Google launched AI overviews in the US in May 2024. Within 90 days, publishers in tested categories reported organic traffic drops of 15-60% on informational queries. Those numbers are not equally distributed. Sites built on original research, case studies, and experience-based content held positions and traffic. Sites built on summarizing existing knowledge lost both. The algorithm did not punish SEO broadly. It punished a specific type of SEO content that was always the weakest form of the strategy.
The structural change to the search results page is the one that matters most. An AI-generated answer now sits between the search bar and the first organic result on a large share of informational queries. That answer pulls from existing sources, often including pages that rank well. Users who get a sufficient answer from the AI overview do not click any of the organic results beneath it. The ranked pages earn the citation but lose the visit.
Which queries took the traffic hit
Not every query triggers an AI overview. Transactional queries returned traditional results before the update and still do. “Buy X,” “price of Y,” “book Z,” “download A” look the same as they did in 2022. Local queries return maps and listings. Brand queries return brand pages.
The queries hit hardest are informational: “how does X work,” “what is Y,” “why does Z happen,” “difference between A and B.” These are the queries that most content marketers target because they carry high search volume and reasonable competition levels. They also carry the lowest purchase intent. A user asking “what is programmatic advertising” is learning, not buying. AI overviews absorbed those queries because they are the exact use case the feature was built for: synthesizing factual answers from existing sources.
The implication for content strategy: keyword research tools still show high search volume for informational queries. That volume is now an overestimate of available clicks. The gap between search volume and actual click opportunity keeps widening. Build content strategy on click-through data from Google Search Console, not on keyword volume alone.
The content floor that moved permanently
Before 2024, a thorough 1,500-word article answering a question clearly could hold a ranking position for years. Google’s AI absorbed those articles. They may still rank in position 4, but the click-through rate dropped because the AI overview above them answered the question before the user reached the organic results.
The content still driving clicks in 2025 contains something the AI cannot generate from sources that already exist on the web: original claims. Your own survey data with numbers from your customer base. Client case study results with specific figures that appear on your site and nowhere else. A methodology you developed from running campaigns directly. A position that contradicts the standard industry advice and supports the contradiction with evidence from your own experience.
Generic SEO content, summaries of what industry publications say, thin how-to guides assembled from secondary sources, fills AI overviews without generating the clicks that would justify the content investment. You produce the content, Google cites it in the overview, and the traffic goes to the AI-generated synthesis rather than to your page.
Zero-click growth and what it means for keyword strategy
Zero-click searches grew from roughly 50% in 2019 to an estimated 65% in 2024. Semrush and SparkToro have tracked this trend across multiple analyses. The growth accelerates as AI overviews expand across query categories. The keyword volume numbers in Ahrefs, Semrush, and Google’s Keyword Planner do not adjust for zero-click rates. A keyword showing 10,000 monthly searches may generate 3,500 actual clicks, not 10,000.
Keyword categories that still generate clicks at strong rates: product comparison queries where users need to evaluate options an AI cannot resolve for them, brand-specific queries where users want to go to a specific site, local intent queries where Maps and Business Profile results dominate, and original research queries where no existing source covers the specific data the user needs.
How to audit your own traffic for AI overview impact
Pull Google Search Console data for the period before AI overviews launched (pre-May 2024) and compare to Q3 and Q4 2024. Sort your pages by the change in click-through rate, holding position constant. Pages where position held but CTR dropped are the ones with AI overviews above them absorbing clicks. Pages where both position and CTR declined had authority problems, not AI absorption problems. The two failure modes look identical in a blended traffic report but require different fixes.
For pages with declining CTR despite stable position: check whether an AI overview appears for those queries. If it does, the content type is the problem. That topic now feeds overviews rather than earning clicks. You have two choices: build content that makes a specific claim no other source makes (so the AI must cite your page to include the claim), or pivot the keyword target toward a buying-intent variant where AI overviews appear less frequently.
The supply-side shift that compounds the problem
AI changed how much content gets produced alongside changing how traffic flows. Claude, ChatGPT, and Gemini let teams produce 10 times the content volume with the same headcount. Search results flooded with mediocre AI-generated pages starting in late 2022. Google’s Helpful Content Updates targeted this supply explosion starting in late 2023 and escalated through 2024. Pages with thin, undifferentiated content lost positions to pages with original perspective and verifiable expertise.
The two shifts interact: AI overviews reduced clicks on generic content while Google’s quality updates reduced rankings for generic content. Sites built on experience-based writing with original claims got pushed up in rankings at the same time their category of content retained higher click-through rates. Sites built on summarized, researched content lost both rankings and clicks simultaneously.
The practical strategy: produce less content and invest more per piece. A single well-researched, experience-based article targeting a buying-intent query outperforms 10 summarized articles targeting informational queries in 2025. The economics of content quantity shifted dramatically.
The sites that gained organic traffic through the 2024 algorithm changes share a production model: they publish four to eight pieces per month at 1,500-3,000 words each, with original data or direct experience embedded in each piece, rather than producing 20-40 shorter summaries. Their authors are named, their credentials are visible, and their content makes specific claims that other sources do not. That model requires more production investment per piece and produces fewer pieces per month than the quantity-first approach. The tradeoff is that each piece earns more traffic, more citations in AI answers, and more backlinks from editors who find it worth referencing, because it contains something worth referencing. The quantity-first model produces a content library that gets absorbed into AI overviews. The quality-first model builds a content library that earns citations and compounds authority over time. The inflection point arrives when your topic cluster has enough original, experience-based content that AI models treat your site as the reference source for that niche rather than one source among many. Converting more of the traffic that does arrive matters more as organic click volume gets harder to generate. GEO as a parallel visibility channel is worth building alongside a reformed content approach because the two visibility channels reward the same underlying content quality.