Web Summit 2026 AI Companies: Which Ones Matter for AI Search and GEO?

Web Summit 2026 AI Companies: Which Ones Matter for AI Search and GEO?

For anyone tracking how brands appear in AI answers, the Web Summit 2026 sessions worth planning around come from three groups: agent builders (Manus, You.com and Recursive, Temporal), model developers (MiniMax, ModelBest) and the compute and data companies underneath them (Google Cloud, IBM, Cerebras, Elastic, Qualcomm). The event runs in Lisbon from 9 to 12 November, and a small cluster of Portuguese AI startups on the ALPHA floor shows what early-stage agent tooling looks like.

Why should a GEO practitioner care about a conference line-up?

Generative engine optimization depends on systems you don't control. A model writes the answer. A retrieval layer chooses which sources feed it. More and more often, an agent acts on that answer before a person reads it.

Web Summit is one of the few places where several of the people building those layers speak in public in the same week.

The scale also tells you something. At the 2025 edition, AI and machine learning was the biggest category among 2,725 exhibiting startups from 108 countries, at nearly one in five. The 2026 exhibition space is reported to have sold out.

A caveat before the list: line-ups change up to the last minute. Treat every name below as announced, not guaranteed, and confirm sessions in the Web Summit app.

Which agent and model companies are speaking?

These names come from the official speakers page, where AI and robotics scaleups, many from China, were among the first announced.

• You.com and Recursive. Richard Socher, founder of the AI search and agent company You.com, is speaking as co-founder and CEO of Recursive and You.com. Of everyone on the list, this is the session closest to AI search itself.
• Manus AI. Manus builds a general-purpose agent that plans and completes multi-step tasks. Co-founder and CPO Tao Zhang is on the programme. Listen for how an agent like this picks and weighs its sources, because that choice decides which brands it ever sees.
• MiniMax. A Chinese developer of large language, video and voice models, represented by Alan Shen, MD Global.
• ModelBest. Known for compact, efficient models. Co-founder and CEO Dahai Li is speaking.
• Temporal. An open-source platform that keeps long-running software workflows reliable, increasingly used for AI agents. Co-founder Samar Abbas has an AI Summit slot titled "Bet on the foundation, not the framework".

The common thread is that answers are turning into actions. A buyer's question may soon be handled by an agent that compares options and completes a task, so the question for GEO teams shifts from "is the brand cited?" to "is it chosen when nobody is reading?"

What do the compute and cloud partners signal?

The 2026 partner page lists the infrastructure side:

• Google Cloud, Google's cloud and AI platform for businesses.
• IBM, listed under AI and machine learning and named the official AI Summit partner. Its venture and AI teams have judged PITCH in the past.
• Cerebras, which makes wafer-scale processors for training and running large AI models.
• Elastic, search, observability and security software that companies increasingly use to ground AI applications in their own data.
• Qualcomm, whose processors power phones, PCs, cars and on-device AI. CEO Cristiano Amon spoke at the 2025 event.

For search people, Elastic is the most direct link. Grounding an AI application in company data is retrieval by another name, and it is the same mechanism that decides which content a public assistant quotes.

The AI Summit track frames the week around questions that matter here: whether organisations are ready for AI agents, whether AI investment pays back, how open and closed models will compete, and what happens when the internet fills with AI-generated content. That last one is a source-quality question, and source quality is what citation depends on.

Where do the humanoid robots fit?

Embodied AI sits further from search and is the first thing to drop if your time is short, but it shows where large models are heading next.

• Unitree makes quadruped and humanoid robots. Co-founder Li Chen is speaking.
• AgiBot develops humanoid robots. Its A3 robot is listed, and William Shi, president for Europe and the Americas, is speaking.
• Galbot builds general-purpose robots powered by large models. Its G1 robot and CSO Yuli Zhao are both listed.

Which Portuguese AI startups deserve a detour?

Portugal has won the last two PITCH competitions, most recently with Granter, a Lisbon startup that uses AI to help companies find and apply for grants. According to Portugal Business News, Portuguese AI startups in this year's ALPHA programme include:

• Arsia Labs: a compliance layer for AI agents, built around rules such as the EU AI Act and GDPR.
• ByChat: AI FinOps, helping teams understand and cut AI and cloud spend.
• Concha AI: a decision-control layer that keeps human and AI decisions aligned with strategy.
• Flowzi: connects company systems so AI can run operational work.
• Ottex: a meeting notetaker and dictation tool for organisations with strict data residency needs.
• Sentinela: a control system for AI-powered customer service.

Read as a group, these are mostly control layers for agents: compliance, cost, alignment and oversight. The early-stage market already assumes agents are in production.

The featured startups page points the same way, with AI and machine learning by far the most common category. One name stands out for content teams: Primitive, which says it makes enterprise email ready for AI.

Startups usually exhibit for one assigned day, so check the app before you walk over.

What does the funding data say?

The Web Summit x Crunchbase report for 2025 found that nearly 200 startups from the 2024 programme raised $715.5 million in the year after the event. AI and machine learning startups took the largest share, $334.6 million, up 360% on the year before.

Where does AI visibility measurement fit in?

Among the exhibitors is MarkGrid, at Booth B2-14 in Pavilion 2 on Thursday 12 November. Its Model Share metric measures how often ChatGPT, Gemini, Perplexity, Claude and Copilot recommend a brand instead of its competitors, and traces each gain or loss to the prompt and the source the model cited. The model sessions explain how answers are built; measurement shows who ends up in them. For the wider list of exhibitors and speakers, see the full guide on MarkGrid, and for the stand itself, its Booth B2-14 briefing.

How should you plan an AI-focused week?

1. Book the talks first. Pick the agent and AI Summit sessions, then fit the floor around them.
2. Filter the app by stage. ALPHA, BETA and GROWTH labels separate early tools from funded scaleups.
3. Confirm exhibit days. Most startups are on the floor for one day only.
4. Prepare your questions. How does the agent choose sources? Which models does it run on? How does it treat AI-generated content?