gemini may generate a custom explainer for why a product fits the query. A user may ask questions inside the ad itself. A brand agent may answer from your website before a human ever sees the lead. Google says these formats will be labelled as sponsored, and advertisers can provide guardrails through tools such as AI Brief.⁶ But marketers still need to get comfortable with a reality that would have been unthinkable two years ago: your ad may speak in real time, and you may not have written what it says.

This is where weak content becomes expensive. If your website is vague, your product data is thin, your claims are unclear, and your tone is generic, the AI has very little to work with. It will either flatten your brand or expose the weakness that was already there. The old ad asked for attention. The new ad may be asked to explain itself, and that is a fundamentally different standard. It rewards brands that have done the harder work of making their positioning clear, specific, and defensible. Search Engine Land's coverage of Google Marketing Live confirmed the breadth of this shift: conversational ads, richer product guidance, Direct Offers, and more native routes to checkout are all in testing or actively rolling out.⁷

MarTech's analysis of agentic commerce made the sharpest version of this argument: consumers may choose emotionally, but AI agents will evaluate pricing, service, delivery, and loyalty value with structured precision.⁸ A human buyer can be persuaded by mood, memory, and timing. An agent buying on behalf of a person will still need signals, but those signals will be more structured and less forgiving. Can the product data be read? Is the price clear? Do reviews match the promise? That means brand work is not disappearing. It is becoming testable in ways that make vague positioning a measurable liability rather than a tolerable weakness.

The steering problem

Google Marketing Live also brought a flood of operational AI updates: Asset Studio for image, video, and text generation from a brief; Demand Gen changes for YouTube and Google surfaces; Meridian coming into Analytics 360; journey-aware bidding; and demand-led budget pacing.⁷ Microsoft Advertising rolled out a new Import Center, cross-account portfolio bidding, and more flexible reporting metrics in the same week.⁹ The common thread across both platforms is not "make more ads". It is "let AI do more, while giving humans fewer but sharper controls".

Asset Studio can generate creative themes from brand guidelines and goals. Journey-aware bidding looks beyond the first form fill. Budget pacing adapts to demand rather than forcing teams to push and pull spend manually every day. This sounds like automation, and it partly is. But the real skill it demands is steering. Marketers will need to define the brief, protect the brand, understand what the system is optimising for, challenge the measurement model, and decide when more reach is not better reach. That is not less work. It is different work, and it requires a kind of judgement that no platform can supply on the marketer's behalf.

The danger is that teams treat AI controls as comfort buttons. They set a goal, accept the recommendation, and assume the machine understands the business. It does not. It understands signals. The marketer's job is to decide whether those signals are pointing towards the right commercial outcome, or merely the easiest measurable one. Marketing has never had a shortage of numbers. It has had a shortage of clear interpretation. AI may finally reduce the reporting grunt work, but the marketer's job gets harder in the place that matters most: deciding which question deserves an answer, and whether the fluent response the system returns actually reflects reality.

How do brands keep their voice consistent when using AI for Instagram?

This is the week's quieter question, but it may be the one that matters most for marketers running AI content marketing day to day. MarTech reported that the real risk of AI in marketing is commoditisation: if every brand uses the same tools, the same prompts, the same templates, and the same "best practice" tone, the output collapses towards the same middle.¹⁰ Research from Canva's 2026 marketing report and separate MarTech coverage found that consumers accept AI in advertising when it adds value, but reject content that feels generic, intrusive, or emotionally hollow.¹¹ The issue is not whether AI touched the work. Most people will not notice or care if AI helped resize assets, test copy, or draft options. They care when the final thing feels empty.

AI can help a brand show up more consistently on Instagram. But if it strips away the messy, local, specific cues that make a business believable, it creates reach without trust. For small businesses, this is especially important. A local café, salon, boutique, or family restaurant cannot afford to sound like every other business in the feed. Their edge is specificity: the owner's eye, the staff, the regulars, the products, the place. Instagram content strategy for these businesses has to start from what makes them recognisable, not from what a prompt can generate from scratch. This is the problem that tools like Asteris are designed to address: Instagram AI content management that begins from the business's own photos, voice, and identity rather than filling in blanks with generic output.

Forbes reported that 43% of small businesses are now using AI for marketing, but customer trust remains weaker when the content feels disconnected from real experience.¹² That gap between business adoption and customer trust is where a lot of AI content quietly fails. The business sees speed. The customer sees sameness. The brands that close that gap will be the ones that use AI to become more consistently themselves, not the ones that publish faster than they learn. Fashion group LPP demonstrated the production side of this equation, reporting that AI now generates 80% of its marketing visuals and has cut content costs by 60%. The savings are real. But when anyone can produce polished content at that speed, the differentiator is no longer production. It is whether the content reflects something true about the brand or merely reflects the default output of a tool everyone shares.

The trust floor

The FTC's settlement with Cox Media over claims that it could use AI to target ads based on conversations near smart devices is a blunt reminder that AI trust has hard limits. At the same time, EU complaints against Google, Meta, and TikTok over scam ads reinforce the same point from a different angle. Platforms and advertisers cannot keep claiming AI improves targeting while harmful content still reaches people at scale. The new marketing edge is not "our AI knows everything about you". It is "our AI helps us serve you better without crossing lines we have no right to cross".

This matters for small businesses too. Trust is hard to earn and expensive to repair. A local salon, retailer, or independent maker cannot afford to sound careless with customer data or to have their brand represented by systems making claims they never approved. Spotify and Universal Music's move towards AI remixes built around opt-in artist participation, credit, and compensation shows what responsible AI content production can look like: powerful tools paired with clear consent and fair attribution. The consent, trust, and quality questions do not disappear because production costs fell. If anything, cheaper production makes those questions louder, because the market fills with more content and consumers become better at detecting the difference between something that was made with care and something that was generated to fill a slot.

The legibility test

The thread running through this week is not that AI is changing marketing. Everyone already knows that. The thread is that the surface area where brands get to explain themselves is shrinking. The platform absorbs the funnel. The ad becomes a conversation. The agent evaluates the product before the customer sees the page. The dashboard hands back a fluent answer before anyone checks whether the question was right. At every layer, the space where human judgement once sat is getting thinner.

That does not mean human judgement matters less. It means the places where it matters are fewer and higher-stakes. The marketer who used to spend hours assembling reports may now spend minutes reviewing AI-generated analysis, but the decision about what to do with it still requires understanding the business, the customer, and the competitive context that no platform can fully encode. LinkedIn's move to reduce the reach of AI-generated content while simultaneously becoming one of the most-cited domains in AI chatbot answers captures the paradox neatly. The platforms are both the amplifier and the filter. Being visible through them requires being genuinely worth citing, not simply present.

For marketers at every scale, the practical test is becoming clearer each week. Can your brand be understood by a machine and still be desired by a human? Can your product data survive an agent's evaluation? Can your ad survive being generated in real time? Can your content survive being summarised? The brands that pass this test will not be the ones that produced the most content or spent the most on ads. They will be the ones with the clearest identity and the cleanest information, because in a world where AI mediates more of the journey, those two things are no longer separate problems. They are the same problem, viewed from different ends of the same shrinking funnel.

Sources