Getting cited by AI assistants: what makes a page citable

en 7 min read

Getting cited by AI assistants: what makes a page citable

AI assistants that search the web pick sources the way a hurried researcher does: they run a search, open a handful of pages, and quote the ones they can reach, read and repeat without guessing. You cannot buy a citation, but the groundwork is ordinary search work: the SEO foundation is a one-off EUR 450 to 750, and on a new site it is decided at scoping and built in, from a single-page landing at EUR 600 to 900 upward.

How an AI answer chooses its sources

An assistant can answer in two ways. From memory, it draws on what it learned in training, which stops at a cutoff date and links to nobody. With search switched on, it looks things up: it sends queries to a search index, pulls back a set of pages, reads passages from them and writes an answer that links to the pages it used. Only the second kind cites you, so that is the one worth working on.

Which index it searches matters. Google's AI Overviews and AI Mode draw on Google's ordinary index, so a page has to be indexed and allowed a snippet in Google Search before it can appear there. Microsoft Copilot runs on Bing. Other assistants use their own crawlers, search partners, or both. A site that has only ever been registered with Google has left at least one door shut.

Then it reads, looking for a passage that answers the question, stated clearly enough to repeat, from a source it has reason to trust. Each of those can fail on an otherwise good website.

Five things that make a page citable

1. It can be reached

Check that robots.txt does not block the crawlers of the assistants you care about. OpenAI calls its search crawler OAI-SearchBot, Perplexity calls its crawler PerplexityBot, and each provider publishes the names it uses. Some CDN and security services also offer a switch that blocks AI crawlers, so check that setting too. A "noindex" tag left over from a staging site shuts out Google, AI Overviews included.

2. The words are in the HTML

Several AI crawlers read the HTML a server sends and do not run JavaScript. If your prices, service descriptions or opening hours only appear after scripts load in the browser, a person sees them and those crawlers may not. A site rendered on the server sends the full text in the first response, which is one reason we build with Next.js.

3. The answer comes first

An assistant pulls passages, not whole pages. A page that opens with three paragraphs of welcome before saying what a service costs makes the useful sentence hard to find. Put the direct answer at the top of the page and of each section, under a heading that names the topic plainly, so each section makes sense on its own.

4. It states specific facts

This is the part you control. An assistant asked what a website costs can cite a page that names a range. It cannot cite "every project is unique, get in touch". The same goes for what is included, how long the work takes, which areas you serve and what you do not do. Vague copy is not wrong, it is simply unusable as a source, which is one reason this site publishes its prices.

5. The facts agree everywhere, and carry a date

An answer about a local business can cite a directory, a review site or a forum thread as easily as your own page. If your opening hours say one thing on your site and another on your Google Business Profile, you have given it two answers and no reason to prefer yours. Keep your name, address, phone number and hours identical everywhere, and date any page whose facts change, such as prices.

Where the work sits, and what it costs

What the assistant needsWhat gets in the wayWho fixes itCost
To reach the pageBlocked crawlers, a leftover noindex, pages missing from Google or BingTechnical setupSEO foundation, EUR 450 to 750 one-off
To read the textContent that only appears after JavaScript runsThe buildA rebuild at your step, only if the current site cannot be fixed
To find the answerLong introductions, vague headingsCopy and page structureYour time, or copy help agreed at scoping
To quote a factNo prices, no scope, no service areaYouA decision, not a budget
To trust the factDifferent details on different sites, no datesYou, with a checklistYour time

The order to do it in

  1. Check access. Search Console for Google, Bing Webmaster Tools for Bing, robots.txt and any crawler-blocking setting. This is the indexation and technical hygiene inside the SEO foundation, a one-off EUR 450 to 750 that also covers analytics, titles, internal links and a 3 to 6 month plan.
  2. Rewrite the pages that answer buying questions. Prices, what is included, how long it takes, where you work. This is the part that waits on you, because the facts and what to publish are yours to decide.
  3. Rebuild only if the site cannot be read. If the text lives behind JavaScript, or the structure fights every edit, a rebuild follows the week range for its step: 1 to 2 weeks for a single-page landing, 2 to 3 weeks for a mini-site, 3 to 6 weeks for a business website of 4 to 7 pages, and 6 to 12 weeks for an advanced site or web app. You can see what each step includes.
  4. Keep adding answers. New service and location pages, internal links and reporting are the work of SEO growth, at EUR 500 to 800 a month.

When to spend less

If your customers come through referrals, repeat business or a Google Business Profile that already fills your calendar, you do not need a project called AI visibility. Run the access checks from the first item above yourself with the free tools named there, publish your prices and service area in plain words, and stop.

Do not pay for an "AI SEO" or "GEO" package that cannot say what it will change on your site, and hold off on monitoring subscriptions until the basics are done, because an answer checked on one day tells you little about the next. A proposed file called llms.txt is also sold as a shortcut, though Google has said it does not use it. And if your site is already indexed, sends its text without relying on scripts and states specific facts, you may need nothing from a studio at all. Then do not hire us yet.

Questions buyers ask

Does schema markup get me cited?

Google says pages need no special structured data to appear in its AI features. Markup that matches the page helps search engines read details such as your business name and address correctly, but it is groundwork, not the thing that gets you quoted.

Should I block AI crawlers to protect my content?

Blocking a provider's crawler keeps your pages out of what that crawler collects, and it can also keep them out of that assistant's answers. Some providers, OpenAI among them, run separate crawlers for training and for search, so you can block one and allow the other.

How do I know if an assistant has cited me?

When someone clicks a citation, the visit shows up in your analytics as a referral from the assistant's domain, such as chatgpt.com or perplexity.ai. You can also ask the assistants the questions your customers would ask and note which sources they name. Repeat the same questions over time, since a single answer varies.

Can you guarantee an assistant will mention my business?

No, and nobody honest can. The providers decide what their systems cite, the answers change from one run to the next, and nobody sells a placement inside them. What a studio can deliver is a site that is reachable, readable and specific: the part within anyone's control.

If you want a site that answers your customers' questions in plain words, tell us what you sell and where.

Building something?

JP Studio designs and builds websites, storefronts and product interfaces.

Straight to the person who would do the work. No newsletter, no call centre. Prefer email? support@jp-studio.com