AI Overviews FAQ and How-To Content: The Formats Google Cites Most in 2026

Key takeaways

• How-to content held a steady 12–15% share of AI Overview citations across all of 2026 while listicles fell from 26% to 18%, making it one of the most reliable formats you can invest in.
• Google killed FAQ rich results in May 2026, but FAQPage schema still matters: pages with FAQ markup were cited by AI Overviews at roughly 3x the rate of pages without structured data.
• 44.2% of AI citations are extracted from the first 30% of a page. If your answer isn't near the top, it probably isn't getting cited.
• Question-format headings, 40–60 word answer capsules, and coverage of follow-up questions are the three structural moves that matter most.
• AI Overviews cite about 10 sources per response, roughly double ChatGPT's average, which makes it the most accessible AI surface for newer and mid-authority domains.

What Google actually cites in 2026

If you've been following AI search advice for more than a few months, you've probably heard that listicles are the format to chase. That was true entering 2026. It's considerably less true now.

Promptwatch's classification of AI Overview citations in Q1 2026 (over 20 million classified citations from January through early April) put listicles at 26.1% of all citations, well ahead of news articles at 15.3% and how-to content at 12.1%. By July 2026, the picture had shifted quite a bit. Listicles had dropped to 18.0%, product pages had climbed to 16.3%, and how-to content held at 15.1%.

Here's the quarter-over-quarter comparison:

The July data has one detail worth pausing on: from July 28 onward, product pages overtook listicles as the single most cited format every single day, ending the month at 17.9% versus 16.2%. That's the first time listicles lost the top spot. You can see the full daily breakdown in Promptwatch's AI Overviews citation types report for July 2026.

What does this mean for FAQ and how-to content? Two things.

First, how-to content is the most stable performer in the entire taxonomy. While listicles swung from 26% down to 18% and product pages nearly doubled, how-tos sat in the 12–15% band all year. If you want a format that won't be obsolete in six months, this is it.

Second, the rise of product pages tells you something about Google's intent-matching. Google is increasingly citing the page that best matches the query's job, whether that's a buying decision or a task. A well-structured FAQ section on a product page and a how-to guide that walks through using the product are both riding this shift. The same pattern shows up in ChatGPT citations, where product pages hit 32.8% of citations in July while how-tos and comparison pages grew steadily through the month with relatively low competition.

The FAQ paradox: rich results died, FAQ content didn't

Here's where a lot of 2026 advice goes wrong.

Google deprecated FAQ rich results. As of May 7, 2026, the expandable FAQ dropdowns stopped appearing in Search entirely, and Search Console reporting plus the Rich Results Test for FAQ markup are being phased out. HowTo rich results went even earlier: mobile in August 2023, desktop a month later, and the documentation has since been removed altogether.

So FAQ content is dead, right?

No. And this distinction matters a lot. Google killed the display, not the markup and not the content format. The paradox is that while FAQ rich results vanished from SERPs, FAQ-structured content became more valuable for AI citations, not less.

An SGA Index analysis of AI Overview citations found that pages with FAQ schema were cited at a 13.6% rate versus 4.2% for pages with no structured data. That's roughly a 3x lift from markup that no longer produces any visible SERP feature. The reason is straightforward once you think about how AI Overviews are built: Gemini reads schema to understand what a page is about, what question it answers, and who published it. Schema that never triggers a rich result can still decide whether your content gets cited.

There's a wrinkle, though. Google's March 2026 core update cracked down on what some are calling schema padding, markup that describes supplementary content rather than the primary content of the page. FAQ rich result impressions dropped by nearly half across tracked sites after that update. If you slap FAQPage schema on a blog post where the FAQ section is an afterthought at the bottom, you're in the crosshairs. If your page genuinely is a Q&A resource, or the FAQ section carries real substance, the markup still works in your favor.

Only about 12% of websites use structured data at all, which means clean, accurate FAQ markup remains a genuine differentiator even after the deprecation.

How to write FAQ content that earns citations

The mechanics of a citable FAQ section come down to how easily Gemini can extract a self-contained answer from it. Here's what the evidence supports.

Lead with the answer, always

Wix's extraction-position research found that 44.2% of all AI citations are pulled from the first 30% of a page. That applies within sections too, not just pages. Every FAQ answer should put the actual answer in the first sentence, with the answer complete within two or three sentences. If someone copied just your first sentence and pasted it into an AI-generated summary, it should make sense on its own.

A practical pattern many SEO teams now use is the answer capsule: a 40–60 word direct answer immediately under each question heading, followed by elaboration. It feels repetitive when you write it. It works anyway, because you're not writing for a human reading top to bottom, you're writing for a system looking for a clean extract.

Use question headings that mirror real queries

"Email Marketing Best Practices" is a topic. "What are the most effective email marketing strategies for small businesses?" is a question, and it's the kind of heading AI Overviews can match directly to a query. Prioritize questions with intent modifiers like "how to", "what is", "why does", and "when should".

Base your FAQ questions on real language, not invented ones. Support tickets, sales calls, Quora threads, People Also Ask boxes, and Reddit posts in your niche are all good sources. People Also Ask remains the most common SERP feature, and it's essentially a free list of the questions Google already associates with your topic.

