AI Overviews and AI Mode: what they do to search and to PPC
A randomised field experiment from January and February 2026 measured about 38 % fewer outbound clicks on queries with an AI overview, and zero-click searches rising from 54 to 72 %. What that means for paid advertising — and what has not changed.
- A randomised experiment measured outbound clicks falling about 38 % on queries with an AI overview; zero-click searches rose from 54 to 72 %.
- The effect concentrates in overviews at the top of the page; those lower down barely changed clicks at all.
- For PPC this currently means relatively more, not less: paid positions remain while organic space shrinks.
Ever since Google started showing a generated summary above the results, its effect on traffic has been discussed more than anything else. The figures circulating range from “no change” to “the end of SEO”. The problem is that the vast majority of them are observations, not experiments — they compare queries with and without an overview, which mixes the effect of the overview with the effect of it appearing on a different kind of query.
Since the start of 2026 there is also a measurement without that problem.
What the randomised experiment showed
A field experiment run in January and February 2026, in which AI overviews were randomly removed for participants over two weeks, measured the following:
Two findings deserve emphasis, because they get lost in the debate:
- Position matters. Removing a top-position AI overview nearly doubled outbound clicks. Overviews lower on the page had no measurable effect. “We have an AI overview” is therefore not a diagnosis in itself.
- Satisfaction did not change. When the overview disappeared, people did not report a worse or better experience. That is an uncomfortable finding for both sides of the argument.
How it looks in observational studies
For context, because you will meet these numbers everywhere:
| Source | Finding | What limits it |
|---|---|---|
| Field experiment (Jan–Feb 2026) | −38 % outbound clicks, zero-click 54 → 72 % | The only causal method here; two weeks per participant |
| Pew Research | Clicks in 8 % of cases with an overview vs. 15 % without | Observation, not an experiment |
| Ahrefs | −58 % CTR for top-ranking pages | Correlation; single tool’s data |
| Sistrix (Germany, Feb 2026) | Position 1 CTR falls 27 % → 11 %, aggregate loss about 6.6 % of clicks | German market; the gap between per-query and whole-site impact is crucial |
| Seer Interactive | CTR drop 1.76 % → 0.61 %, later a partial recovery | Agency data, moves over time |
“−38 % clicks on queries with an overview” and “−6.6 % of clicks in total” are not in conflict. The first applies to affected queries, the second to a whole site across all queries. Which one concerns you depends on how many of your queries trigger an overview at all — and those are typically informational queries, not transactional ones.
What it means for PPC
Here it is less dramatic than the debate suggests, and in the short term rather favourable:
- Paid positions have not gone away. Google shows ads above the overview, below it and inside it. When organic space shrinks, the relative value of a paid position rises.
- The query mix shifts. Informational queries (“how to choose”, “what does X mean”) can be answered in the overview and no click follows. Transactional queries (“buy”, “price”, “near me”) behave differently, because the person needs to get somewhere. Those are the core of PPC.
- Pressure on budgets rises. If you lose informational traffic from organic, companies often replace it with paid. That increases competition in the auction and therefore the cost per click — not because of AI as such, but because of how advertisers react.
- Measurement gets worse. A customer journey through an AI assistant often hides under direct traffic in analytics. More on that in traffic from AI assistants.
What to do about it in practice
- Split your queries by intent. Informational separately, transactional separately — in campaigns and in the report. Only then will you see where traffic is really falling and whether those are even the queries you earn on.
- Check in Search Console how many of your queries trigger an overview. The difference between “this affects 5 % of our queries” and “60 %” is the difference between a nuisance and a strategic problem.
- Do not shift budget in a panic. The same arithmetic applies as always: what decides is marginal CPA against margin, not that clicks fell somewhere.
- Invest in the queries where the purchase is decided. An overview will not resolve those, because the person needs a supplier, not a definition.
- Watch where ads already live inside AI. Formats inside AI answers are a separate topic — see paid advertising in AI search and assistants.
What we do not yet know
An honest list of open questions, because this topic attracts a lot of false certainty:
- How the quality of the remaining traffic behaves. Fewer clicks may also mean better-decided visitors.
- How it plays out long term. All the data comes from a period when the format was still changing every few months.
- What the effect is on non-English markets. Most studies come from English-speaking markets; the closest is the German Sistrix data.
- How the ratio of ads inside the overview to ads above it will develop.
What to take away
AI overviews do take clicks, demonstrably and measurably, but unevenly: most from informational queries and from overviews in the first position. For PPC that currently means rising rather than falling importance — with one caveat, that higher competition in the auction eventually shows up in the cost per click. The sensible reaction is neither panic nor denial, but splitting queries by intent and checking how much of this concerns you at all.
Common questions
How much do AI Overviews reduce clicks?
A randomised field experiment from January and February 2026 measured outbound clicks falling by about 38 per cent on queries where an AI overview appeared, and the share of zero-click searches rising from 54 to 72 per cent. Observational studies report higher figures, commonly between 34 and 61 per cent CTR decline, but they cannot separate the effect of the overview from the effect of it appearing on a different kind of query. It also matters whether you look at the impact on an affected query or on a whole site: the latter is considerably smaller.
Do AI Overviews threaten paid advertising?
So far rather the opposite. Ads appear above the overview, below it and inside it, so when organic space shrinks the relative value of a paid position rises. The indirect risk lies elsewhere: companies that lost organic traffic often replace it with paid, which increases competition in the auction and with it the cost per click. The second risk is degraded measurement, because a journey through an AI assistant often hides under direct traffic in analytics.
Which queries are most affected by AI Overviews?
Informational queries of the “how to choose”, “what does it mean” or “what is the difference between” kind — there the overview can supply the answer outright and no click follows. Transactional queries, where someone is after a specific product, price, supplier or location, are affected considerably less, because the user needs to get somewhere. Those queries are the core of performance advertising, which is why the impact on PPC is smaller than the impact on content sites.
- Study Confirms Google AI Overviews Cut Organic Clicks 38%, Search Engine JournalReport on the randomised field experiment of January and February 2026 (clicks −38 %, zero-click 54 → 72 %, overview on 42 % of queries, effect concentrated in the top position). The only causal method among the sources cited here; two weeks per participant.
- Google AI Overviews Statistics: CTR & Traffic Impact, SlateOverview of observational studies including Pew Research (8 % vs. 15 %), Ahrefs (−58 % CTR) and Seer Interactive. These are correlations, not causal measurements, and the sources disagree with each other.
- Google AI Overviews & Organic CTR in 2026: Data + Strategy, Relevant AudienceSource of the Sistrix figures from the German market (position 1 CTR 27 % → 11 %, aggregate loss about 6.6 % of clicks). A vendor blog; we did not verify the original Sistrix data.