Key takeaways
- Original research and comparison content are the top-performing content types in AI search, with performance scores of 82% and 76% respectively (Neil Patel data).
- How-to guides and listicles targeting informational queries are the hardest hit by AI Overviews -- organic CTR for these formats has dropped as much as 61% when AI Overviews appear.
- Commercial and transactional intent content (pricing pages, case studies, comparison pages) retains clicks because AI can't fully replace the decision-making experience.
- The content that survives AI summarization is content that requires interactivity, fresh data, or nuanced judgment -- not content that answers a single factual question.
- Tracking which of your pages actually get cited by AI models (and which drive traffic from those citations) is now a core part of any content strategy.
Something shifted in 2025 and it didn't shift back. Organic clicks are down 42% from pre-AI Overviews baselines, according to Define Media Group's analysis of Google Search Console data across 64 sites. That number has been cited enough times that it's easy to gloss over -- but it represents years of content investment quietly losing its return.
The agencies and content teams struggling most aren't the ones who made bad content. They're the ones who made the wrong kind of content for the current environment. How-to guides, definition posts, listicles answering common questions -- these formats built entire editorial strategies. Now Google (and ChatGPT, and Perplexity) answers them directly, often without sending a single click.
So which formats actually work in 2026? The answer is more nuanced than "comparison pages win, everything else loses." Let's go through the data.
The zero-click reality you're working against
Before comparing formats, it helps to understand the baseline. Zero-click searches now account for roughly 65-70% of all Google queries. AI Overviews trigger on about 18% of searches, and when they do, Seer Interactive found that organic CTR drops from 1.76% to 0.61% -- a 61% decline for the #1 result.
That's not a ranking problem. A page can rank first and still lose most of its clicks because the AI summary above it answered the question well enough.
The content types hit hardest are exactly the ones that were easiest to rank for: how-to content, definition posts, comparison guides targeting informational queries, and listicles answering "what are the best X" questions. These formats worked because they matched search intent cleanly. That same quality -- clean, extractable answers -- now makes them easy for AI to summarize away.
The content that retains clicks tends to share one characteristic: it can't be fully summarized without losing something important. Either the experience of reading it matters, the data is too fresh or specific to reproduce, or the decision it supports requires more than a paragraph.
How comparison pages perform in AI search
Comparison content ranks second overall in AI search performance at 76%, behind only original research (82%), according to Neil Patel's dataset on content types in AI search. That's a meaningful gap from the next tier.
Why do comparison pages hold up? A few reasons.
First, comparison queries tend to carry commercial intent. Someone searching "Notion vs Coda for project management" or "best CRM for small agencies" isn't looking for a quick definition -- they're making a decision. AI models can summarize the key differences, but the click often still happens because the user wants the full breakdown, wants to see pricing, or wants to trust a source before committing.
Second, comparison content is harder to compress. A good comparison page has structured data, nuanced tradeoffs, and specific use-case recommendations. AI Overviews can pull a few bullet points, but the full page still has value.
Third, comparison pages are frequently cited as sources within AI responses. When ChatGPT or Perplexity answers "what's the difference between X and Y," it often cites a comparison article. That citation drives clicks even when the AI has already summarized the answer.
The catch: not all comparison content performs equally. Thin "X vs Y" pages that just list features side-by-side are increasingly summarized away. The comparison pages that drive clicks in 2026 are the ones with original takes, real testing data, and specific recommendations -- not generic feature tables.
How how-to guides perform in AI search
How-to content is in the most difficult position of the three formats. It's the format most likely to be fully answered by an AI Overview, and it's the format where the 61% CTR drop is most pronounced.
The problem is structural. A how-to guide answers a procedural question: "how do I do X?" AI models are very good at answering procedural questions. They can reproduce the steps, add context, and often do it better than a generic guide written to rank for a keyword.
That said, how-to content isn't dead -- it's bifurcating. Two types of how-to content still drive clicks:
How-to guides tied to specific tools or platforms. "How to set up a Google Tag Manager container" is answerable by AI. "How to configure server-side tracking in Elevar for a Shopify store" is specific enough that AI either can't answer it confidently or cites your page as the source. The more specific the tool context, the more likely your guide survives.
How-to guides with genuine depth and original process. If your guide documents a proprietary process, includes screenshots of your actual workflow, or reflects hard-won experience that isn't available elsewhere, it has citation value. AI models cite sources they trust. A guide that reads like it was written by someone who actually did the thing is more likely to be cited than one that assembled information from other sources.
Generic how-to content targeting high-volume informational queries -- "how to write a blog post," "how to set up email marketing" -- is the most exposed. These pages are being summarized at scale.
How listicles perform in AI search
Listicles are in an interesting middle position. The format itself isn't the problem -- lists are actually a format AI models cite frequently, because they're easy to extract and present as structured answers. The problem is the intent behind most listicles.
"Best project management tools" or "top 10 SEO tips" listicles target informational queries where AI can reproduce the list directly. The user gets the answer without clicking. Position Digital's data makes this explicit: top-funnel content (what is, how-tos, guides) is losing traffic, while case studies and pricing pages are gaining it.
