• AI visibility

Guide · AI search

Will AI assistants put our robots on a manufacturer’s shortlist before an integrator is called?

They can, when a buyer’s question matches an application you document and outside sources confirm it. Robot demand is spreading to industries with little automation history, and those first-time buyers research on their own before they call a robot maker or an integrator.

By Underneath Team Guide Updated

This guide is for executives at robot makers: industrial arms, collaborative robots (cobots), autonomous mobile robots and the first commercial humanoids. It treats system integrators as a sales channel. Integrators themselves, and warehouse automation systems, have their own articles in this series.

The short version

  1. The market is growing again: the International Federation of Robotics (IFR) counted 603,307 industrial robots installed worldwide in 2025, up 11%, and forecasts 655,000 in 2026. The US installed 38,400, up 12%, and is now the second-largest market after China.
  2. New industries are buying: in the first half of 2026, North American robot orders from automotive manufacturers fell 25% while semiconductor and electronics orders rose 35% and life sciences 32%, according to the Association for Advancing Automation (A3) (opens in a new tab).
  3. Cobots are a large share of new buyers’ orders: cobots made up 15.4% of North American robot units ordered in the first half of 2026, and 43.7% in life sciences.
  4. Mobile and service robots are growing fastest: the IFR reported professional service robot sales up 24% to almost 250,000 units in 2025, with transport and logistics robots at 117,500.
  5. Reliability decides selection: in PMMI’s 2026 robotics study (opens in a new tab), 72% of packaging and processing end users already used robots, and 69% rated reliability and uptime as very important when choosing a supplier.

Who buys robots, and what is a customer worth?

Engineers and operations leaders at manufacturers, usually buying through an integrator, with fleet orders following a first success.

A robot purchase usually starts with an operations problem: a labor gap, a quality issue, a safety risk or a throughput target. A manufacturing or automation engineer scopes the application, an operations or plant leader owns the business case, and finance approves the capital spend. In many projects a system integrator designs the cell and installs it, so the integrator often decides which robot brand goes in. Other capital equipment follows a similar path; see how machinery makers get shortlisted.

The size of the market is public. The IFR put the global market value of industrial robot installations at US$18.6bn in 2025. In North America, A3 counted orders for 17,995 robots worth $1.166 billion in the first half of 2026, 6.6% more by value than a year earlier.

What one customer is worth is not public, and it varies widely: one cobot for machine tending, or hundreds of robots across plants. The pattern that matters is expansion. Our inference: a robot maker’s best customers are plants that succeed with a first cell and then repeat it, so the first shortlist decision carries the value of later fleet orders and service contracts. The IFR adds that robot-as-a-service (RaaS), where customers pay over time instead of buying outright, is already the dominant model for professional service robots in the US. Robots sold into distribution centers meet a different buyer, covered in how warehouse automation vendors make shortlists.

Competition is heavy. Teradyne, which owns Universal Robots and the mobile robot maker MiR, reported robotics revenue of $100 million in the second quarter of 2026, up 33% from $75 million, according to The Robot Report (opens in a new tab). But that unit’s annual revenue had fallen from a peak of $326 million in 2022 to $293 million in 2024. Growth is available, not guaranteed.

Where does AI search already sit in robot buying?

In early research, by engineers who use AI but verify it with suppliers, peers and video.

The best evidence on technical buyers comes from the 2026 State of Marketing to Engineers (opens in a new tab) survey by TREW Marketing and GlobalSpec, with Elektor. It found that 69% of technical buyers use generative AI during the purchasing process, but they rate their trust in its answers at 4.7 out of 10. On average, 62% of the buying process happens online before they talk to a vendor.

Across all business purchases, a Gartner survey of 645 buyers (opens in a new tab) found that 45% used generative AI in a recent purchase, mainly to gather information on vendors and products, and that 69% prefer to validate AI-generated insights with sales reps.

Robot buying adds two habits. First, buyers want to see the robot work. In our study of YouTube citations in Google’s AI Overviews, only 15.1% of cited videos were ones Google also showed on page one of the same search, and the text Google displayed came from what was said in the video, not its description. Our inference: a captioned application video from a robot maker or integrator can be read as evidence, even if it has few views. Second, industry associations and trade press carry weight: the IFR, A3 and PMMI publish the numbers that journalists and assistants repeat.