Cover the follow-up questions

AI Overviews are built through query fan-out: Google expands the original query into multiple sub-queries and pulls citations for each. A page that answers the initial question plus the obvious follow-ups can get cited across several of those sub-queries from a single URL.

If someone searches "how to start a podcast", the fan-out includes equipment costs, hosting platforms, and promotion. Cover those. A tool like Promptwatch can show you the actual fan-out queries for your target prompts, which beats guessing.

Keep it neutral and factual

AI Overviews tend to skip pages heavy with promotional language, opinion, or speculation. This is one of the quieter findings across multiple studies: neutral, factual, clearly written content wins. Save the persuasion for your landing pages and keep FAQ sections clean.

How to structure how-to content that earns citations

How-to guides have quietly become one of the best risk-adjusted content investments in AI search. The share is stable, the competition is lower than for listicles, and the format maps naturally to how people actually query. Here's how to build them well.

Numbered steps with extractable step text

Each step should be a discrete, self-contained unit: a clear action in the first sentence, then the detail. Avoid steps that only make sense if you read the previous three. Gemini assembles answers from fragments, and fragments that stand alone get used.

Match length to intent, not to a word count

Cited content averages around 18 words per sentence with heavy use of H2/H3 structure. Document length varies by engine: Copilot skews short (~964 words), Gemini skews long (~1,977 words). For AI Overviews specifically, Ahrefs found the average cited page runs about 1,282 words. Write the length the task requires and stop.

One question per page, answered completely

Pages that try to cover everything often lose to pages that answer one thing well. Google's systems match content to the question, not just the topic. If your how-to covers setup, troubleshooting, and pricing on one page, consider whether splitting it would give each question a cleaner, more citable answer.

Add video where it fits

Video citations in AI Overviews roughly doubled across 2026, from about 2.7% in January to nearly 6% by late July. YouTube accounts for a meaningful share of citations even when the video doesn't rank in the traditional top 10. A short walkthrough video embedded in your how-to guide gives Google a second format to cite from the same URL.

Schema: what still matters after the March 2026 update

With FAQ and HowTo rich results gone from display, the strategic value of schema has shifted from "trigger a SERP feature" to "help AI systems verify and understand your content." The March 2026 core update made this explicit.

What works now:

• FAQPage markup on pages where Q&A is the primary content, not padding
• Entity schema (Organization, Person) with SameAs identifiers pointing to Wikidata, LinkedIn, Crunchbase. Sites with clean entity schema saw a reported 3.2x AI Mode citation lift in one analysis
• Product and Offer markup, which retains full rich-result support and aligns with the product-page citation surge
• JSON-LD in the document head, which remains Google's preferred delivery format

What to stop doing:

• HowTo markup expecting a rich result (the feature no longer exists)
• FAQ schema on pages where the FAQ section is decorative
• Review and editorial markup on pages it doesn't accurately describe

The mental model shift is simple: AI Mode and AI Overviews don't display your schema. They read it. Markup that never produces a visible feature can still determine whether you get cited.

Which format should you invest in?

Here's a quick decision guide based on the 2026 citation data:

One more consideration: AI Overviews cite about 10 sources per response, roughly double ChatGPT's average of 5, and that number has been notably stable over time. More citations per answer means more slots to win, which is why AI Overviews is often described as the most forgiving AI surface for newer and mid-authority domains. Classic SEO signals still carry weight there.

Measuring whether it's working

You can't manage what you don't measure, and AI citation measurement is its own discipline now. The metrics that matter have shifted: instead of tracking FAQ rich result impressions in Search Console (which no longer exist), track citation frequency in AI Overview and AI Mode answers for your target prompts.

A few practical steps:

• Track a fixed set of prompts monthly and record whether your domain appears in the AI Overview citations for each
• Watch which of your page types earn citations. If your how-tos get cited and your listicles don't, that's your answer about where to invest
• Monitor citation churn. Monthly citation churn is high across AI search, so a citation you earn this month is not a citation you keep by default
• Check that AI crawlers can actually reach your pages. Crawler log data shows which of your pages AI systems visit, what they read, and whether they hit errors. Promptwatch's Agent Analytics does this, and it explains a lot of "why am I not cited" mysteries that visibility scores alone can't

If you want to compare the full landscape of tools in this category, the GEO software directory at bestgeosoftware.com covers the main options.

Common mistakes to avoid

That last row deserves emphasis. Google AI Overviews and Google AI Mode cite the same URLs only about 13.7% of the time despite often reaching similar conclusions. Each surface has its own source preferences. A brand that wins AI Overview citations can be completely invisible in Perplexity or ChatGPT Search, and vice versa.

The bottom line

The 2026 data tells a clear story: the formats Google cites are diversifying. Listicles are no longer the automatic answer, product pages are surging, video is doubling, and how-to content is the steady performer that never left. FAQ content survived the death of its rich result because its value was never really about the SERP dropdown. It was about structure, and structure is exactly what AI citation selection rewards.

If you do one thing after reading this, audit your ten most important pages and check two things: is there a direct, extractable answer in the first 30% of each, and does the markup accurately describe the page's primary content? That combination, repeated across your library, is most of the game.