Introduction

Everyone wants to know what they can do to increase their brand’s visibility in AI-assisted search. So, the SALT team conducted research to see if traditional SEO ranking signals had any impact on brand mentions in AI responses.

AI search is disrupting traditional SEO. Success is no longer as simple as ranking on page one of Google. Brand visibility now means being included in Google’s AI Overviews, ChatGPT answers, or Perplexity citations.

While the opportunity is huge for any company able to get ahead of the competition in these spaces, the drivers that determine whether your brand appears in AI-generated answers remain unclear.

The conversation around traditional ranking signals has shifted. We need to understand which long-standing SEO principles still matter when it comes to appearing in AI-generated responses.

The question we wanted to answer was this:

Do traditional SEO signals and metrics, like backlinks and Domain Rating, still matter in the era of AI search?

Collecting the data

The sheer number of Large Language Model (LLM)-powered features and platforms out there would make an exhaustive study impossible. So, we chose to investigate visibility within four of the most popular:

  • Google AI Overviews (Google AIO)
  • ChatGPT
  • Perplexity
  • Gemini (previously Google Bard)

Meanwhile, to reduce any variables in content and search behaviour between industries that might skew our findings, we chose to focus on websites in the travel industry.

We collected a sample dataset of 5,825 URLs across multiple travel and content-heavy sites. For each URL, we gathered:

  • Traditional SEO metrics: Domain Rating (DR), number of referring domains, and total backlinks (from Ahrefs)
  • Search metrics: Estimated organic keyword and traffic data
  • Visibility markers: Mentions and citations in each of the four AI models.

We ran various statistical correlation tests to see if changes in one variable, such as more backlinks, coincided with changes in the AI visibility markers.

Then we used visual similarity modelling to compare the results. This would tell us if the impact on the AI visibility markers was consistent across all four AI models.

Key findings

AI Model Strongest Correlated Metric Correlation Strength
Google AI Overviews Domain Rating (DR) Weak (0.25) 
Chat GPT Backlinks Moderate (0.39) 
Perplexity Referring Domains Moderate (0.42) 
Gemini Referring Domains Moderate (0.41) 

Not all AI models care about backlinks

SEO and content teams need to rethink how they measure authority when optimising for AI visibility.

First, we wanted to understand whether Domain Rating and backlinks have any impact at all on AI search visibility.

We found a moderate correlation between backlink-based metrics and responses in ChatGPT, Perplexity and Gemini. These tools are more likely to surface pages that have higher Domain Ratings, more backlinks, and a greater number of referring domains.

But what about Google AI Overviews, with their coveted Position Zero in the SERPs? Perhaps surprisingly, we found little to no correlation with those same metrics.

In other words, Google’s AI model, Gemini, doesn’t rely on traditional link-based authority when choosing which pages to cite. This is a major change to how organic rankings have historically worked.

The impact of link-based authority varies between AI models

Each LLM or model uses a different weighting system, and marketers need to optimise accordingly.

Next, we used similarity modelling to compare how each of the AI tools responds to DR, backlinks and referring domains. Do they value link signals in a similar way?

ChatGPT, Perplexity and Gemini showed very high alignment, with similarity scores between 0.95 and 0.98. If a page performs well in one, it is likely to perform well in the others.

However, once again, we found that Google AI Overviews behave differently, with almost no alignment to the others in how it values link signals. This is only to be expected, given it is an extension of Google Search.

This suggests there is no one-size-fits-all approach to optimising for AI search. While three of the four AI models we tested showed similar results, we can’t say whether the same would be true for others, such as Claude or Bing Generative Search.

SEO teams may need to develop a tailored approach for each AI mode they hope to target.

Context and structure matter more than link signals

Format your content with easy-to-extract information that highlights key points and is scannable by users.

While we found a moderate correlation between link metrics and visibility in some AI models, they aren’t sufficient by themselves.

  • 18.5% of domains with a DR of 80 or higher fell into the lowest 25% of total AI citations.
  • No domain with a domain rating of 40 or lower appeared in the top 25% of domains based on total AI citations.
  • Structured, well-organised information with clear sections and descriptive headers appears to play a role in information retrieval, from the hundreds of thousands of LLM citations and responses we’ve analysed.

We recommend running regular audits of your visibility in AI-driven search experiences. These audits help you understand which of your pages are being surfaced and cited as trusted sources, and which ones are missing from the conversation altogether.

Research summary: out with templates, in with audiences

The old rules of SEO aren’t dead, but they aren’t enough either.

AI models clearly use different signals, relying on traditional authority metrics to varying degrees. The biggest takeaway?

You cannot rely on existing backlinks strategy to earn your spot in AI answers.

Your SEO strategy needs to evolve to include the following:

Key Takeaways

Backlinks still matter for some, but not all, AI models

While ChatGPT, Gemini and Perplexity still lean on link-based authority, Google AIO does not.

Google’s AI Overviews are fundamentally different

Pages that perform well in Google’s AI Overviews succeed based on clarity and user-first formatting, regardless of whether they have high DR or backlink volume.

Authority beats quantity

Many high-performing pages had just one or two backlinks from highly authoritative, topic-relevant domains.

It’s not about the number of links. It’s about the right publications linking to your content.

Cross-domain links are not valued the same

The AI models did not count backlinks from internal brand clusters (e.g. Kayak to Priceline), as “good examples” when selecting which pages to cite.

Informational content wins citations

Pages offering tips, safety guidance, government travel notices, or evergreen advice were far more likely to be cited, especially in AI and media coverage. Build for Asking.

Some pages with zero traffic are AI darlings

We found pages with fewer than 15 organic visits per month appearing in thousands of LLM citations. Visibility in AI isn’t just about traffic; it’s about trust and content structure.

Structured URLs dominate AI citations

Clean, semantic URLs (such as … /travel/en/advisories/europe/) were overrepresented in high-visibility AI citations, with 24% of the most cited domains following this format.
This may indicate AI models prefer content that is easier to parse, understand, and link to.

Marketers must optimise their content for each platform

You can’t take a blanket approach towards AI search. ChatGPT, Perplexity, and Gemini are separate AI Models with different data training sets and grounding methods.

On the other hand, Google’s AI Overviews and AI Mode are extensions of traditional Search, applying an LLM layer over the top of the same infrastructure.

As it currently stands, treat LLMs like ChatGPT, Gemini and Perplexity as one cluster, but develop a separate approach for Google AI Overviews and AI Mode, especially for transactional and informational pages.

Structure beats scale

Content with clear subheadings, direct answers, and semantic markup consistently outperformed less structured, wordier pages.

Run regular AI citation audits

Knowing where and why your content appears in AI responses gives you a competitive edge and helps you prioritise what to optimise next.

Contact us if you’d like more information about how SALT can help you with your SEO content strategy