SEO for AI search: how to get cited by ChatGPT, Claude and Google AI

A growing share of searches now end in an AI-written answer — in Google’s AI results, in ChatGPT, in Claude, in Perplexity — instead of a click on ten blue links. SEO for AI search is about making sure that when those assistants answer a question in your field, your site is one of the sources they read, trust and cite.

Some people call it generative engine optimization. The label matters less than the fact that most of it is familiar work done more carefully.

How AI assistants pick their sources

When an assistant searches the web, it usually runs a normal search, fetches a handful of pages, extracts the relevant passages and writes an answer, citing the pages it used. That tells you what matters:

• You still need to be found — if you do not rank in the underlying search, you are not in the pool of pages fetched.
• Your page has to be readable by a bot — fast, server-rendered HTML, not content that only appears after JavaScript runs.
• The answer has to be extractable — a clear passage that answers the question directly beats a clever intro and a buried conclusion.

Technical foundations first

This is the part developers own, and I cover it in more depth in technical SEO for people who build the site. For AI search specifically:

• Render content on the server. Many AI crawlers do not execute JavaScript. If your text is injected client-side, they see an empty page.
• Check robots.txt deliberately. Decide which AI crawlers you allow. Blocking them all means you will not be cited; allowing search-focused ones while blocking training-only ones is a reasonable middle ground.
• Use structured data. Article, FAQ, Product, Organization and LocalBusiness schema make it easier to understand what a page is and who is behind it.
• Keep pages fast and stable. Crawlers fetching dozens of pages per answer give up on slow ones.

Write passages that can be quoted

AI answers are assembled from passages, so structure content in self-contained blocks:

• Use headings that match real questions (“How much does X cost?”, “Is X better than Y?”).
• Answer in the first one or two sentences under each heading, then explain.
• Define terms plainly. A clear one-sentence definition is exactly what gets lifted into answers.
• Use lists and tables for comparisons, steps and specifications.

This is also just better writing for humans in a hurry, so there is no trade-off.

Be a source worth citing

Assistants favour pages that add something the others do not. Generic content restating the top ten results gives them no reason to pick yours. What does:

• First-hand experience — what happened when you actually built, tested or ran the thing.
• Specific numbers and examples — your own data, benchmarks, before-and-after results.
• Clear authorship — a real person with a real track record, an about page, links to work.
• Freshness — dates on articles, and actually updating them when things change.

Brand mentions matter more than before

Assistants learn which names are associated with which topics from across the web. Being mentioned — in directories, comparisons, community discussions, podcasts, partner sites — increases the chance your brand comes up even when your page is not the one fetched. Classic link building still helps, but unlinked mentions in the right context count too.

How to measure it

Tracking is still immature, but you can do a lot by hand:

1. List 20–30 questions your customers actually ask.
2. Ask them in the main assistants every month and note who gets cited.
3. Watch referral traffic from AI assistants in your analytics — it is small but usually high-intent.
4. Track branded search volume, which tends to rise when assistants mention you.

Where to start

Fix rendering and crawl access, add structured data, then rewrite your five most important pages so each section answers one question directly. That covers most of the gap for most sites. If you want help with the technical side, that is what my SEO and technical performance work covers — get in touch.