What do robot buyers ask AI assistants?

Application questions first, then brand comparisons, business model and support questions.

The questions below are written by us to show typical patterns. They are examples, not records of real searches.

StageIllustrative question
Application fit“Best collaborative robot for palletizing 20 kg cases at the end of a packaging line”
Robot type“Cobot or industrial robot for tending two CNC lathes?”
Mobile robots“Mobile robots that move totes between work cells without changing the floor”
Emerging tech“Are humanoid robots actually working in factories yet?”
Business model“Robot as a service versus buying robots for a small food plant”
Support“Robot brands with service engineers and spare parts in the Midwest”
Comparison“Compare these two cobot brands on payload, reach and ease of programming”

Two things stand out. Many questions describe a task, not a product, so a robot maker named only for its model numbers can miss them. And new buyers ask basic questions: the A3 data shows growth in semiconductor, pharmaceutical and food plants, many with smaller automation teams than an automaker. Our inference is that these buyers lean harder on outside research, including AI answers, before they know which integrator to call.

The humanoid question shows the risk of hype. The IFR estimates that about 7,000 full-size humanoids were sold in 2025 for commercial and professional uses, and its industrial robot report says most humanoids remain prototypes or pilots. An assistant that repeats press releases can leave a buyer with the wrong idea about what is ready today, in either direction.

How does an AI answer become a robot order?

Through a brand shortlist, an integrator conversation, an application test, a pilot cell and then fleet orders.

  1. Shortlist. The engineer asks which robots suit the task. Brands named in the answer, and in the pages it cites, become the starting list.
  2. Check. The engineer looks for specifications, application videos, safety information and nearby integrators or distributors. Missing information ends consideration quietly.
  3. Integrator. An integrator is brought in, or asked which brand it recommends. Integrators research too, so the same visibility matters at this step.
  4. Application test and pilot. The robot maker or integrator tests the part, cycle time and gripper. A first cell is installed.
  5. Fleet and service. If the pilot meets its targets, the plant repeats it, signs service or RaaS terms, and may standardize on the brand.

AI visibility can influence steps 1 to 3. Cycle time, price, support and the integrator relationship decide the rest. In PMMI’s study, integration, serviceability and total value drove adoption decisions, which is why those facts belong in what an assistant can find.

What decides whether an assistant names your robots?

Outside coverage it can find and specifications it can check; the platforms document their search, not their choices.

Each platform explains its search but not its picks. According to OpenAI (opens in a new tab), ChatGPT search breaks an engineer’s question into one or more narrower queries for its search providers, and a robot maker’s site is eligible only if it admits OpenAI’s crawler, OAI-SearchBot. Google (opens in a new tab) says AI Mode fans a question out into related searches on its subtopics, which it calls “query fan-out.” Neither publishes how a robot brand is chosen.

Observed in our studies:

  • Assistants look for rankings and awards. In our hidden searches study, ChatGPT ran a search aimed at a named publication, ranking or award in 43.8% of its answers.
  • Many “best” lists are self-serving. In our study of cited best-of lists, 24.2% of numbered lists with an identifiable publisher ranked their own publisher first. Robot makers that publish their own rankings should expect buyers, and possibly assistants, to discount them.
  • Brand familiarity still tips decisions. In the engineers survey, 70% were likely to pick the better-known brand when two solutions look technically similar, and 53% said familiarity influenced their most recent purchase.

Our inference for robot makers: the trust signals are the ones an engineer already checks. Payload, reach, repeatability and speed for each model; the applications it is proven in; safety standards and certifications; integrator and service coverage by region; and reference deployments that customers allow you to name. The IFR notes that local engineering, system integration and service infrastructure matter more as production regionalizes, and it says most robots installed in the US are still imported from Japan and Europe. Where your support is, and who installs your robots, are facts buyers ask about. Plant software vendors face similar checks, covered in how MES and IIoT platforms get found.

What does a robot maker lose when assistants leave it out?

First-time buyers in growing industries, and the integrators who standardize on a brand they find first.

