Key takeaways
- Zero-click searches now make up 65-70% of all Google queries, and organic clicks are down 42% from pre-AI Overviews baselines -- but some content formats are bucking that trend
- Breaking news content saw 103% traffic growth even as informational content collapsed, showing that recency and original data are now major differentiators
- The formats most likely to get cited AND clicked are original research, industry reports, comparison content, and tool/product roundups -- not generic how-to posts
- Structured content (clear headings, FAQ sections, tables) gets pulled into AI summaries more often, but the click happens when the AI can't fully answer the question from the snippet alone
- Tracking which of your pages actually get cited by AI models is now a core part of any content strategy -- tools like Promptwatch make this measurable
The headline number is brutal: organic clicks are down 42% from their pre-AI Overviews baseline, according to Define Media Group's analysis of Google Search Console data across 64 sites. That's not a blip. It's a structural change in how Google works, and it's been compounding since AI Overviews expanded at scale in 2025.
But here's the thing most coverage misses: the 42% drop is an average. Some content formats are declining far faster. Others are actually growing. The difference isn't topic -- it's format and intent.
This guide breaks down which content types are generating clicks from AI search in 2026, which ones AI models cite but don't send traffic from, and how to think about your content mix going forward.
Why format matters more than topic now
Before getting into the breakdown, it's worth understanding the mechanic. AI search engines -- Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini -- don't just rank pages. They synthesize answers. They pull from multiple sources, compose a response, and deliver it directly.
For a user asking "what is compound interest," Google's AI Overview answers the question completely. There's no reason to click. The page that used to rank #1 for that query is now invisible.
But for a user asking "best project management tools for remote teams in 2026," the AI can summarize a few options -- but the user often wants to read the full comparison, check pricing, or see screenshots. That's a click.
The difference is whether the AI can fully satisfy the query from a snippet, or whether the content has enough depth, specificity, and recency that the user needs to visit the source.
Seer Interactive's study across 3,100+ queries found that organic CTR dropped 61% for searches where AI Overviews appeared (from 1.76% to 0.61%). But that average masks huge variance by content type. A PikaSEO study cited by Leadfeeder found the top search result sees about 58% fewer clicks when an AI overview is present. For some formats, the drop is near-total. For others, it's minimal.
Content formats that still drive clicks
Original research and proprietary data
This is the format that AI models most want to cite and that users most want to visit. When you publish data that doesn't exist anywhere else -- a survey, an analysis of your own dataset, an industry benchmark -- AI models have no choice but to attribute it to you. They can quote the finding, but they can't replicate the full methodology, the tables, or the context.
One LinkedIn analysis of 30 AI search content types in 2026 argued that industry reports are "becoming the most valuable asset because they provide source-of-truth data that LLMs crave." That tracks with what we're seeing in citation patterns.
The click happens because the AI summary says something like "according to [Brand]'s 2026 State of X report..." and curious readers follow the source. The citation itself is the traffic driver.
What makes this work: a clear, quotable headline stat, a methodology section, and data that's specific enough to be useful but complex enough that the AI can't just reproduce it in a paragraph.
Comparison and "best of" content
Comparison content is surviving the AI search transition better than almost any other format -- but only when it's genuinely comprehensive and current.
The reason is intent. Someone searching "Notion vs Coda for teams" or "best email marketing tools under $100/month" has a decision to make. The AI can give them a summary, but the decision usually requires reading the full breakdown, checking current pricing, and seeing real-world use cases. That's a click.
The formats that work here: head-to-head comparisons with a clear verdict, roundups with 5-10 options that include pricing and use-case specificity, and "alternatives to X" pages that address a specific switching scenario.
What kills comparison content: generic feature lists that the AI can reproduce in three sentences, outdated pricing, and "it depends" conclusions that don't actually help anyone decide.
Breaking news and timely analysis
This is the one category where traffic actually grew. Define Media Group's data shows news content traffic up 103% even as informational content collapsed. The reason is simple: AI models can't synthesize news that happened an hour ago. They have to cite the source.
The opportunity here isn't just for news publishers. Any brand that publishes fast, original takes on industry developments -- a product launch, a regulatory change, a major study -- can capture this traffic. The window is short, but the clicks are real.
The key is speed and specificity. A generic "here's what the new Google algorithm update means" post published three days later competes with hundreds of similar pieces. A post with original analysis, published within hours, with a specific angle that no one else has taken, gets cited.
Tool and product roundups with real testing
"Best [tool category] for [specific use case]" content is one of the most durable formats in AI search -- but the bar has risen significantly. AI models are getting better at detecting thin affiliate content, and they increasingly favor sources that show evidence of actual testing.
What this means practically: include specific observations from using the tools, note limitations alongside strengths, and update the content when tools change. A roundup that was accurate in 2024 but hasn't been touched since is a liability, not an asset.
The click happens because users want to read the full methodology and see the details that didn't make it into the AI summary. If your roundup is just a feature table that the AI can reproduce, you won't get the click.
In-depth guides with original perspective
Long-form guides still work, but "long-form" alone isn't the differentiator anymore. The AI can summarize a 3,000-word how-to post in four bullet points. What it can't summarize is a guide that's built around a specific perspective, a novel framework, or a set of hard-won observations that aren't available anywhere else.
The guides generating clicks in 2026 tend to have a clear point of view, use specific examples from real experience, and go beyond what any AI model could synthesize from existing sources. They're not just comprehensive -- they're irreplaceable.
Content formats that are struggling
Generic how-to content
This is the category taking the hardest hit. "How to write a cover letter," "how to set up Google Analytics," "how to make sourdough bread" -- these queries are now almost fully answered by AI Overviews. The CTR for informational how-to content has collapsed.
