How AI Assistants Find Office Space, and What Operators Should Do About It
AI assistants find offices in three ways: what they learned in training, what they can read on the web when they search, and what they can query through connectors. Operators can influence all three by publishing clear, consistent and current facts that assistants are allowed to read.
Reviewed on
What Are the Three Routes?
When a founder asks an assistant where to find an office for a team of twelve in Shoreditch, the answer names a handful of businesses and, sometimes, a few buildings. If your offices are not among them, it is worth knowing why.
First, training. Large models learn from a vast amount of public text, so businesses that have been written about often, over years, are more likely to be named from memory. Second, live search. Many assistants can search the web as they answer, read some pages and cite them. Third, connectors. An assistant can query a service directly through a connector built on the Model Context Protocol[1] and receive structured facts, such as a listing's desks, price and availability.
What Controls Whether Assistants Can Read Your Website?
Your robots.txt file, mostly. Each AI company publishes the names of its crawlers. OpenAI says it uses OAI-SearchBot to surface websites in ChatGPT's search features, and that sites opted out of it will not be shown in ChatGPT search answers, though they can still appear as navigational links[2]. It uses a separate crawler, GPTBot, for training, and a site can allow one while blocking the other[2]. Anthropic publishes the robots it uses for training, for search and for retrieving pages at a user's request[3]. Google says its Google-Extended control does not affect a site's inclusion in Google Search[4].
Should an Operator Block AI Crawlers?
That is a business decision, and the answer can differ by crawler. If you block search crawlers, you are unlikely to be cited in those assistants' search answers. Blocking a training crawler is a separate choice about whether your content is used to train future models. Check your robots.txt and your security settings today, because a crawler can end up blocked without anyone in the business deciding it, and you may be invisible without knowing it.
What Do Google's AI Answers Need?
Google says there are no additional requirements to appear in AI Overviews or AI Mode, and that the usual best practices for search still apply[5]. In practice that means pages that can be crawled and indexed, that load without depending on scripts to show their main content, and that answer a question clearly.
What Should Be on Each Office Page?
The facts an assistant needs to match a request, written in plain text: the address and area, the number of desks, the price and what it covers, the term each price applies to, whether VAT is included, the date availability was last confirmed and the name of the operator. If those facts appear only in an image, a PDF or after a script runs, some crawlers may not see them at all.
Does Structured Data Help?
It helps machines read a page, but it must match what people see. Google's structured data guidelines say to provide up to date information, not to mark up content that is not visible to readers, and not to mark up misleading content[6]. A price in the markup that no longer matches the page is worse than no markup at all.
What Is llms.txt?
It is a proposal for a plain text file at the root of a website that tells language models what the site is and where its most useful pages are[7]. It is not a formal standard, and assistants do not all read it, but it is cheap to publish and does no harm if it is kept accurate.
Why Does Consistency Matter?
Assistants decide which business a name refers to partly by checking that independent sources agree. If your operating company, brand name, address and description differ between your website, your Companies House record, your listings and the directories that mention you, an assistant is more likely to confuse you with someone else or leave you out. Use one name, one description and one set of facts everywhere.
How Do Connectors Change Things for Operators?
A connector lets an assistant ask for offices that match a request and get current facts back, rather than relying on what it remembers or can find on the web. That rewards accurate, structured listings. District (district-hq.com), a managed office marketplace for London, offers one. District offers a free, read-only connector that works with AI assistants such as Claude and ChatGPT. An assistant can search District's published, verified listings and open a single listing, and every result links back to the listing page on District. The connector cannot book anything, cannot see anyone's account and never shows an operator's contact details, so enquiries and viewings still happen on District. District's listings are being onboarded, and none are published yet.
What Should an Operator Do This Month?
Six things:
- Check your robots.txt against each AI company's published crawler names.
- Make sure each office page states its facts in plain text.
- Keep structured data in line with the visible page.
- Use the same company name and description everywhere.
- Publish an llms.txt if your site can.
- List your offices where assistants can query them directly, keeping availability current in every place you list.
None of this is a trick. Assistants recommend what they can read, understand and trust, so the work is making your facts readable, clear and the same everywhere.
Sources
- What is the Model Context Protocol (MCP)?, Model Context Protocol. Read on .
- Overview of OpenAI crawlers, OpenAI. Read on .
- Does Anthropic crawl data from the web, and how can site owners block the crawler?, Anthropic (Claude Help Center). Read on .
- Google's common crawlers (Google-Extended), Google Search Central. Read on .
- AI features and your website, Google Search Central. Read on .
- General structured data guidelines, Google Search Central. Read on .
- The /llms.txt file, llms-txt. Read on .