Last week, Cyrus Shepard published a study on AI citation features, and it caused quite a stir on X, LinkedIn, and a number of private WhatsApp groups I’m in. It is not only the difference between what is a feature, and what is a correlation, especially considering that many subjects in SEO and AI are diverse and have a high level of incommensurable complexity. To be clear, this is not a criticism of Cyrus’ work; The study is excellent, and the correlation/causation caveat is one that he clearly makes.
This made me think in parallel with other factor studies done before, which revealed that direct traffic is a common factor for SEO ranking. At the time, these studies received a lot of negative feedback, and this was discussed again by many on the Internet after the documents in the Google search of the DOJ revealed a “popular” signal.
It makes sense that direct traffic is part of how popularity is measured with Chrome. Google uses Chrome data to find new websites. It also judges the “quality” of a page based on how users interact with it after clicking, but the atomic levels of how this is done, and how much weight the variables carry here, are not public knowledge.
Direct Traffic x Popularity Correlation
Direct traffic is widely considered a sign of efficiency, not a primary driver of search ranking.
Treating direct traffic as a ranking factor leads to a loop of misinformation, encouraging external tactics, less effective, such as buying bot traffic, in a wrong attempt to increase popularity, as it is more likely to have high levels of direct traffic and poor SEO performance.
A broader perspective suggests that high direct traffic is often an indicator of a strong brand, coupled with real ranking factors such as multiple product searches, high-quality backlinks, and strong social connections.
These elements are the real causes of high positions; direct traffic simply serves as a measured measure of a brand’s health and success, the result being “all ships rise in great waves”.
If Chrome’s data was a specific feature, a sudden spike in browser activity on a particular URL would push it up the SERPs, and this would be a playable exploit.
This would also be something that Google would choose as it looks to eliminate the obvious manipulation of search rankings, and this would have happened many years ago.
Some Insights From DOJ Files
NavBoost and Glue are specialized systems within Google’s infrastructure that focus on user interaction signals rather than raw volume of direct traffic.
NavBoost looks at historical click data and user behavior in search results to see which pages are most relevant to specific queries, effectively acting as a memory of what users have found useful.
While NavBoost focuses on traditional organic results, Glue extends those same principles of user interaction to all other SERP elements: information panels, video carousels, image packs, and featured snippets.
They allow Google to measure a site’s authority based on how users interact with it in the search ecosystem, regardless of the source of the user’s traffic.
→ Read more: What the Google Antitrust Verdict Says About the Future of SEO
So, What Thunder?
Based on what we know from various official (and unofficial) sources, research, and the common sense of the SEO hive, we can define popularity as a sign of brand strength that is reflected in user behavior such as auto-complete and bookmarks.
It works as a top ranking link because it naturally matches the various signals that make up a page’s ranking.
Google may avoid using Chrome data directly as a ranking feature, opting to use it as a dataset to train or validate its AI models. We don’t know this, and we probably won’t be able to prove or disprove it through research.
Thanks to Ryan Jones, Mark Williams-Cook, Chris Green, Gerry White, Kristine Schinger, Charlie Whitworth, Emina Demiri Watson, (and anyone else I remembered) for the weekend’s lively discussions on this topic.
Additional resources:
Featured Image: PerfectWave/Shutterstock
