Ranking isn't enough: how to get cited inside AI answers

For most of my career, search visibility meant one thing: where you showed up on a results page. That still matters. But a growing share of questions now get answered before anyone sees a list of links. Someone asks ChatGPT, Perplexity, Claude or Google's AI features a question, gets a paragraph back, and maybe a few sources underneath.

If you're not one of those sources, ranking fourth doesn't help much.

I've spent years working on high-value legal websites in one of the most competitive spaces on Google, and part of that work has been making those sites authoritative not just in classic results but inside AI answers. Here's what I've learned about what actually moves that needle, and what I think is noise.

A caveat: nobody outside these companies knows exactly how their systems choose sources, and they change constantly. What follows is based on what's observable, what the platforms document, and how retrieval sensibly works. Be suspicious of anyone claiming to know the internals.

How AI citation differs from ranking

Classic ranking is a list. The searcher does the reading and deciding.

AI answers work differently. Most of these tools retrieve a set of candidate pages, pull passages from them, and write a synthesized answer. The citation goes to the source whose content made it into that answer. That shifts what you're competing on:

Passages, not pages. A page can rank well overall and still not have a single passage that cleanly answers the question. The page that gets cited is often the one with the clearest, most self-contained answer.
Confidence, not just relevance. An answer engine that states something as fact has reason to prefer sources that are corroborated elsewhere and clearly attributable to a real entity.
Access. If the crawler for a given tool can't fetch your page, or it isn't in the index that tool searches, you don't exist to it.

Ranking still feeds this. Google has said its AI features draw on the same core systems as Search, and a page has to be indexed and eligible to show a snippet to be used as a supporting link. Classic SEO is the foundation. It's just no longer the whole building.

Be crawlable by the tools you want to be cited by

This is the unglamorous part, and it's where I'd start, because it's binary. Check your robots.txt, your CDN or firewall bot rules, and your server logs.

Things I check:

Which bots you allow. The major AI companies publish their crawler user agents, and several separate "search/retrieval" bots from "training" bots. You can make a deliberate choice about training while still allowing the bots that fetch pages for live answers. Blocking everything with a blanket rule usually isn't a decision anyone actually made.
Firewall and bot-protection settings. Aggressive bot rules at the CDN layer can silently challenge or block legitimate crawlers. Your robots.txt says yes; your firewall says no.
Content that only appears after JavaScript runs. Not every crawler renders JavaScript the way Googlebot does. If your key answer is injected client-side, some tools may never see it. Server-rendered HTML is the safe default.
Google's controls. Google-Extended is a robots.txt token for Gemini-related uses; Google's documentation says it doesn't affect inclusion in Search, and AI Overviews are governed by regular Search crawling and snippet controls. Know which lever does what before you pull one.

The simplest test: check your raw access logs for the bots you care about getting 200 responses on your important pages.

Write content that can be lifted cleanly

When a system pulls a passage, it wants something that stands on its own. Most web copy doesn't. It's written as a slow build toward a point, with the actual answer buried in paragraph six.

What works better:

Answer first. Put the direct answer in the first sentence or two under a heading that matches the question. Then explain.
One idea per section. Headings that describe what's underneath, not clever teasers. "How long does the process take?" beats "The waiting game."
Self-contained passages. Avoid "as mentioned above" and pronouns that only make sense in context. Name the thing.
Specific and precise. Vague copy gives a model nothing to quote. Defined terms, clear conditions, and plain statements of what is and isn't true are what get lifted.
Current and dated. Show when a page was updated, and actually update it.

On the legal sites I've worked on, this often meant restructuring existing pages more than writing new ones. The expertise was there; it just wasn't shaped for quick grabbing.

Make it obvious who is speaking

An answer engine deciding whether to trust a claim benefits from knowing who made it. That's where entity signals come in: clear, consistent evidence of who you are as a person or organization.

In practice:

A real About page that says who you are, what you do, and why you're credible.
Author information on content, with a consistent name that connects to a profile page.
Organization and Person structured data, with sameAs links to your genuine profiles elsewhere.
The same name, description and details everywhere you appear.

This is the piece most sites skip. If Google and AI tools can't confidently connect your site to a known entity, you're asking them to trust an anonymous page.

Get corroborated across the web

This is the part you can't fully control, and it matters. If the only place a claim about you exists is your own website, that's weak evidence. When independent sources describe you the same way, it's stronger.

That doesn't mean chasing links for their own sake. It means showing up where your audience already looks:

Industry directories and professional listings that are actually maintained.
Podcasts, interviews and long-form media where you say something useful, with a link back to your canonical site.
Being quoted or referenced because you published something worth referencing.
Consistent business profiles, so every mention points at the same entity.

Media is undervalued here. A podcast or video with a proper description, transcript and link becomes another corroborating source that connects your name to a topic.

Structured data: helpful, not magic

Schema markup gets oversold. It won't make a weak page get cited. What it does is remove ambiguity: this is an organization, this is its logo, this person wrote this article, this page was updated on this date, these profiles belong to the same entity.

Use the types that genuinely describe your content: Organization or LegalService for the business, Person for authors, Article for posts, FAQPage only where there really is a Q&A. Validate it with Google's Rich Results Test and the Schema.org validator. Make sure it matches what's visible on the page. Markup that contradicts the page is worse than none.

How I'd approach it

If I were starting on a site tomorrow, this is the order:

Access. Confirm the relevant crawlers can reach and read your key pages.
Foundation. Make sure the pages are indexed and performing reasonably in classic search.
Identity. Fix the About page, author pages, Organization/Person schema and profile consistency.
Shape. Restructure your most important pages so the answers are direct, specific and self-contained.
Corroboration. Build a presence beyond your own domain that says the same things about you.
Observe. Ask the tools the questions your customers ask. Note who gets cited, including you. Repeat regularly, because the answers change.

That last step is the cheapest and the most skipped. If it isn't measured, you're guessing. It's manual and imperfect, but it beats any dashboard claiming to track "AI visibility" with suspicious precision.

The takeaway

Getting cited inside AI answers isn't a new trick layered on top of SEO. It's what good SEO was always supposed to be, with less tolerance for shortcuts: be reachable, be clear, be identifiable, and be confirmed by people other than yourself.

The sites that win tend to be the best answer and make that easy to verify. That's also why I think of this as a systems problem, not a content problem, which I wrote about in every web problem eventually becomes a systems problem. If you want to talk through how this applies to your site, get in touch.