If your organic traffic is falling while your rankings look stable, AI Overviews may be one factor, and your analytics alone cannot prove it. A Pew Research Center study shows that click behavior differs when an AI-generated summary appears. Google’s reporting, meanwhile, tells you whether your pages showed up in those features, not whether anyone opened them.
Our article on the future of websites in the age of AI introduced the larger shift: a website can influence a decision without receiving a visit. This piece stays on the harder follow-up, which is what the available data actually measures.

What the Pew Study Found
Pew Research Center analyzed the browsing of 900 U.S. adults in March 2025. The dataset held 68,879 unique Google searches. Pew later re-ran those queries and found an AI summary on 12,593 of them.
The difference was substantial:
- When an AI summary appeared, users clicked a traditional search result in 8% of visits.
- Without an AI summary, they clicked a traditional result in 15% of visits.
- Users clicked a source link inside the AI summary in 1% of visits to pages that contained one.
Those figures describe what Pew observed. They do not prove that an AI summary caused every difference. The queries that trigger summaries may differ from those that do not, and the study reflects one group of U.S. adults during one month. Pew also reconstructed the results pages after the browsing period, and an AI summary can change over time. Google disputed the study’s methodology and said its sample of queries was not representative of Search traffic.
The careful reading is not that AI Overviews cut every site’s traffic by a fixed percentage. In this dataset, result pages with AI summaries were associated with substantially fewer outbound clicks.
What Google’s Report Adds
In June 2026, Google announced a dedicated Generative AI performance report in Search Console. It shows how often URLs from a site appeared in AI Overviews and AI Mode, with breakdowns by page, country, device, and date. Google says that view reached websites worldwide on August 31, 2026.
The report is useful, and it is still incomplete. It reports impressions, not clicks, click-through rate, or queries. An impression means a URL from your site appeared in a supported generative feature under Google’s counting rules. It doesn’t tell you whether the user opened the link, read your content, remembered your brand, or took action later.
A property can still be missing the report when it has too few impressions in those features, or when the site is excluded from them. Not seeing the report is not proof that your pages never appear in an AI result.
Build a Measurement Baseline
No single metric replaces organic clicks. Use the signals you do have together:
- Record AI impressions by page. Note which URLs appear and how that visibility changes.
- Compare impressions with organic performance. Look for pages whose traditional impressions or rankings stay stable while clicks decline. That pattern is worth a look. It is not proof of cause.
- Separate branded and non-branded demand where your tools allow it. A change in searches for your company name is context. It is not, by itself, evidence of an AI citation.
- Track business outcomes. Leads, sales, bookings, calls, and qualified inquiries matter more than visibility alone.
- Annotate major changes. Record content updates, seasonality, tracking changes, algorithm updates, and the date the AI report showed up. Otherwise, unrelated events can look like one trend.
The Bottom Line
Pew’s study measures observed click behavior across thousands of searches. Search Console measures how often your own URLs appear in Google’s generative features. Neither one, on its own, measures the path from visibility to a business result. Being named in an answer is also not the same thing as earning the visit, which is the practical limit in what generative engine optimization changes for your site.
Treat AI impressions as a new diagnostic signal, not a replacement for clicks or conversions. The goal is a clearer baseline now, so you can recognize a real change as the reporting improves.