AI visibility: being the thing an assistant cites

Someone asks an assistant which accountant, installer or agency to use in their area. The answer may mention businesses, cite supporting pages or offer links for further reading. The presentation varies by platform and question. AI visibility work examines whether useful business information can be found and accurately represented, then records the mentions, citations or visits actually observed.

Much of this overlaps with ordinary search work: accessible pages, useful answers, clear facts and evidence readers can inspect. Different platforms retrieve and select sources differently; no page format guarantees selection.

What the work involves

Check whether the crawlers can reach you at all

Start by identifying the purpose of each access control. Search retrieval and model training are separate uses, so allowing every AI bot is not a prerequisite for search visibility.

OpenAI documents OAI-SearchBot as its search crawler and GPTBot as a crawler for content that may be used in model training. Their controls are independent: a site can allow search access while declining training access. ChatGPT-User handles some user-initiated visits and is not the search-index control.

For Google's AI Overviews and AI Mode, Google's guidance points to normal Search access and snippet eligibility. Google-Extended controls training and grounding in specified Gemini products; it does not determine inclusion or ranking in Google Search. Check each other provider's current documentation before changing its permissions.

Review robots.txt and server logs to see which bots arrive and what status codes they receive. Check whether content is present in the HTML or only appears after JavaScript runs. Investigate rate limiting, bot protection and geographic rules that may block a crawler while allowing a human browser.

Crawler access is a deliberate publishing decision. Some businesses want particular AI crawlers to access their content; others do not. Understand each crawler or token’s documented purpose before changing permissions.

Make the answers liftable

Clear passages help readers understand an answer even when they arrive partway through a page. Use descriptive headings and state important facts in text. A short answer near the top can help, but keep qualifications beside the claim they limit. These are editorial practices, not documented rules that guarantee AI citations.

When a page buries its specifics or puts essential information only in a graphic, readers and retrieval systems may have difficulty interpreting it. Keep important information accessible in the page's text.

So a chunk of this work is rewriting and restructuring. Putting the answer first. Turning implied claims into stated ones. Making sure that your service area, your opening hours, your qualifications and what you actually do are written as plain text in a sensible place. It improves the page for readers too, which is the usual sign that a technique is sound rather than a trick.

Structured data

Schema markup tells a machine what your page is about in a format it does not have to guess at. The vocabulary is published at schema.org and the type names are spelled the American way: Organization, LocalBusiness, Service, Person, Article. FAQPage earns far less in search results than it once did, so use it only where it is relevant and supported. Markup is not magic and it does not make you rank, but it removes ambiguity about your name, location, contact details and what you sell.

Add markup where it describes something real and remove unsupported claims. Markup that claims things the page does not say is a liability, not an advantage.

Google does not require special AI schema or an AI text file for its Search AI features. Relevant structured data should match visible content; it is not a separate route to guaranteed citations.

Factual consistency across the web

An answer may draw on sources beyond your website. An old directory address or an inaccurate business description can create conflicting information. How a platform resolves that conflict is not fully documented. Correcting the records you control improves the information available to readers without promising a ranking or citation benefit.

List significant business records and update errors under your control: your website, eligible Google Business Profile, industry directories and professional registers. Describe each address accurately; a registered office and a customer-facing trading address can legitimately differ.

Measurement

You cannot improve what you are not watching, and this is the part most people skip because it is fiddly.

Use a repeatable set of questions a real prospect might ask. Record the assistant, date, wording and relevant context, whether the business was mentioned, how it was described and who else appeared. That gives you a baseline and a trend line rather than an anecdote. Answers vary between runs, so a single result proves very little and the pattern over months is the thing to read.

Also review server logs for AI crawler activity and identifiable referrals from assistant interfaces. It is imperfect. Some of this traffic is invisible, because someone reading an answer that mentions you and then typing your name into a browser leaves no trace connecting the two. Keep those attribution gaps explicit.

A crawler request does not prove that an answer cited your page. Google includes its AI-feature traffic within Search Console's overall Web reporting; that total is not a separate count of AI citations.

What nobody can promise

This work is new. The systems change without notice, the crawlers change their behaviour, and nobody outside those companies knows exactly how a citation is chosen.

No provider can promise a particular organic citation in another company's assistant. A submission mechanism, product feed or paid advertising placement is not a guarantee of organic citation. Ask a proposal to distinguish the platform, placement and outcome it can actually deliver.

Practical work can remove obstacles, improve source clarity and measure observed results. That is an honest description of the work, and it is the same honest description that applied to search engine optimisation before anyone called this AI visibility. If you want the longer version of the reasoning, we wrote it up in getting cited by AI, and the underlying mechanics are covered in how search decides what to show.

When you do not need this

If important pages cannot be crawled or basic facts are wrong, investigate those foundations first. That overlaps with ordinary SEO work; the scope depends on the cause and the site.

If you sell to a handful of clients through personal relationships, or you are at capacity and turning work away, the likely value should be assessed before committing to the work.

Further reading and practical help

Start with the guide to getting cited by AI and our SEO audit overview. For hands-on support or a quotation, contact our sister agency SuffolkWeb. Scope, deliverables and commercial terms depend on the project.

Read our editorial approach, or contact The SEO Blog about a correction or question.