ContentRanking #1 doesn’t mean what it used to. When an AI Overview sits above the results, organic click-through on the top-ranking page falls by roughly 58%, according to an Ahrefs analysis cited in a Search Engine Journal field study covering 300,000 keywords. Pew Research found people click an organic result only 8% of the time when an AI Overview appears, versus 15% when it doesn’t. 📉 That’s not a reason to quit writing blog posts. It’s a reason to get specific about which formats still earn the click, or better yet, earn the citation inside the AI answer itself.
Here’s the part that changes the whole calculus: being cited inside an AI Overview isn’t the same as merely ranking below it. Brands the AI actually quotes capture dramatically more of whatever traffic remains than brands sitting in position one underneath the summary, some analyses put that gap well over 100%. 🎯 So the real question for 2026 isn’t “how do I rank higher.” It’s “which formats get lifted into the answer.” Five formats keep showing up in that research, and they happen to be the same five that convert once someone actually clicks through.
Ranked listicles still win the algorithm’s attention
Numbered “best of” and “top N” posts are unglamorous, and they’re also the single most-cited format across large-scale studies of AI search behavior. One analysis of roughly 400 million citations across 25,000 URLs found that 63% pointed to listicle-format pages, per HubSpot’s 2026 State of AEO research. Even more conservative studies still put listicles at the top of the pile, just at a lower share. 🏆
The reason is mechanical, not mysterious. A numbered list with named items is easy for an AI system to lift cleanly, it doesn’t have to interpret tone or untangle a narrative. Structure it right and you get pulled into the answer nearly intact.
Use a real ranking, not a random order. “Top 5” or “Best 7” formats dominate the citation data, unordered roundups perform worse
Front-load a one-line verdict for each item before the deep dive, that’s the sentence most likely to get quoted
Add a small comparison table or “best for” tag per item, structured summaries get cited 3 to 5 times more often than plain prose lists
Keep descriptions specific, “handles low light exceptionally well” beats “great camera” every time
This format wins commercial, decision-stage searches specifically, the “which one should I get” queries closest to an actual purchase. That’s also why it converts. Someone scanning a ranked list is already comparing options, not learning what a category even is.
Comparison posts capture the traffic AI Overviews can’t summarize away
“X vs Y” content does something the listicle can’t: it resolves a single, specific decision. And it turns out AI engines treat that resolution as authoritative. Comparison content wins ChatGPT citations outright, with a citation rate as high as 95%, the highest of any format measured on any engine, per the same HubSpot research. ⚔️
There’s a second reason this format holds up better than most against the AI Overview traffic drain: transactional, buyer-intent queries trigger AI Overviews far less often than informational ones. E-commerce queries sit around 4% AI Overview coverage compared to 80%+ for health or B2B topics. That means the clicks on a genuine “vs” comparison haven’t been siphoned off nearly as hard.
Pick two options people are already deciding between, don’t invent a comparison nobody’s searching for
Structure it around the actual decision criteria, price, ease of use, support, not a generic feature checklist
Declare a winner, or at least a “pick X if / pick Y if” split, comparisons that refuse to conclude anything get skimmed and abandoned
Use the exact product or tool names in your headings, entity-specific language is what both search engines and AI systems match against
Comparison posts convert because the reader has already done the hard part, narrowing the field to two. Your job is just to break the tie.
FAQ blocks are the highest-leverage real estate on the page
A standalone FAQ page performs poorly. A tight FAQ block embedded inside a longer, authoritative post performs very differently. Pages with FAQPage schema applied are roughly 3.2 times more likely to appear in AI Overviews, and pages with FAQ sections plus inline citations rank about 40% higher in source selection than pages without them. ❓
The mechanism makes sense once you think about it: a question-and-answer pair is already shaped exactly like a user’s prompt to an AI system. You’re not making the engine guess what question your paragraph answers, you’re handing it the question pre-written.
Pull FAQs from your actual comment section, sales calls, or support inbox, invented questions read as generic and get skipped
Keep each answer to two or three sentences, this is the one place brevity beats depth
Mark it up with FAQPage schema, the structural signal does real work here, it isn’t decorative
Place it after your main content, not instead of it, FAQ blocks perform best as reinforcement, not as the whole page
This is exactly the kind of topic the BizWhat Membership digs into properly, with a dedicated ebook and real implementation details on structuring posts for both human readers and AI extraction at the same time.
Step-by-step guides win by front-loading the answer
How-to content still earns real citation share, but the research on where AI systems pull from inside these posts is the useful part. Roughly 44% of all citations come from the first 30% of a document, with only about a quarter coming from the final third, according to the same large-scale citation analysis behind the listicle numbers. 🔧 Translation: your intro can’t be throat-clearing anymore, it has to be the answer.
That changes how a step-by-step guide should actually open. Don’t build up to the method, state it, then unpack it.
Open with a one- or two-sentence direct answer before your first numbered step, not after
Use HowTo schema and numbered H2s or H3s so each step is independently extractable
Keep sentences tight, heavily cited content averages around 18 words per sentence, not 30
Match length to the query, a quick fix doesn’t need 2,000 words padded around a five-step process
This is also where BizWhat’s own faceless blog guide is worth revisiting, it makes the case that Google traffic earned this way is slower to build than social traffic but far more durable once it’s ranking. A tight how-to converts for the same reason it gets cited, the reader gets what they came for immediately instead of scrolling past your life story to find step one.
Original research and case studies are what Google’s latest update actually rewards
This is the format most publishers skip because it’s the most work, and that’s exactly why it’s the highest-leverage one left. Google’s March 2026 core update rebalanced its ranking signals toward genuine, firsthand experience, and sites with named authors, original data, and documented outcomes gained ground on competitive queries while thin, rewritten content lost it. Google’s own guidance is blunt about this: content should demonstrate first-hand expertise, not just cover a topic, per Google Search Central’s people-first content documentation. 🔬
A documented case study checks every box in Google’s E-E-A-T framework simultaneously. You did the thing, you understood what happened, other people cite your numbers, and your methodology is visible enough to trust.
Publish your own numbers, even a small sample size beats a rewritten summary of someone else’s study
Show the method, not just the result, “here’s exactly how we tested this” is what earns the backlink
Update it when the underlying data changes, stale original research loses its edge fast
Pair it with a real author bio and credentials, anonymous “research” reads as unverifiable and gets discounted
BizWhat’s own roundup on recession-resilient online businesses leans on this same idea, specific niches and specific search demand, not vague optimism about “content that works.” That specificity is the whole point. Which of these five formats is your last post actually missing, the ranking, the comparison, the FAQ block, the front-loaded answer, or the original data? Pick the gap and fix the next post before you write another one from scratch.