What the studies measured
TESTED Two US studies are the best available evidence, and both are worth reading for their limits as much as their figures.
Pew Research Center analysed browsing data from 900 US adults for March 2025. 18% of Google searches produced an AI summary. Users who saw one clicked a traditional result link in 8% of visits, against 15% for users who did not, and clicked a link inside the summary in 1% of visits. Users were also more likely to end their browsing session after a page with a summary.
Ahrefs compared 300,000 informational keywords, half of which triggered an AI Overview in March 2025, using aggregated Search Console click-through rates for March 2024 and March 2025. Its estimate was that the presence of an Overview reduced the click-through rate of the top-ranking page by about 34.5% relative to the trend for informational keywords without one. Ahrefs notes 99.2% of Overview-triggering keywords in its data were informational.
Neither study is Australian, neither covers AI Mode, and neither says anything about a specific Perth site. What they establish is direction: when a summary answers the question on the page, fewer people click.
Which queries lose clicks first
INFERRED The unevenness follows from what a summary can do. Definitional queries (“what is entity SEO”, “how long does a settlement take”) are answered completely, and the informational skew in Ahrefs’ data is consistent with that. Simple factual queries behave the same. Comparison queries hold up better: a summary can list options but cannot do the evaluation a buyer weighing two specific choices needs. Transactional and local queries hold up best, because nothing gets booked, bought or enquired about inside a summary, and local results are shaped by the map pack and Business Profile as much as by generation.
Auditing your own exposure
DOCUMENTED Google reports AI Overview and AI Mode traffic inside Search Console’s Performance report under the Web search type, not separately, so exposure has to be estimated by intent rather than read off a filter.
Export the top hundred pages by organic clicks. Classify each page’s dominant query intent as definitional, factual, comparison, transactional or local. The definitional and factual pages are the exposed set. A site where most clicks come from explainers has a structural exposure no amount of optimisation removes; a site where the exposed set is small has a manageable one. The exercise usually produces an uncomfortable realisation: a decade of top-of-funnel content strategy depended on the click, and the click is the part being removed.
What to do about the exposed set
- Consolidate. Several thin definitional pages on adjacent topics become one substantial page that also answers the comparison and cost questions around them.
- Add the decision layer. A page explaining what something is gains sections on how to choose, what it costs and who it does not suit. Those are not substitutable by a summary.
- Retire, rarely. Pages that exist only to capture a query that no longer produces a click can be removed and redirected. Last resort.
What not to do is produce more definitional content faster.
What a citation delivers
In descending order of certainty: brand exposure in the context of an authoritative answer, largely unmeasurable per impression and visible indirectly in branded search; shortlist inclusion where the summary names providers, which arrives later as a branded search rather than a referral; and some direct clicks, at rates the studies above put well below a classic result. INFERRED The measurement consequence is that a working programme often shows flat referral numbers, rising branded search and better assisted-conversion presence. Reading only direct sessions from the surface will conclude nothing happened.
Reading traffic data without false conclusions
- Segment by intent before concluding. A decline concentrated in definitional pages is a different situation from one spread across commercial pages.
- Read impressions alongside clicks. Steady impressions with falling clicks suggests summary absorption. Falling impressions points at ranking or indexing, which is a different fix.
- Give it more than a fortnight. Overview presence for a query is rarely stable.
A halved informational traffic number alongside steady enquiries is the version people find hardest to read, and it is a reasonable outcome if the visitors lost were never going to enquire. The failure picture is different: total clicks down, commercial pages down, branded search flat. That points at an ordinary SEO problem.
Why no traffic outcome can be promised
DOCUMENTED Google decides whether to generate an Overview, whether a page is among its sources, and states that “indexing and serving isn’t guaranteed”. The user decides whether to click. A provider promising a traffic figure from this surface is promising all of those. What can be offered is work on the inputs, and reporting that says which numbers are observed and which are inferred.
What is documented and what is inferred
- Documented: Search Console reporting; Google’s no-guarantee caveat.
- Tested: the Pew and Ahrefs click figures, both US, both pre-AI Mode.
- Inferred: the intent ordering; the consolidation advice; the reading habits.
- Not established: any Australian or Perth click effect. No panel study exists for either.