What AI Overviews did to organic traffic, six months in.

Google's AI Overviews went US-wide on 14 May 2024. By November they were appearing on a majority of commercial and informational queries in the US, UK, and Canada, with material rollout across ten more markets. We have been tracking query-level click-through on the same 2,200 queries across 19 client domains since April. Here is what the data actually says after six months of data, not what the trade press keeps saying.

1. The click loss is real but uneven.

Across those 19 domains, query-level click-through rate on pages ranking 1 to 3 dropped a median 29% between April 2024 and October 2024 on queries where an AIO was present. Queries without an AIO did not move. This tracks within a few points with Ahrefs' September 2024 study, which showed a 34.5% drop for position-one results on AIO-affected queries.

2. Informational queries were hit much harder than commercial.

"How do I..." and "what is..." queries lost a median 46% of clicks. Queries with clear purchase intent (brand plus product, "buy", "pricing") lost a median 7%. The spread is wide inside each category; the median is a rough guide, not a rule. Local queries ("near me") and transactional queries ("book", "reserve") were almost unaffected because AIO appears on those much less frequently.

3. The citation stack inside AIOs is narrow.

Across 3,100 sampled AIOs between June and October 2024, cited sources skewed toward a small set of domains: Wikipedia, Reddit, official product documentation, and a handful of heavy editorial sites (NYT, Wirecutter, Forbes Advisor for consumer finance, Healthline for health). For most B2B categories, the top three cited sources inside AIOs were the official product docs, one G2-style review site, and one vertical publication.

If you are not in this citation stack, the AIO is costing you clicks without giving you visibility. This is the worst position to be in.

4. Reddit is surfacing at a rate that surprised us.

Reddit appeared as a cited source in 23% of results in our 2024 sample. The pre-deal baseline in early 2024 was 9%. Google's content-licensing deal with Reddit, publicly reported at around $60M per year, was announced in February 2024, and the citation frequency scaled with its rollout. For consumer and DTC categories this ratio is even higher.

5. Clicks that do come through convert better, not worse.

Session-level conversion rate from AIO-impacted queries (where the user clicked past the overview) ran 16 to 22% higher than the same queries before AIO. This is a selection effect: users who click past an AIO are further along the buying cycle, or have a harder question than the AIO answered.

The tempting conclusion is "volume is down but quality is up, net-neutral". The data does not support that. The revenue math is almost always negative once you account for the volume loss, even at the higher conversion rate.

6. Brand queries are an anomaly.

Branded search click-through has been almost unaffected. AIOs fire on brand terms only about 9% of the time, and when they do, the top organic result almost always gets the click. This is the shape we see protected across our sample.

7. AIO presence is correlated with helpful-content update damage

Pages that took heavy losses in Google's September 2023 and March 2024 Helpful Content updates are over-represented in AIO-impacted queries. That is, AIO hits hardest on queries where Google is already unsure which result is authoritative. Pages that survived the HCU well are under-represented in AIO-citation loss.

What to actually do about it

• Rewrite informational content around "answer" shape. One question per page. Canonical answer in the first 80 to 120 words. Supporting detail below. This is the format AIO lifts from, and it is not the format most SEO content was written in.

• Ship FAQPage and HowTo structured data. Not a magic bullet. Helps the content be quotable.

• Monitor query-level AIO presence monthly. Tools: ZipTie.dev, AlsoAsked, Semrush's AI Overviews report, SE Ranking. The rollout is not static. Categories where AIOs are infrequent today will likely have them in six months.

• Stop conflating AIO inclusion with LLM answer inclusion. Google AIO and ChatGPT/Perplexity/Gemini-app answers are different distributions cited from different sources. An LLM-SEO strategy and a Google-AIO strategy are two projects, not one.

• Audit your presence inside the narrow citation stack. If your category's AIOs cite G2, Wikipedia, Reddit, and your competitor's docs, fix the order of investment.

What is overrated

• Trying to "optimise to be cited inside AIO". You cannot reliably. You can be cited more often by being good content, but the deterministic lever does not exist.

• Schema markup as the dominant fix. FAQPage helps. Article schema helps. The biggest lever by far is content shape.

• The assumption that AIO will be rolled back or softened. Google has doubled down repeatedly through 2024.

Sources: internal query-level tracking across 19 domains, April to October 2024; Ahrefs September 2024 study; SparkToro 2024 Zero-Click Study; Reddit/Google licensing deal coverage.