Key takeaway: A lower search share does not prove fewer search visits.
The short answer
If you are trying to determine whether AI search is weakening website traffic, do not begin with search share alone. Chartbeat’s regional data shows that search accounted for 12% of pageviews at North American publishers in its network in the third quarter of 2026, down from 21% in the third quarter of 2025. That is a meaningful change in the composition of traffic, but it does not by itself prove that search sent fewer pageviews.
A share is a ratio. The numerator may fall, the denominator may grow, or both may happen at the same time. Search pageviews could decline while other channels grow faster. Search pageviews could remain stable while direct, social, referral, or other traffic expands. Total traffic could also fall, changing the interpretation of the percentage. Therefore, the 12% figure is a structural benchmark, not a measurement of absolute search visits or a forecast of future traffic.
Separate the three measurements
A publisher should track three related but distinct numbers by quarter. The first is pageviews referred by search engines. This shows whether search visits themselves are rising, stable, or falling. The second is total pageviews, which explains whether the denominator has changed. The third is pageviews from other sources, which can reveal whether another channel is replacing or outgrowing search.
A practical sequence is to compare search-referred pageviews first, total pageviews second, and the major non-search sources third. If search pageviews and total pageviews both decline, search may be part of a wider contraction. If search pageviews remain broadly stable while total traffic grows through other channels, the lower share is primarily a mix shift. If search pageviews increase while their share falls, it would be inaccurate to describe the result as a loss of search traffic.
The quarterly pattern in Chartbeat’s cited data also matters. North American search share was 16% in the fourth quarter of 2025, 13% in the second quarter of 2026, and 12% in the third quarter of 2026. A sequence can help a team monitor direction, but a regional benchmark should not be treated as a target, a prediction, or a direct proxy for any single publisher.
Treat AI as a hypothesis
Earlier Chartbeat reports connected the decline in search share with AI’s growing impact on search and content discovery. The cited third-quarter 2026 takeaway, however, does not establish a cause. Chartbeat also groups all search engines into one category in the regional chart. That means the data cannot show whether the change came from Google, Bing, another engine, or a combination of sources. It also does not isolate traffic from AI search products.
For editors and growth teams, AI can therefore be a hypothesis to test, not a completed attribution. Useful checks include separating search engines in analytics, identifying referring products where possible, examining landing paths, and comparing content types over the same periods. Without an identifiable source signal, a change in organic traffic should not automatically be labeled an AI-search effect.
The second report, about Eka Robotics, offers evidence from a different field rather than from publisher traffic. Eka’s system combines vision, force inputs, and action to improve robotic manipulation, including the handling of objects that can be damaged by excessive force. The report also says that the company still needs to prove the system on real industrial lines over sustained periods. This cannot validate a claim about AI search, and it is not a supplementary traffic sample. Its relevance is methodological: a demonstrated mechanism or impressive capability is not the same as a verified business outcome in a live environment. The same distinction applies when interpreting AI search.
Where the benchmark is useful
The regional figure is useful for publishers that already have consistent analytics and want to ask a narrow question: is the composition of their search traffic moving in a direction similar to a public benchmark for the same region and network? It is not suitable for predicting revenue, promising a traffic decline, or claiming that a particular AI feature has replaced search visits.
There are also important comparability limits. The cited report does not state how many websites were included, or whether the annual comparisons use exactly the same sites. That makes the scope of the year-over-year comparison uncertain. The report describes overall traffic to the network as steady, but it does not provide an overall pageview count in the cited takeaway. Readers therefore cannot use the percentage alone to calculate the absolute size of publisher traffic.
A reporting rule for content teams
When presenting the change internally, show search-referred pageviews, total pageviews, and search share together. Label the quarter, region, and measurement scope. Avoid writing, “Search share fell from 21% to 12%, so AI caused search traffic to fall.” A defensible formulation is narrower: the public regional benchmark shows that search represented a smaller portion of publisher pageviews, while the share measure alone cannot establish whether search pageviews decreased; AI remains one possible explanation.
The operational next step is not to chase an unspecified AI-search channel. It is to maintain quarterly source records and compare actual changes across content types and entry points. Only when a publisher sees a decline in search-referred pageviews and has identifiable evidence from an AI-related source is it reasonable to discuss an AI-search impact. Even then, the conclusion should remain limited to the site, period, sources, and measurement method that were actually observed.