We found no measurement of robot sales lost to AI absence, so this is reasoning, labeled as such:

  • The growth is in new hands. Non-automotive customers made up 56% of robot units ordered in North America in the second quarter of 2026. Our inference: many of these plants have no incumbent robot brand, so early research decides who is considered.
  • Brand loyalty forms after the first cell. A plant that standardizes on one brand for training and spare parts is hard to win later. Missing the first shortlist can cost the fleet.
  • New models lag in AI answers. Assistants often answer from older training data. Our guide on why ChatGPT misses new products explains why a new cobot or mobile robot can be invisible for months.
  • Errors spread. A wrong payload, a discontinued model or an outdated support region in an answer filters you out. Our guide to fixing wrong brand information in AI answers walks through the correction.

How does GEO work for a robotics company?

Generative engine optimization (GEO) helps AI assistants find, describe accurately and verify your robots for the tasks buyers ask about.

For robot makers, the work usually covers:

  1. Application pages. One page per task, such as palletizing, machine tending, welding, inspection or material transport, with the models that fit, typical cycle times, payload limits and industries served, written in text.
  2. Specification facts that match everywhere. The same figures for payload, reach, repeatability and safety ratings on your site, distributor pages, datasheets and marketplaces.
  3. Integrator and service directory. Who installs and supports your robots, by region, in a page an assistant can read.
  4. Captioned video. Application demos with spoken explanations and accurate captions, since AI Overviews quote what is said in a video.
  5. Independent coverage. Trade press, association data, conference talks, awards and case studies published with customer permission. Our article on how brands build authority for AI search covers the outlets that matter, and why “best of” lists matter explains why third-party rankings carry weight.
  6. Honest answers to hard questions. When a cobot is the wrong choice, what RaaS costs over time, and what your humanoid or AI features can and cannot do today.
  7. Measurement. Ask a fixed set of application, comparison and support questions across ChatGPT, Gemini, Perplexity, Copilot and Google’s AI features, repeatedly, and record who is named and which pages are cited. Then compare with demo requests and integrator leads.

None of this guarantees a recommendation. It makes your robots easier to verify for the tasks you actually do well.

What is still unknown about AI and robot purchases?

How many robot orders begin with an AI answer, and how assistants handle robotics questions specifically.

  • No attribution data. The surveys show that technical buyers use AI. None links AI answers to robot orders.
  • Sources have interests. The IFR, A3 and PMMI represent industry members, TREW and GlobalSpec sell marketing to engineering firms, and Teradyne reports its own results.
  • Our studies were broad. They covered buyer questions across several industries, not robotics. Applying them here is our inference.
  • Integrator influence is unmeasured. We found no public data on how often integrators pick the brand, or how they use AI themselves.

Where should a robotics company start?

Begin with the ten applications you most want to win, and ask assistants which robots suit each one.

Phrase the questions as a plant engineer would, with the task, payload, industry and region. Run them across the main assistants and note whether your robots appear, whether the specifications are right, which integrators and sources are cited, and which brands are named instead. That shows where buyers and integrators are being steered, and what evidence is missing.

If your growth depends on more qualified demos, application tests and pilot cells, book a review of your robots’ AI visibility with us. We will show where your robots appear when engineers and integrators ask AI for options, why other brands are chosen, and which changes are most likely to bring in the right projects. How the ongoing work is run, from application pages and spec consistency to captioned video, is set out on our generative engine optimization service page.

Frequently asked questions

Do integrators use AI assistants to choose robot brands?

We found no public data on that. Integrators research applications and components like other engineers, so it is reasonable to expect some do. Ask your integrator partners directly.

Should we publish our own “best cobots” list?

It may help buyers, but self-ranking lists are common and easy to discount. Of the numbered lists assistants cited in our study, 24.2% put their own publisher at the top. Independent rankings and reviews are stronger evidence.

Will humanoid robot news crowd our products out of AI answers?

It can shape broad questions. Task-specific questions with payload, cycle time and industry details are where proven robots are more likely to appear.

How soon will assistants reflect a new robot model or an updated spec sheet?

There is no fixed timeline. Assistants that search the web can use new pages quickly; answers from training data change only when models are updated.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work