What MWC 2026 actually cared about.

MWC 2026 cared about agent-readiness, even though almost nobody on the main stage said those words. The keynotes sold the same things they sell every year: faster connectivity, bigger models, a device with a new fold. But the operators I talked to in the hallways were worried about something quieter and more practical. Their customers are starting to find them through an answer, not a search result, and most of them have no idea whether the answer is right. That gap is the real story of this year.

This was my second year speaking at MWC in Barcelona, both times about Answer Engine Optimization. Last year AEO felt like a term I had to define before I could use it. This year I didn't. People already knew the problem in their bones, they just hadn't named it. That shift, from explaining the idea to debating the tactics, told me more than any panel did.

The keynote and the hallway were talking about different companies

On stage, the framing was abundance. More compute, more capability, more autonomy coming soon. In the hallway, the framing was exposure. A CMO from a mid-size retailer asked me, plainly, what happens when a buyer asks ChatGPT for the best option in her category and her brand never comes up. She wasn't asking about model capabilities. She was asking whether she still exists to the customer at the moment of decision.

That is the divide I kept hitting. The platform companies are building the future of the demo. The operators are trying to survive the present of the funnel. Both are real, but only one of them has a budget line and a deadline attached, and it isn't the one with the brightest lighting rig.

The shelf is the answer now

Here is the change I built the talk around. For twenty years the shelf was the search results page. You optimized to rank on it, you bought ads next to it, you measured your position on it. That shelf is being replaced. The new shelf is the answer that ChatGPT, Claude, Perplexity, and Google's AI surfaces hand the user directly. There is no page of ten blue links to climb. There is a paragraph, and either you are in it or you are not.

This frightens people for a good reason. The old shelf was visible. You could see your rank and watch it move. The new shelf is opaque. You ask the question, you read the answer, and you have no scoreboard telling you why you were left out. Operators feel this loss of visibility before they can articulate it. A lot of the anxiety in the room was really anxiety about not being able to measure the thing they suddenly depend on.

You can't optimize for a shelf you can't see. So the first job isn't ranking, it's making yourself legible to the machine reading the room.

That is the practical core of what we do at Almost a Lab. Make a company readable to the agents that now stand between it and its customer. Clean, structured, verifiable content the model can quote without guessing. It is unglamorous work. It is also the work that decides whether you appear in the answer or watch a competitor get quoted instead.

Agent-native commerce was the undercurrent nobody put on a slide

The phrase I heard least on stage and thought about most was agent-native commerce. Buyers are starting to delegate the search, the comparison, and soon the purchase to an agent acting on their behalf. When the buyer is a model, your beautiful homepage is irrelevant. What matters is whether your product, your pricing, and your terms are something an agent can read, trust, and act on without a human in the loop.

Very few of the companies I spoke to were ready for that. They had spent a decade optimizing for human eyes and human patience. Now the first reader is a machine with neither. The teams that clicked with this weren't the biggest ones. They were the ones already shipping AI products inside their own agencies, because they had felt the difference between a model that can use your data and a model that gives up on it.

What I'm telling people to do about it

My advice this year was less abstract than last year, because the problem is closer. Don't wait for a scoreboard the platforms have no incentive to give you. Start by asking the engines your own buyer's questions and reading the answers honestly. You will learn fast whether you exist in them. Then fix the readability problem before you fix anything else.

• Ask ChatGPT, Claude, and Perplexity the questions your customers actually ask, then read whether you show up at all.
• Treat your structured content as a product the agent consumes, not as copy a human skims.
• Stop measuring search rank as the only scoreboard. It is describing a shelf that is shrinking.
• Assume the first visitor to your site is a model, and make sure it can read you without guessing.

None of this is what the keynotes were about. That's exactly why I think it's the part of MWC that mattered. The hype tells you what vendors want to sell. The hallway tells you what operators are afraid of. This year those two stories drifted further apart than I've seen, and the quieter one is the one I'd bet on.