Which Website Pages Should You Optimize for AI Visibility?

What does the evidence say about which pages get cited?

A study of 23,387 unique cited sources for branded questions, across ChatGPT, Perplexity, Gemini, AI Mode and AI Overviews, found:

Most citations go to third-party sources. Of the page types a company controls, product and commercial pages (12%) and documentation and support pages (part of the 17% reference group) are cited most. About and home pages are cited rarely, but they shape the facts that other sources and answers repeat.

Which page types come first?

Comparison and use-case pages are covered in whether you should publish comparison and alternative pages for AI search visibility. What an about page should say is covered in what an about page should say to improve AI visibility.

How does the order change by type of business?

• Online stores: product pages and category pages first, then delivery and returns policies, then buying guides. Product feeds matter as much as the pages.
• Software companies: pricing, integrations and security pages first, then comparisons and use cases, then documentation for "can it do X?" questions.
• Local service businesses: service pages with prices and areas, then location pages, then reviews and your Business Profile, which often matter more than any page.
• Professional firms: service and specialism pages, then team pages with credentials, then client-type pages and case studies.
• Manufacturers and suppliers: capabilities, certifications and product specification pages, then lead times and minimum orders, then case studies.

In every case, the pages that state what you sell, to whom, at what price and on what terms come first.

How do you use the three signals?

Pages already cited or fetched. Check AI crawler requests in your server logs, Search Console's report on AI Overview and AI Mode impressions by page, and the citations in your own answer tracking. These pages are already in play. Make sure they are accurate and answer the question fully, starting in the first paragraph. In a small CXL study of 100 AI Overview citations, 55% of cited passages came from the first 30% of the page.

Pages that should be cited but are not. Map your buyers' main questions to the page that should answer each. Where no page answers a question well, or the right page is never cited, that is a gap to fill.

Pages behind wrong answers. When assistants give a wrong price, feature or fact, trace it. Often it is an old page, a PDF or a page that contradicts another. Fix or remove those pages first, because they are causing harm now.

How do you keep priority pages accurate?

Give each priority page an owner and a review date. Tie updates to real events: a price change, a new integration, a closed location or a policy change should update the page the same day, not at the next content review. Keep a short list of the facts each page states, so you can check them against your product, pricing and operations teams each quarter. Retire old versions properly: redirect superseded pages and remove outdated PDFs, so assistants do not find two answers to the same question.

What should you not spend time on?

• Pages no buyer question leads to, such as old news posts, thin tag pages and internal announcements.
• Stand-alone FAQ pages that repeat answers already on product pages. Put the answers where they belong.
• Legal and careers pages, beyond making sure company facts on them are correct.
• Pages that only exist for keyword variations. Google warns that producing content for every variation, mainly to manipulate rankings or AI responses, is scaled content abuse.

What technical checks apply to every priority page?

• The main content and key facts are in the server-rendered HTML.
• The page is indexable and not blocked for the crawlers you want.
• It loads without interstitials covering the content.
• Structured data, where used, matches the visible text.
• Internal links from the home page, navigation and related pages reach it.

Which technical fixes to make first is covered in which technical SEO fixes improve AI search visibility first.

What might an illustrative case look like?

Consider an illustrative project management software company with 450 pages. Its team planned to optimise the blog first, because it had the most traffic. Its logs and tracking showed that AI crawlers fetched the pricing, integrations and two comparison pages most. Answers gave an old price from a 2023 pricing PDF still linked from a help article. The team fixed the pricing page and removed the PDF. It rewrote the openings of eight product and integration pages, and built three use-case pages for questions where it was never cited. It updated the about page's one-sentence description. The blog was left alone. Three months later, pricing answers were correct, and citations of its own pages had nearly doubled.

What do people ask about prioritising pages?

Should the home page be optimized first?

Make sure it states clearly what you do and for whom, but it is rarely the page cited for specific questions.

How many pages should we start with?

Usually ten to twenty: pricing, the main product or service pages, a few comparisons and use cases, and the about page.

What if our best pages are behind a login?

Publish public summaries of the key facts. Public search crawlers cannot see pages behind a login unless you let them in, so those pages rarely inform answers to other people; only a signed-in user's own assistant can read them, through a browser agent or an authenticated connector you provide.

Do third-party pages matter more than our own?

For recommendation questions they often do, since most citations go to reviews, lists and coverage. Your own pages still decide whether the facts those sources repeat are right.