Google's AI Overviews now appear on roughly 15-25% of all queries, and as of December 2025 they've cut click-through rates for #1 organic positions by 58%. Zero-click searches keep climbing.
Around 60% of Google searches now end without a click, and when an AI Overview is present that jumps to roughly 83%. But there's a split happening that most teams aren't watching closely enough: pages that get cited inside an AI Overview see meaningfully higher organic CTR, around 35%, compared to pages that don't.
The game didn't end. It just got a different scoring system.
What actually happened to organic traffic in 2025?
The damage was real. Global publishers lost roughly a third of their Google traffic. CNN dropped 27-38%.
Business Insider and HuffPost fell about 40%. A Semrush 10-million-keyword study found that informational query coverage by AI Overviews dropped from 91% in January 2025 to 57% by October, meaning Google pulled back on some queries. But that pullback didn't undo the structural shift.
The informational queries that still trigger AI Overviews are the ones with the highest volume. And those queries now send less traffic to websites because Google answers them directly.
Here's what the traffic impact looks like by query type:
| Query type | AIO frequency | CTR impact on #1 result | Citation opportunity |
|---|---|---|---|
| Informational | High (57-91% depending on month) | -58% average | High, if you're the cited source |
| Commercial investigation | Medium | -30 to -45% | Medium |
| Transactional | Low | Minimal change | Low |
| Navigational | Very low | No meaningful change | N/A |
The pattern is clear. If your traffic strategy depended on ranking #1 for "what is [topic]" queries, you've already felt this.
Does schema markup actually help you get cited?
This is where the industry is overselling. Only a sliver of the sources cited in AI Overviews lean on FAQPage schema, and Reddit, one of the most-cited domains in AI results, uses almost no Schema.org markup at all.
Schema helps Google understand your content structure. That's real, and clean, well-structured pages with sensible schema and a tidy H1-H3 hierarchy do get pulled into AI answers more often. But schema isn't the magic switch agencies are selling it as.
Here's the more interesting finding: JavaScript-injected schema is completely invisible to GPTBot, ClaudeBot, and PerplexityBot. Those crawlers don't execute JavaScript. If your schema is only rendered client-side (common with React apps and some WordPress plugins), it doesn't exist as far as AI systems are concerned.
Editor's Note: Server-side rendered schema works. JS-injected schema doesn't. That distinction alone is worth more than most "AI SEO audits" will tell you.
What does an AI Overview actually want to cite?
The Princeton GEO (generative engine optimization) study tested which content attributes correlate with getting cited by AI. The biggest levers were citing authoritative sources, adding specific statistics, and quoting experts directly.
Together those methods lifted visibility inside generative engines by up to 40%, and lower-authority domains gained as much as 115%. That last part matters if you aren't a household name yet.
Short, self-contained passages that fully answer a query outperform long ones. The best-performing content isn't a 3,000-word pillar page with the answer buried in paragraph twelve. It's a page where the answer sits at the top, complete and specific, with the supporting evidence right there in the same passage.
Multimodal content, meaning text paired with original images, video, and structured data, gets pulled into AI answers far more often than text alone. Video sources like YouTube are now among the most-cited in AI Overviews. If you're only producing text, you're competing with one hand behind your back.
Is generative engine optimization different from regular SEO?
Yes and no. The technical fundamentals overlap, but the priorities are different.
Regular SEO rewards comprehensive pages. GEO rewards self-contained, directly answering passages. Regular SEO rewards backlink authority. GEO rewards source citation and factual density. Regular SEO lets you bury the answer below a 400-word introduction. GEO punishes that.
The practical checklist:
- Write tight TL;DR passages at the top of every section that answers a query, a focused paragraph, not a wind-up. Don't tease. Answer immediately.
- Cite your sources in the same sentence as your claim. Not in a bibliography at the bottom.
- Include specific numbers. "Revenue grew 43%" beats "revenue grew significantly."
- Use named entities. "Google's March 2026 core update" not "the latest algorithm change."
- Add original visual content. A custom chart or explainer video puts you in a different tier than text-only competitors.
- Render schema server-side. Test by disabling JavaScript and checking if your structured data still appears in the page source.
- Build clean heading hierarchies. H1 for the page topic, H2s as questions people search, H3s as specific sub-answers.
What about AI Mode and agentic search?
Google's AI Mode has expanded to 180+ countries and territories in English, and agentic booking has rolled out worldwide. It started with restaurant reservations in the US in August 2025, and AI Mode can now check real-time availability and complete the booking inside search.
This isn't a separate product. It's the next step in the same direction, the same push that's reshaping local search too.
For marketers, agentic search means Google doesn't just answer the question. It completes the task. If someone searches "book a haircut near me Saturday morning," Google's AI can now check availability and make the reservation without the user ever visiting a website.
The businesses that benefit are the ones with structured data Google can act on. The businesses that lose are the ones where the website was the funnel. When the funnel gets bypassed, the traffic numbers become irrelevant.
How should content teams restructure for 2026?
Stop measuring success by ranking position alone. A #1 ranking with an AI Overview above it is worth dramatically less than it was two years ago. Start tracking:
- 1AIO citation rate across your target queries
- 2Branded search volume (if people search your name directly, AIO can't steal that)
- 3Click-through rate per query, not just position
- 4Multimodal content coverage for your top 50 keywords
The teams winning right now aren't doing anything exotic. They're writing specific, well-sourced content in tight passages. They're producing video. They're rendering their structured data server-side. And they're tracking the AI referral traffic that standard analytics quietly miss.
The teams losing are the ones still writing 2,500-word blog posts with the answer in paragraph eight, checking their ranking position, and wondering why traffic dropped 40%.
This is the work we do for clients at Sparksbox: tight, well-sourced content, server-side schema, and measurement built around citations, not just rankings.
FAQ
AI Overviews appear on approximately 15-25% of Google queries as of early 2026. Informational queries see the highest frequency, though Google reduced informational AIO coverage from 91% to 57% between January and October 2025 based on a Semrush study of 10 million keywords.
No. Reddit is one of the top-cited sources and uses almost no Schema.org markup at all. Schema helps Google understand your content and well-structured pages do get cited more, but it's one signal among many. Content quality and source authority matter more.
Short, self-contained passages that directly answer a query consistently outperform longer sections. This doesn't mean your overall page should be short, just that each answerable section needs a tight, complete response near the top, ideally a single focused paragraph.
No. GPTBot, ClaudeBot, and PerplexityBot don't execute JavaScript. If your schema is injected via JS (common with React apps and some CMS plugins), AI crawlers can't see it. Server-side rendering is required.
Pages ranking #1 see a 58% drop in click-through rate when an AI Overview appears above them, according to Ahrefs research from December 2025. However, pages cited within the AI Overview see roughly 35% higher organic CTR, creating a significant gap between cited and non-cited pages.
GEO is a layer on top of SEO, not a replacement. Traditional ranking signals still matter. But optimizing for AI citation (source attribution, statistical density, self-contained passages, multimodal content) is now a separate, measurable discipline that affects traffic independently of ranking position.