AI Is Rebuilding the Shopping Funnel From the Inside Out — And Comparison Platforms Need to Adapt

Three announcements landed in the same week of March 2026. Shopify turned on Agentic Storefronts, putting 5.6 million merchants' products inside ChatGPT, Copilot, and Gemini by default. Meta added AI review summaries and one-tap checkout to Instagram and Facebook ads. And OpenAI killed Sora, its video generation app, to redirect compute toward robotics and enterprise productivity — while simultaneously upgrading ChatGPT's shopping experience.

Each announcement on its own is a product update. Together, they outline how AI is restructuring the way consumers discover, evaluate, and purchase products. I've spent the past year building Mubboo across five countries, and what I'm seeing from the inside is that the traditional shopping funnel — discover on Google, compare on a review site, buy on a retailer — is being compressed into a single AI conversation.

The old funnel had five steps. The new one has two.

The shopping funnel I grew up with looked like this: see an ad → search Google → visit comparison sites → read reviews → buy from a retailer. Each step had its own industry. Google owned discovery. Sites like Wirecutter owned evaluation. Retailers owned the transaction.

What happened this week collapses that into: ask AI → buy. Shopify's Agentic Storefronts mean AI already knows what's available and at what price. Meta's AI review summaries mean evaluation happens in the same interface as discovery. One-tap checkout means the transaction completes without leaving the conversation.

When we started building Mubboo, our assumption was that consumers would always want a dedicated comparison step — a place to see multiple options side by side, with structured data and scenario-based recommendations. That assumption is being tested. Not because comparison is less valuable, but because the friction cost of leaving an AI conversation to visit a separate site is higher than most consumers are willing to pay for slightly better information.

What AI shopping gets right

I'll give credit where it's due. AI-driven product discovery solves a real consumer problem: choice overload. A category like wireless earbuds has hundreds of options. Traditional comparison sites show you 10 to 20, organized by some editorial judgment of "best." AI can narrow the field to 3 based on your actual requirements — budget, use case, brand preference — in seconds.

Shopify's data backs this up. AI-attributed orders on their platform have grown fifteenfold since January 2025. Conversion rates from AI-assisted shopping run at 12.3%, roughly four times the unassisted rate of 3.1% (Envive AI, 2026). Consumers who arrive through AI are higher-intent because the AI has already filtered for relevance before presenting options.

Meta's AI review summaries address another real pain point. Reading 400 reviews of a hair dryer to figure out whether it works for thick hair is something nobody enjoys. An AI that compresses those reviews into "works well for thick hair, multiple users report overheating after 20 minutes of continuous use" saves real time.

What AI shopping gets wrong

The problems are structural, not cosmetic.

First, AI review summaries flatten nuance. A product might carry a 4.5-star average rating and get a glowing AI summary, but bury a pattern of failures for a specific use case in the long tail of reviews. We see this in our own data at Mubboo. When we built comparison pages for robot vacuums in Australia, the models that ranked highest by average rating were not always the best choice for specific floor types. A pet owner on hardwood needs different things than a family with carpet and small children. AI summaries trained on aggregate sentiment miss these distinctions.

Second, product data quality determines visibility in AI shopping, but consumers don't know that. When ChatGPT recommends a product, the consumer assumes it's the best match. In reality, it might be the product with the most structured data in Shopify Catalog. A better product from a brand with poor metadata simply doesn't surface. This is the same problem Google Shopping had in its early years — garbage in, garbage out — but wrapped in the authority of a conversational AI that sounds like it knows what it's talking about.

Third, the economics point toward ads. Shopify and OpenAI both say there are currently no sponsored placements in AI shopping results. EMARKETER analyst Sarah Marzano warned at Shoptalk that "agentic infrastructure is being built faster than the underlying consumer behavior is evolving" and urged brands not to "reallocate meaningful capital ahead of proof." But the platforms generating AI shopping traffic need revenue. Sponsored AI product placements are coming — the only question is when. Once they arrive, the organic visibility window that currently benefits small brands will close, and AI shopping will look a lot more like Google Shopping: pay to play.

What we're doing at Mubboo

We've made three decisions based on what we're seeing.

First, we're doubling down on scenario-based comparison rather than aggregate ranking. AI can tell you the three highest-rated air fryers. We tell you which air fryer works best if you cook for one person in a small kitchen versus a family of five that wants to roast a whole chicken. That level of specificity requires editorial judgment and real-world testing data that AI summaries don't yet capture.

Second, we're optimizing every piece of content for AI citation. If a consumer asks ChatGPT "what's the best travel insurance for Australians going to Bali?" and Mubboo's comparison page appears as a cited source, we're part of the new funnel even without the consumer visiting our site directly. This is the GEO (Generative Engine Optimization) strategy we've been implementing — structured data, self-contained fact-dense paragraphs, question-based headings that match how people query AI.

Third, we're tracking which AI platforms cite us. The overlap between Google's top-ranked pages and AI-cited sources has dropped below 20% (EMARKETER, 2026). A page ranking well on Google is no guarantee of appearing in ChatGPT's product recommendations, and vice versa. We're treating AI visibility as a separate optimization channel from traditional SEO, with its own measurement and its own content format requirements.

The comparison platform that survives will be the one AI trusts

Here is what I keep coming back to. The traditional comparison site competed for Google rankings. The next generation of comparison platforms will compete for AI citation. That means structured data, named authorship, verifiable statistics, and content formats that AI systems can extract and present to users.

AI shopping will grow. EMARKETER projects AI platform-driven e-commerce will reach $144 billion by 2029, representing 8.8% of total retail e-commerce. The comparison platforms that adapt will be cited alongside product listings inside those AI conversations. The ones that don't adapt will wonder where their traffic went.

The window for positioning is right now, while AI shopping rankings are still organic and the protocols (like Shopify's Universal Commerce Protocol and Google's AI Mode) are still being defined. We started building Mubboo for this moment, even before we knew exactly what form it would take.