Listicles that perform better in AI search tend to be:
- Tied to commercial intent ("best CRM for agencies under $100/month" rather than "best CRM tools")
- Based on original testing or research ("we tested 12 email tools for 6 months, here's what we found")
- Specific enough that AI can't confidently reproduce the list without citing your source
- Updated frequently with fresh data, which AI models can't replicate from training data
The worst-performing listicles are evergreen "best of" posts targeting broad informational queries with no original data. These were the backbone of affiliate content strategies for years. They're now the most exposed format in the AI search era.
The content type performance comparison
Here's how the three formats stack up across the dimensions that matter most in 2026:
| Content type | AI search performance | Click retention | Citation likelihood | Best use case |
|---|---|---|---|---|
| Comparison pages | High (76% score) | Strong for commercial intent | High -- frequently cited as sources | Decision-stage buyers comparing options |
| How-to guides | Mixed | Low for generic queries, higher for tool-specific | Medium -- cited when highly specific | Technical/tool-specific processes with depth |
| Listicles | Mixed | Low for informational, higher for commercial | Medium -- cited when data-backed | Original research lists, commercial-intent roundups |
| Original research | Highest (82% score) | Strong | Very high | Any topic where you have proprietary data |
| Case studies / pricing | Growing | Strong | Medium | Bottom-funnel, conversion-stage content |
The pattern is clear: intent and specificity matter more than format. A comparison page targeting an informational query ("what's the difference between SQL and NoSQL") is as exposed as any how-to guide. A listicle based on original research with specific data is more defensible than a generic comparison page.
What actually drives clicks from AI citations
Being cited by an AI model and driving clicks from that citation are two different things. AI models cite pages regularly without those citations generating meaningful traffic. The pages that convert citations into clicks tend to share a few characteristics.
They offer something the AI summary can't. A tool, a template, a calculator, a database, a community -- something that requires the user to actually visit the page. Content that ends with "download our template" or "use our free calculator" has a reason for the click that the AI summary can't replace.
They're authoritative enough that users want to verify. When an AI cites a source for a claim that matters to the user's decision, curious users click through. This is especially true for statistics, research findings, and specific recommendations. Being the cited source for a data point is more valuable than ranking for the keyword.
They match the user's next step. A comparison page that ends with a clear recommendation and a path to act on it (free trial, pricing page, demo) captures users who are ready to move. The AI summary might have answered their comparison question, but the click happens because they want to take the next step.
Tracking which content types are actually working for you
The data above gives you a framework, but your site's performance will vary. A how-to guide in a technical niche might outperform a comparison page in a crowded market. The only way to know is to track which pages are being cited by AI models and which of those citations are driving traffic.
This is where most content teams are still flying blind. They can see organic traffic in Google Search Console, but they can't see which AI models are citing their pages, which prompts trigger those citations, or how citation rates correlate with traffic.
Promptwatch is built specifically for this -- tracking page-level citations across ChatGPT, Claude, Perplexity, Gemini, and other AI models, then connecting those citations to actual traffic through GSC integration or server log analysis.
The practical workflow: identify which of your existing pages are being cited (and for which prompts), identify which content types are generating citations that convert to clicks, and use that data to prioritize your content production. If your comparison pages are being cited but not clicked, the issue might be that they're too thin to warrant a visit. If your how-to guides aren't being cited at all, the issue might be that AI models are answering those queries without needing your source.
What to build in 2026
Based on the data, here's a practical content prioritization for teams rethinking their strategy:
Double down on comparison content with commercial intent. "X vs Y for [specific use case]" pages targeting decision-stage buyers are the most defensible format right now. Make them genuinely useful -- real tradeoffs, specific recommendations, updated pricing. Thin feature comparisons won't survive.
Shift how-to content toward tool-specific depth. Generic how-to guides are the most exposed. How-to content tied to specific tools, platforms, or workflows -- especially with screenshots, original process documentation, and real examples -- still has citation value and click potential.
Invest in original research. At 82% performance in AI search, original research is the highest-performing content type. Even modest original data -- a survey of your customer base, an analysis of your product data, a benchmark study -- gives AI models something to cite that they can't reproduce from training data.
Build bottom-funnel content. Case studies and pricing pages are growing in the AI search era, not shrinking. These formats serve users who are close to a decision, and AI models regularly cite them when answering commercial queries. If your content strategy is weighted toward top-funnel informational content, this is the rebalancing moment.
Add interactivity and tools. Content that includes a calculator, a quiz, a template, or a comparison tool has a reason for the click that AI summaries can't replace. Even simple interactive elements shift content from "summarizable" to "worth visiting."
The bigger picture
The formats that built content marketing over the last decade -- how-to guides, listicles, definition posts -- aren't going away. But their role is changing. They're increasingly useful as citation sources for AI models rather than as direct traffic drivers. The click happens when the user wants more than the summary, and that requires content with depth, specificity, and something the AI can't fully reproduce.
Comparison pages are the current winners because they serve decision-stage intent that AI can inform but not replace. Original research wins because AI models cite data they can't generate themselves. How-to guides and listicles survive when they're specific enough, deep enough, or interactive enough to be worth the visit.
The teams winning in 2026 aren't the ones who picked the right format. They're the ones tracking which of their pages are actually being cited, understanding why some citations convert and others don't, and building content that serves the moment after the AI summary -- not the moment instead of it.