The fix isn't to stop writing how-to content. It's to make it specific enough that the AI can't fully answer it. "How to set up Google Analytics 4 for a Shopify store with server-side tracking" is a different query than "how to set up Google Analytics." The more specific the use case, the more likely the user needs to read the full guide.
Definition and explainer posts
"What is [term]" content is essentially dead as a traffic driver. AI models answer these perfectly and completely. If your content strategy relies heavily on definition posts targeting high-volume informational keywords, those pages are likely showing the steepest declines.
The path forward is to turn definition posts into something more: a definition plus a framework, a definition plus a comparison, a definition plus original data. The definition alone isn't enough.
Thin listicles
"10 tips for better sleep" or "5 ways to improve your email open rates" -- these are easy for AI to reproduce. If the list items are generic enough to appear in any article on the topic, the AI will just include them in its overview and the user won't need to click.
Listicles that survive are the ones with specific, non-obvious items, real examples, and enough depth per item that the summary can't do it justice.
Format comparison: click potential vs. citation potential
Different formats serve different goals in AI search. Some get cited frequently but don't drive clicks. Others drive clicks when cited. Understanding the difference helps you build a content mix that serves both.
| Content format | Citation frequency | Click-through rate | Best for |
|---|---|---|---|
| Original research / data | Very high | High | Brand authority + traffic |
| Breaking news / timely analysis | High | High | Short-term traffic spikes |
| Comparison / "best of" roundups | High | High | Decision-stage traffic |
| In-depth guides (with POV) | Medium | Medium-high | Long-term authority |
| Tool/product reviews with testing | Medium | High | Affiliate + decision traffic |
| Generic how-to content | High | Very low | Almost no click value now |
| Definition / explainer posts | Very high | Very low | Brand awareness only |
| Thin listicles | Medium | Low | Diminishing returns |
| FAQ pages | Medium | Low-medium | Featured snippet capture |
| Video transcripts / multimedia | Medium | Medium | Multimodal search |
The structural shift: being cited vs. being clicked
One thing that's easy to miss in this conversation: citation and clicks are now two different metrics, and both matter.
Getting cited by an AI model without a click still has value. It builds brand familiarity. Users see your name associated with authoritative answers. Over time, that recognition influences direct search behavior and purchase decisions.
But clicks are what pay the bills for most content teams. And the content formats that drive clicks from AI search share a common trait: they contain something the AI can't fully reproduce in a summary. Proprietary data, specific product testing, a genuine point of view, real-time information.
The practical implication is that your content strategy needs to answer two questions for every piece you publish: "Will AI models cite this?" and "Will users click through to read the full thing?" The formats above score differently on each question, and a healthy content mix needs both.
Structural elements that improve both citation and click rates
Beyond format, there are structural choices that affect how AI models process and cite your content -- and how likely users are to click.
Clear heading hierarchy matters. AI models parse page structure to understand what a piece of content covers. Pages with logical H2/H3 structures that match user intent get pulled into overviews more reliably than walls of text.
FAQ sections at the end of articles capture question-based queries. They're not a traffic goldmine on their own, but they increase the surface area of a page for AI citation.
Tables and structured data are easy for AI to extract and reproduce -- which is a double-edged sword. A pricing table gets cited easily, but the user might not need to click. A table that's part of a larger analysis (like a comparison with methodology notes) is more likely to prompt a click for the full context.
Author credentials and first-person experience signals are increasingly important. AI models, particularly Google's systems, are weighting content from identifiable experts more heavily. A guide written by someone with demonstrable experience in the topic performs better than anonymous content.
Recency signals matter more than they used to. AI models prefer fresh sources for anything time-sensitive. Updating existing content with new data, new examples, and a clear "last updated" date is a legitimate strategy for maintaining citation frequency.
Tracking what's actually working
The challenge with all of this is that it's hard to know which of your pages are being cited by AI models, which prompts are driving those citations, and whether citations are translating into traffic.
Traditional Google Search Console data doesn't capture AI referrals well. You can see some of it in direct traffic or "dark social" patterns, but the picture is incomplete.
Tools built specifically for AI search visibility fill this gap. Promptwatch tracks which of your pages are being cited across ChatGPT, Perplexity, Claude, Gemini, and other AI models, and connects that citation data to actual traffic through GSC integration or server log analysis. That kind of page-level visibility is what lets you test whether a new piece of original research actually moved your citation frequency, or whether a comparison update improved click-through from AI summaries.
Without that data, content strategy in 2026 is mostly guesswork. You're publishing and hoping, rather than publishing and measuring.
What to actually do with this
A few concrete changes worth making based on the format breakdown above:
Audit your existing content for "AI-completeness." If an AI model can fully answer the query using just the first 200 words of your article, that page is at high risk of zero-click status. Add original data, specific examples, or a perspective that requires reading the full piece.
Shift investment toward original research. Even a small survey of your customer base, or an analysis of your own product data, creates citation-worthy content that competitors can't replicate.
Update comparison content aggressively. Stale pricing, discontinued tools, and outdated feature comparisons are worse than no comparison at all. Set a quarterly review cadence for your highest-traffic comparison pages.
Build a news/analysis capability. You don't need a full editorial team. Even one person publishing fast, specific takes on industry developments within hours of them happening can capture the traffic window that AI models can't fill.
Stop publishing generic how-to content at scale. The ROI is gone for most topics. The exception is highly specific, technical how-to content where the use case is narrow enough that the AI can't fully answer it.
The content formats that generate clicks from AI search in 2026 are the ones that contain something irreplaceable: data that only you have, testing that only you did, a perspective that only you can offer, or information so fresh that no AI model has processed it yet. Everything else is increasingly just training data.