*By Waleed Qamar | SEO By Highsoftware99*
—
Impressions were up thirty-eight percent year on year. Sessions were flat. The client had spent four months publishing twice a week, had finally sorted out their crawl budget issues, had a clean internal linking structure, and their pages were appearing in more searches than ever. But the people finding them were not arriving. We both stared at the Search Console graph for a long time before I said anything.
The Stanford study on AI search behavior puts a mechanism to what I had been watching in client data for months without a clean explanation. The headline figure, that users who switch to AI mode ask twice as many questions and click half as often, is being reported as a warning about declining traffic. That framing misses the more important half of what the data is describing.
Image credit: Screenshot from "Are we asking the right questions about artificial intelligence?" by Stanford Digital Economy Lab on YouTube (https://www.youtube.com/watch?v=YzKtlgJ96Wk).
The twice-as-many-questions finding is not just a curiosity. It means AI mode users are running longer research sessions inside the search interface before they decide to visit anywhere. A traditional search session for someone comparing project management software might look like: search, click, read, back, click, read, decide. An AI mode session for the same intent might look like: ask, follow-up, follow-up, compare, narrow, ask again, then click once with a reasonably formed opinion already in hand. The number of questions doubled because the conversation moved inside the interface. The click came later, from a different position in the decision process.
The practical consequence is that the user who arrives from an AI mode session is not the same user who arrived from a traditional organic click. They have already filtered. They have already eliminated some options. They have already formed a partial opinion that the AI shaped. When they do click, they are not starting their research. They are finishing it.
That distinction is not showing up in how most content strategies are being built. I have seen audits from content agencies that use query volume for informational keywords in AI search contexts as the primary justification for publishing volume. The logic looks sensible on the spreadsheet: more queries, more content, more potential traffic. The problem is that impressions in AI Overview contexts do not convert to sessions at the same rate as traditional ranked positions, and building a content operation around query volume without accounting for that conversion gap means building a library that gets cited and not clicked. The site gets visibility without traffic. The visibility is real but it does not pay the bills.
The conventional wisdom response to declining CTR is to publish more content targeting more queries, on the theory that broader coverage creates more entry points. That logic held in traditional search. It does not hold the same way in AI mode. More content targeting more questions does not automatically mean more sessions when the AI is handling the conversation and reserving the click for the moment the user is ready to act. A site that publishes fifty informational articles to capture fifty AI mode queries is not guaranteed fifty entry points. It might be building fifty citations that the AI references before sending the user somewhere else entirely.
What actually changes with this data is where on the decision journey a site needs to be credible. If AI mode users arrive later and more informed, the content that needs to perform is not the content at the top of the research chain. It is the content that handles the comparison stage, the objection stage, the specific "is this right for my situation" question a user asks when they are close to deciding. The client whose impressions were climbing while sessions flatlined had a content library built almost entirely for awareness-stage queries. The pages that would have caught an AI mode user who had already decided they needed a solution and was narrowing their options did not exist.
We built them. Not more volume, not broader coverage. Six pages targeting the specific questions a user asks when they are thirty minutes into an AI research session and getting close to a decision. The CTR on those pages, once they were indexed and appearing in the right context, ran higher than anything else on the site. Not because we had done something clever. Because we had stopped building for where the user used to enter and started building for where they actually are when they finally click.
The Stanford data is being used to argue that AI search is bad for publishers. That argument is incomplete. AI search is bad for publishers who built for the beginning of the user journey. The click did not disappear. It just moved.
Waleed Qamar holds a BSc in Computer Science from Purdue University and has spent the years since turning that technical foundation into something the curriculum never covered: figuring out why websites rank, why they fall, and why most businesses never find out until it is too late.
Pakistan-born and based between the United States and South Asia, he has managed search visibility for e-commerce stores, local service businesses, and SaaS startups across two continents. He started in SEO when guest posting still worked, survived the Penguin update, and has rebuilt client sites from scratch after algorithm hits more than once.
He has watched good businesses get sold packages that looked like progress and delivered nothing lasting. He has also seen the right approach quietly double a site’s traffic without a single press release about it.
His writing on SEO By Highsoftware99 covers Google algorithm updates, autocomplete optimization, semantic SEO structure, and the widening gap between what agencies promise and what Google actually rewards in 2026.
He knows what a traffic cliff looks like in Search Console on the morning you discover it.