by MichaelWe’re at HiveMCR today.
This is one of the reasons we show up. Not for the badges or the coffee. We come to sit in rooms with the best minds in search, take notes, argue with each other on the train home, and work out what any of it means for the people we work with. Then we pass it on to you.
Here’s a look in my notebook from the morning session at New Century Hall. Four lines I’m taking back to the desk.
Agentic commerce is the headline.
Every Hive has a theme the speakers keep circling back to. This year it’s agentic commerce. The proof point on the table is Shopify’s Universal Commerce Protocol, which the company opened to every developer on Monday and which sits behind Google AI Mode, Gemini apps and Microsoft Copilot Checkout. Buying, comparing, deciding, all increasingly happening through agents acting on the shopper’s behalf. If you sell anything online, this matters.
The reason UCP, Stripe’s Machine Payments Protocol and Anthropic’s MCP Apps all landed inside a five-month window is mechanical, not vibey. Agents need a discovery primitive, a payment primitive and a UI primitive before they can transact on a shopper’s behalf, and as of this week all three exist as open standards.
Anthropic donated MCP to the Linux Foundation in December 2025 and added MCP Apps for embedded in-conversation checkout in January 2026. Shopify and Google launched UCP under Apache 2.0 on 11 January 2026. Stripe Sessions on 29 April brought Link Agent Wallet, the Machine Payments Protocol co-authored with Tempo, and Shared Payment Tokens with Klarna and Affirm. Visa publicly joined Stripe on agent payments in 2026. By April, the UCP Tech Council had expanded to include Amazon, Meta, Microsoft, Salesforce and Stripe alongside Google, Etsy, Target and Wayfair.
We’re in a new consideration era.
Sophie Brannon from StudioHawk made this point: Consideration isn’t a single moment any more. It’s spread across more platforms, more SERP features, and more touchpoints than ever, and the competition at that stage has gone through the roof.
Datos and SparkToro’s Q4 2025 State of Search report found that 56% of Google desktop searches are now zero-click. AI Overviews have cut organic CTR by 9 to 15% on affected queries. Gartner’s widely-cited number for B2B buyers is around 27 touchpoints in a typical journey, with only 17% of buying time spent meeting suppliers directly. Dreamdata’s 2025 dataset of 3.5 million journeys puts the average B2B deal at 76 touchpoints, 6.8 stakeholders, and 211 days from first touch to close.
Consideration used to live in two or three browser tabs and a phone call to a colleague. In a 56%-zero-click world it lives across Reddit threads, Trustpilot scores, AI Overviews and three different LLMs running silent comparisons before a human has even typed your brand name.
Brand visibility across the whole funnel matters, but the information and consideration stages are now where the work is. Perspectives, ideas, owned content, your own assets saying your own things. The brands showing up across formats and platforms with a clear point of view are the ones being chosen.
Search is now a hybrid.
Dawn Anderson framed today’s search as a hybrid.
– Classic information retrieval
– LLMs sitting on top with RAG (Retrieval-Augmented Generation)
She also pointed out “lost in the middle” as a real problem with AI outputs. The same biases that affect the human mind, primacy and recency, also affect what these systems pay attention to. What sits in the middle of the context gets dropped.
The “Lost in the Middle” paper by Liu, Lin, Hewitt, Paranjape, Bevilacqua, Petroni and Liang (arXiv 2307.03172, published in TACL 2024) documents a U-shaped accuracy curve: language models recover information best when it sits at the very start or very end of the context window, and degrade sharply when the relevant passage is buried in the middle. The paper reports accuracy drops of more than 20 percentage points on multi-document QA when the answer-bearing chunk is in the middle of a long context.
If a RAG-powered answer engine pulls ten chunks of your content into a long context window, the chunks at positions four through seven are statistically the least likely to make it into the response. Structure and chunking are now a ranking factor in everything but name.
The panel afterwards suggested the hybrid era won’t last. The direction of travel is toward more fully AI-driven search, inevitably, with the classic only information retrieval layer gradually absorbed into the generative one. The work for content teams now is to make every piece useful when pulled in isolation by an LLM, not just when read sequentially by a human visitor.
Practically that means front-loading the answer in every section, using clear semantic chunks with self-contained meaning, naming the entity in every paragraph rather than relying on it carrying from the headline, and structuring content with a question-and-answer logic that a retrieval system can lift directly into a generated response. But, yet again, this has been best practise for journalists and content writers for years, in order to optimise for humans.
Contribution vs attribution.
This one came from Nick Handley, Performance Director at Connective3, In a world of AI search and agentic commerce, the question stops being “which channel gets the credit” and starts being “what did we contribute to the decision.” Attribution was always a flawed model. Contribution is closer to how buying actually works.
The contribution framing isn’t a rebrand of attribution, it’s a concession to reality. Once you accept that multi-touch attribution models over-credit digital channels by 30% or more in the majority of large-account audits, you stop arguing about which click gets the gold star and start asking what would have happened if the channel hadn’t run at all.
For the in-house teams reading this, the practical move is to stop reporting MTA as the single source of truth. Layer in MMM (Marketing Mix Modelling) for budget allocation across channels and incrementality testing for the channels you can’t measure causally. Treat each method as answering a different question rather than fighting over the same one.
If you want more content like this delivered daily? Connect with Myself or Sean from our marketing department on LinkedIn, where they share regular insights, practical tips, and industry analysis to help you stay ahead of the curve.
