Search behavior has changed more in the last two years than in the prior decade, and the biggest driver is generative AI answering questions directly inside the results page. Estimates on the impact vary, but a consistent range keeps showing up: Google AI Overviews are cutting organic click-through by somewhere between 18 and 47 percent for the queries they appear on.
For a B2B SaaS marketing team whose pipeline depends on organic search, that is not a rounding error. It is a reason to rebuild how you set traffic expectations and measure content ROI.
The mindset shift that actually matters
The old model was simple: rank well, earn the click, convert on your own site. The new reality is that a meaningful share of searches never produce a click at all, because the AI-generated answer satisfies the user’s question directly on the results page. Ranking well is no longer the goal, being the source the AI answer cites and trusts is.
This is the shift people are calling Generative Engine Optimization, or GEO, alongside the older term SEO. They are not competing disciplines, GEO is closer to a new layer on top of SEO fundamentals.
What to actually change in your content process
Stop measuring success by clicks alone. Track brand mentions and citations inside AI answers as a separate metric from organic traffic. A citation with no click still builds awareness and trust with a buyer who may search your brand name directly a week later.
Write the direct answer first. Content that leads with a clear, specific answer to the query before building out supporting context is more likely to get pulled into an AI Overview than content that saves the payoff for paragraph four.
Double down on content AI answers cannot easily replicate. Original research, proprietary data, customer case studies, and named expert perspective are harder for a generative answer to summarize away, because the value is in the source, not just the information.
Reset traffic-based goals with your team and leadership now. If your organic traffic dashboard shows a decline this quarter, the honest first question is not “what did we do wrong,” it is “how much of this is the AI Overview effect industry-wide.” Bring the benchmark data into that conversation before anyone assumes a content quality problem that may not exist.
The teams handling this well are not the ones panicking about a traffic dip, they are the ones who already redefined what success in organic content looks like before the dip showed up on a dashboard.
Why the range is so wide, and why that matters for your forecast
The 18-to-47-percent range is wide enough to be almost useless as a single planning number, and it’s worth understanding why the spread is so large rather than just picking the middle. The impact varies heavily by query type: informational queries with a clean, factual, single-answer response (definitions, comparisons, how-to questions) get pulled into AI Overviews far more consistently than queries with ambiguous intent or queries where the “answer” genuinely requires visiting a page — a pricing page, a product comparison with dynamic data, a tool you need to interact with. If a large share of your organic traffic comes from the second category, you should expect an impact closer to the low end of that range. If it comes from the first, expect closer to the high end. The useful exercise isn’t picking a number from someone else’s study — it’s auditing your own top organic landing pages by query type and estimating which bucket each one falls into, which gives you a forecast actually calibrated to your content mix rather than an industry-wide average.
Building the citation-tracking habit before you need it
The “track citations, not just clicks” guidance is right, but most teams don’t have a system for it yet, which means the recommendation goes unactioned. The lightest-weight version that actually gets built: pick your ten highest-value buyer questions, and once a month, manually run each through Google’s AI Overview, ChatGPT, and Perplexity, logging whether and how your brand is mentioned. It’s manual, it’s not glamorous, and it’s still more signal than most teams currently have — a rough, consistent time series beats a perfect measurement system that never gets implemented. Once you’ve run it for a quarter, you’ll have a real baseline to compare against when traffic dips, which turns “is this the AI Overview effect or a content problem” from a guess into an answer backed by your own data.