The team at SALT has carried out research into whether content structure and templates have an impact on your content surfacing in Google AI Mode.

Now that Google AI Mode is available in the UK, CMOs and marketers are naturally keen to know how best to optimise their content.

A quick online search will throw up plenty of advice about how to structure your content and even AI-optimised templates you can use. But before you spend money on restructuring your page layouts and rewriting your copy, does this stuff work?

A templated approach will include multiple structural factors. Testing every element of every page layout would take far too long, so we decided to test just one common content recommendation to see if there was any correlation with visibility in AI Mode.

Is AI Mode more likely to cite text from “above the fold”?

Content marketers have long held that key information should appear “above the fold”; usually taken to be the first 800 pixels of a webpage. And it appears this has crossed over into some of the advice we’re seeing for how to optimise and structure your content for AI search.

If the data reveals some correlation, then perhaps there is a benefit to structuring your content in a particular way.

But if the data shows no correlation, then we can safely say that marketers should be more sceptical of templates or structural advice that promise to increase visibility in AI.

Key findings

1. Pixel depth doesn’t matter

We found no correlation between pixel depth and the text fragments highlighted by AI Mode. Text buried thousands of pixels down the page was just as likely to be highlighted as information prominently positioned near the top.

VERTICAL AVERAGE MEDIAN <800px
Travel 2,400px  1,200px  ~1/3 
E-commerce 3,600px  2,100px  ~1/4
SaaS 4,600px  2,500px  ~1/10 

This makes sense, as AI Mode uses the same fragment indexing technology that’s powered jump links in Google for more than four years. Google breaks webpages into sections during processing, then serves relevant fragments regardless of their position. The visual presentation in AI Mode might be new, but the underlying technology isn’t.

2. Page layout impacts pixel depth, but not AI visibility

The wide variability in depths, with outliers going as deep as 60,000px, suggests that page layout does play a role in these results. However, this doesn’t mean there is a causal link between page layout and AI visibility.

The variations in pixel depth are a product of these wildly different page layouts, not because AI Mode favours one layout over another. And certain layouts or structural elements are more common in some verticals than others.

Simple blog or FAQ pages with concise questions and answers near the top attract earlier highlights. By contrast, pages with large hero images or long‑form storytelling are more likely to push the cited text further down the page.

As for whether specific structural elements such as accordions or FAQs are more likely to be cited, the data suggests not. FAQs, for example, accounted for only 1.08% of the citations we looked at, and most of those were from the same website.

3. Descriptive subheadings matter

The data does reveal one consistent pattern: AI Mode tends to highlight subheadings followed by the first line of text.

This suggests AI Mode is using heading structures to navigate content and then sampling the opening sentences to assess how relevant it is to the query.

Again, this isn’t new, nor is it unique to AI search. Structuring your content with descriptive subheadings and clear opening sentences has long been good content practice, optimising for both human readers and search engine algorithms.

What does this mean for SEO and content teams?

There’s no such thing as an AI-optimised template or page layout

Page layouts and templates have no impact on AI Mode. This is even more true for LLMs, which strip away all HTML, CSS, Javascript, schema markup and formatting before processing the raw text.

The usual SEO fundamentals remain valuable for visibility in AI Mode

AI Mode relies on the same search infrastructure as traditional Google results. Rather than chasing hypothetical AI-specific optimisations, focus on the same E-E-A-T quality signals you always have.

Focus on quality content written for humans

It remains good practice to include a descriptive opening paragraph or summary near the top of your content because this is helpful for human readers. Your content can then explore the key points in greater depth within your content. Depending on the specific nature of the user’s query, AI Mode might determine that one of these more detailed points provides the best match.

This aligns with current advice around content chunking; breaking the information into clear self-contained sections of text.

Conclusion: the best strategy for AI Mode visibility

We may have only tested one aspect of page structure in relation to only one AI model, but the data clearly debunks the idea that where the information sits within a page has an impact on whether it will be cited.

There is no magic template or formula for increased visibility in AI Mode responses.

Instead, the best approach mirrors what’s worked in search for years: create well-structured, authoritative content that genuinely addresses the needs of your ideal customers.

The methodology

When citing content, instead of just providing a URL to the relevant web page, AI Mode will sometimes include a relevant snippet of text from the content. This snippet is accompanied by a text-fragment link, appending the URL with #:~:text=text fragment. Clicking this link takes you directly to the highlighted text within the page.

These text fragments would allow us to see if there is any discernible pattern to where AI Mode pulls information from within a web page.

From previous research, we already had data on 2,318 unique URLs cited in AI Mode responses, for valuable queries related to three industries: travel, e-commerce, and SaaS. Where the citations included a text-fragment link, we used a custom bookmarklet in desktop Chrome with a 1920×1080 viewport to record the vertical position in pixels of the first highlighted character on each page.

For each web page, we also captured information about the page layout or template used, as well as any elements such as hero sections, FAQs, accordions, or tables of contents. This would allow us to see if any of these structural elements were more likely to correlate with citations in AI Mode.

Need help surfacing your content in AI mode? Get in touch