A low AI Overviews click-through rate does not, by itself, settle the traffic question. For agencies, the harder issue is what the opt-out evidence actually isolates, and what it leaves mixed together.
Search visibility, on-page intent, brand demand, and result-page behavior can all shift the final click pattern. That makes the current signal useful, but narrower than many quick takes suggest. The practical takeaway is not panic or dismissal, but a more careful read of what changed and where.
What the opt-out study actually measured, and what it did not
Two boundaries matter most: what changed, what was counted, and what stayed separate. As dsgnbyhl explains, Google’s Search Console setting excludes a site’s links and content from AI Overviews, AI Mode, and generative AI features in Discover, while ordinary Search can still remain available.
That means an opt-out test is not a clean measure of “AI search” as one single surface. It is also not a measure of model training, because Google treats search visibility and AI training as separate choices.
Just as important, the performance report tracks link impressions for AI Overviews and AI Mode, with cuts by page, country, device, and date. So the core reading is narrow: it reflects reported search-surface visibility and clicks, not every way AI may affect discovery or demand.
Why AI Overview impressions can produce a much lower CTR than classic organic results
Three forces usually drive the drop. When Google places an AI Overview at the top of the results page, many searchers get a usable summary before any site visit happens, so an impression is counted without the old click pattern that classic blue links relied on.
Bennett Uwagwu wrote on LinkedIn that AI Overviews tend to redistribute traffic rather than erase it, which fits this logic: broad informational queries are easier to satisfy inside the search results.
The format also favors extractable answers. Uwagwu argues that models pull distinct entities, relationships, and concise facts, and may skip long preambles or fluffy copy. That means impression volume can rise while clicks lag.
For agencies, the key shift is measuring which queries still require a visit, not treating all impressions as equal traffic opportunities.
Does a 0.9% CTR mean AI Overviews hurt traffic, or just change where clicks happen?
Because Google was exploring opt-outs for AI Overviews and AI Mode, as SEO Masters Academy noted in a Facebook post, the practical question is not whether every impression should have clicked like a classic result.
It is whether the same query still sends visits through another path, later in the session or on a different result. A very low click-through rate can signal substitution inside Google’s interface, but it can also coexist with traffic on higher-intent or branded searches.
That makes the number directional, not final. Agencies should read it as evidence of changed click distribution first, then test whether total qualified visits actually fell. That distinction matters when reporting impact to clients and when choosing whether to optimize for visibility, deeper consideration, or direct conversions.
The counterintuitive opt-in finding, and the explanations that best fit the data
One result needs a different lens. If an opt-in page outperformed an opted-out page, the cleanest explanation is exposure, not magic. A Google Search Community thread titled “I opted out of the generative AI features in google but am still getting AI overviews in search” points to an important boundary.
Opting out does not necessarily make AI Overviews disappear from the results page itself. That helps explain the surprise. A page that remains eligible to be cited inside those features may gain another route to discovery, even when many impressions never become clicks.
The finding also fits mixed search behavior. Some users take the summary and leave, while others click the cited source that looks most relevant or trustworthy. For agencies, the practical read is simple: eligibility may still matter, even when click-through looks weak.
How query type, brand strength, and ranking position can distort the apparent impact
Three variables can bend the read. Informational queries are easiest to satisfy inside an AI answer, so they can look worse than navigational or high-intent searches. Strong brands can also hold clicks better, because users may still seek a known source even after reading a summary.
Ranking position matters too. A citation or top organic result can capture the limited clicks that remain, while lower placements lose out fast. That is why one blended CTR can hide very different query realities.
As SEOSherpa reported in 2026, Google’s explored opt-out mechanism emerged from UK Competition and Markets Authority discussions and reflects a real control-versus-visibility tradeoff. For agencies, the practical move is segmenting by query class, branded demand, and position before treating any AI Overviews click-through rate shift as a uniform loss.
What agencies should check before blaming AI Overviews for a traffic drop
Start with the simpler checks. A Google Search Community thread titled “I opted out of the generative AI features in google but am still getting AI overviews in search” is a useful reminder that surface behavior can stay messy even after a control change.
Before pinning a drop on AI Overviews, agencies should verify whether demand fell, rankings slipped, pages changed, tracking broke, or Google introduced other result features on the same queries. It also helps to separate homepage, blog, and product-page losses, because the cause may sit in content fit rather than in AI visibility alone.
The practical implication is restraint: treat AI Overviews as one possible contributor, then rule out ordinary search and measurement issues before making a client-facing call.
Where the evidence is still thin, including scope, sample, and causation limits
Three limits matter most. First, AI Overviews are a generated synthesis, not a direct ranking signal or a validated recommendation. As iatroX notes in its medical-search analysis, Google can cite several sources and still produce a summary that is wrong, incomplete, or poorly fitted to the context.
That matters for click-through interpretation because a cited link is not proof that the surface itself worked cleanly. Second, the evidence is narrow. One opt-out case can show what happened in that setup, but not across industries, query mixes, or user intents.
Third, causation stays weak. If Google has removed some medical Overviews after misleading outputs, then observed traffic shifts may reflect answer quality, query type, or trust, not one stable AI Overview effect.
That makes broad agency claims premature.
Taken together, the opt-out evidence points to a narrower answer. AI Overviews can depress click-through rate without proving a uniform traffic loss, because the measured change sits inside Google’s search surfaces and can reflect redistributed clicks, query intent, brand demand, and ranking position.
The strongest practical reading is cautious, not absolute. Agencies should treat a very low CTR as a signal to segment performance first, then judge impact by qualified visits and page type instead of assuming every lost impression would once have produced a comparable organic click.