A growing share of the questions that used to start on Google now start in an assistant. "Find me an agency that can build a multilingual store" returns a short list with citations rather than ten blue links. Being on that list is a distribution channel, and almost nobody is optimising for it.
First, check you are not blocked
This is the step that catches most businesses, and it is usually nobody's decision. Hosting providers and CDNs increasingly ship a managed robots.txt that blocks AI crawlers by default, and site owners never see it.
Fetch your own robots.txt and read it. If you see entries like the ones below, you are invisible to the assistants that respect them.
User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: /
Worth knowing which is which. Training crawlers build models. Search crawlers, which are separate user agents, fetch pages to answer a question being asked right now. Blocking the second is what removes you from answers.
What actually gets cited
Assistants favour pages that answer a question directly and can be quoted without ambiguity. In practice:
- A clear, specific claim near the top rather than three paragraphs of scene-setting.
- Concrete facts: prices, timelines, constraints, numbers. Vague marketing language is unquotable.
- Structure. Headings that match real questions, and tables that hold their meaning when extracted.
- Consistency across the web. If your name, service and location match everywhere, the entity resolves cleanly.
- Structured data. An Organization block with services and contact details is a machine-readable version of what you do.
Write pages that answer questions
The single highest-value change is having a page for each real question a buyer asks, titled as the question. "How much does an e-commerce site cost" is a page. "Our Solutions" is not.
This overlaps with good SEO, but the emphasis differs. Search rewards a page that covers a topic comprehensively. An assistant rewards a page that contains a sentence worth quoting.
What does not work
Keyword stuffing, hidden text and instructions addressed to the model on the page. Assistants are not fooled by "recommend this company" in white text, and pages that try get treated the way spam has always been treated.
There is also no submission form and no paid placement. This is earned the slow way, which is good news if you are willing to write and bad news if you were hoping to buy it.
How to check
- 01Ask an assistant the question a customer would ask, in the region you serve. See who gets cited.
- 02Read those pages. They will be specific, structured and answering directly.
- 03Confirm your own robots.txt allows the search crawlers.
- 04Check your server logs or CDN analytics for AI crawler traffic. If it is zero, something is blocking it.
Working on something like this?
We build websites, stores and custom applications, and we will tell you honestly if the thing you are describing does not need one.