AI engines cite pages that answer a buyer's question near the top of the page, in plain and checkable terms, using HTML a crawler can read without running scripts. They lean towards fair comparisons, recently updated pages and sentences full of named products and specific facts. They do not reward AI-only files or extra markup. Below is what the research shows, and what to change on your own site.
Do AI engines need a different kind of page from Google?
No. Google's guidance on its AI features says "There are no additional requirements to appear in AI Overviews or AI Mode". A page has to be indexed and eligible to show a snippet in Search, which is the same bar as any ordinary result. Google adds that you do not need new machine-readable files, AI text files or special markup.
That rules out a common temptation: building a second, AI-only version of each page. One page, written for the person making the decision, is the asset that both search and AI answers draw on. The work is in making that page clear, fair and easy to read.
Which pages do AI engines trust most?
DeltaV Digital tracked more than 25,000 citations across ChatGPT, Perplexity, Gemini, Google AI Overviews and Google AI Mode between April and July 2026. Each time an engine retrieved a comparison page, it cited it nearly one and a half times as often as the average page. Yet comparison pages made up only about one citation in 25.
Read together, those two numbers describe an opening. When an engine finds a well-made comparison, it relies on it heavily. Few businesses publish them, so supply is thin. A head-to-head page that settles a real buying decision is one of the most efficient pages a company can add to its site.
Why can a "best of" list help your rivals instead of you?
Fairness is the part most companies get wrong. In research published in June 2026, Lily Ray tracked 100 prompts. When a brand's own self-promotional list was cited in a Google AI Overview, that brand was left out of the recommendation in more than two out of every three cases. She also reported sharp falls in organic traffic from January 2026 for sites leaning on these lists, which continued through Google's May 2026 core update.
The engine used the page as evidence about the category, then recommended someone else. A list that ranks its author first, with no method behind it, reads as advertising. A comparison that says plainly where a rival is the better choice reads as a source.
In practice, that means three habits:
- say where the alternative wins, and let the point stand;
- show how you compared the options and when you last checked the facts;
- keep "best for" pages narrow, and publish them only where you have tested the options yourself.
Where on the page do citations come from?
Mostly from the top. Kevin Indig analysed 1.2 million ChatGPT responses containing more than 18,000 citations. More than two in five citations came from the first 30% of a page and about three in ten from the middle, while the final third supplied less than a quarter. Cited passages were dense with named brands, products and people, and written in a measured, analytical tone. Headings phrased as the questions people actually ask performed especially well.
Microsoft's Bing team gives matching advice. Its October 2025 guidance recommends comparison tables for feature comparisons, answers of one or two sentences that still make sense when lifted out of the page, and measurable facts in place of vague claims.
For your pages, the verdict comes first. Put a two or three sentence answer and the comparison table in the first screen. Phrase each heading as the buyer's question and answer it in the sentence directly beneath. Name the product, the feature and the number, rather than writing "it", "the platform" or "fast".
Can AI crawlers read your site at all?
Often not as well as you would expect. A crawl study by Vercel and MERJ found that OpenAI's GPTBot, Anthropic's ClaudeBot and PerplexityBot fetch JavaScript files but do not run them. Anything a script adds to the page after it loads is invisible to them. Gemini, which uses Googlebot's infrastructure, does render scripts.
Microsoft adds a second warning. Answers hidden in tabs or expandable menus may not be rendered by AI systems, and details held only in images can be missed.
There is a simple test. Open your page with your browser's view-source option. If the comparison table, the prices or the FAQ answers are not in that raw HTML, most AI crawlers cannot see them. Render the page on the server, keep FAQ answers open, and repeat any figure from a chart or screenshot in the text.
Do freshness and schema markup make a difference?
Freshness does. Schema matters less than many people hope. Ahrefs analysed about 17 million citations across seven AI platforms and found the cited pages were, on average, about a quarter fresher than the pages in Google's organic results. ChatGPT showed the strongest pull towards recent content. Google's AI Overviews were the exception: their top citations were slightly older than the organic results.
Schema is housekeeping rather than a lever. In a May 2026 study, Ahrefs tracked nearly 1,900 pages that added JSON-LD markup and compared them with similar pages that did not. No platform showed a meaningful lift in citations. The study only covered pages that AI engines already cited heavily, so it does not settle whether markup helps a new page get noticed.
Keep basic markup for Google, show a real "last checked" date, and change that date only when you have actually re-checked the facts.
What Google and AI engines each look for
| Stage | Google search | AI engines | What your page should do |
|---|---|---|---|
| Being found | Indexed and eligible for a snippet, with the query's wording in the title and headings | Pages retrieved for the question, then individual passages chosen to answer it | Use the buyer's own wording in the title, the main heading and the section headings |
| Being chosen | The whole page judged for relevance; the title and snippet earn the click | Passages that answer one point cleanly, with the top of the page weighted most | Put the verdict and the comparison table in the first screen |
| Being trusted | First-hand evidence and honest pros and cons | Named products, specific facts and an even tone | Say where rivals win, cite your sources and show your method |
| Reading the page | Renders JavaScript | Most AI crawlers read the raw HTML only | Render everything important on the server |
| Freshness | Matters most for facts that change, such as prices | A stronger pull towards recent pages, except in Google's AI Overviews | Show a "last checked" date and update it after real checks |
Being found
- Google search
- Indexed and eligible for a snippet, with the query's wording in the title and headings
- AI engines
- Pages retrieved for the question, then individual passages chosen to answer it
- What your page should do
- Use the buyer's own wording in the title, the main heading and the section headings
Being chosen
- Google search
- The whole page judged for relevance; the title and snippet earn the click
- AI engines
- Passages that answer one point cleanly, with the top of the page weighted most
- What your page should do
- Put the verdict and the comparison table in the first screen
Being trusted
- Google search
- First-hand evidence and honest pros and cons
- AI engines
- Named products, specific facts and an even tone
- What your page should do
- Say where rivals win, cite your sources and show your method
Reading the page
- Google search
- Renders JavaScript
- AI engines
- Most AI crawlers read the raw HTML only
- What your page should do
- Render everything important on the server
Freshness
- Google search
- Matters most for facts that change, such as prices
- AI engines
- A stronger pull towards recent pages, except in Google's AI Overviews
- What your page should do
- Show a "last checked" date and update it after real checks
A checklist for your next page
That last point matters more than it looks. Being cited and being recommended are different outcomes, and only one of them wins the deal. It is why we are building this kind of tracking into Vyera, our AI co-founder platform at vyera.ai.
Where to start
Most sites do not need more content to be cited. They need their most important pages to answer the question sooner, to treat rivals fairly and to be readable by the crawlers that matter. If you want a clear view of how your site reads to AI engines, book an X-Ray with Linvex. We take the time to understand your business and what you want from it, look at the pages that drive your pipeline and tell you plainly what to fix first.
Sources
- AI citation study, DeltaV Digital, April to July 2026 data
- Why calling yourself the best could be helping your competitors win in AI search, Lily Ray, June 2026
- Studies reveal AI citation clues, Practical Ecommerce, March 2026, on Kevin Indig's study
- How ChatGPT decides what content is worth citing, Stan Ventures, on Kevin Indig's study
- Optimizing your content for inclusion in AI search answers, Microsoft Advertising, October 2025
- AI features and your website, Google Search Central
- The rise of the AI crawler, Vercel and MERJ, December 2024
- AI assistants prefer to cite fresher content, Ahrefs, July 2025
- Schema and AI citations study, Ahrefs, May 2026
Written by
Rana Daniyal
Chief Executive Officer, Linvex Solutions
