What is a Featured Snippet? How to Win Position Zero and Its Link to AI Overviews

The featured snippet sits in position zero on Google: a box above the first organic result that answers the query directly, without the user needing to click through. The text, image or figures inside always come from a page that is already indexed and that, on top of that, ranks within the top ten organic results for that query. That's a requirement, not an option. Google doesn't pull snippets from pages that aren't already on the first results page.

Google states plainly in its own documentation that there is no way to request or force this position directly. Its systems decide automatically whether a page fits the query. The only thing a site can actively control is the opposite: opting out with the nosnippet tag, or shortening what Google can show with max-snippet. There's no equivalent tag that forces entry.

Before AI Overviews existed, the featured snippet showed up on roughly 12 to 15% of informational searches in the US, according to Ahrefs analyses from that period. Along with the local pack, it was the SERP feature with the biggest pull on clicks away from the first organic result: whoever won position zero usually also got the click, though not always, since part of the audience reads the answer and never opens a page. That share of clickless searches is exactly what has shifted with AI Overviews, covered further down.

A featured snippet also isn't permanently locked to one page for a given query: Google often tests several candidate pages in parallel and switches the source once another page delivers the structure more clearly. On mobile devices and with voice assistants such as Google Assistant, the featured snippet is frequently the only answer read out loud, with the user never seeing a results list at all. For a highly searched, competitive term like "featured snippet" itself, clean structure pays off twice over: it counts for classic desktop search just as much as for voice output.

Google doesn't assemble the snippet from the general meaning of a page, it pulls it straight from the HTML structure. The first step to winning a featured snippet isn't better writing, it's organizing the content in the format that particular snippet type needs. There are three types, and each one demands a different structure.

Paragraph snippet. The most common type, it appears for definition and direct-explanation questions ("what is", "why", "how much"). The format that works best puts the question as a heading (H2 or H3, close to how a user would phrase it), followed directly by the answer in two or three sentences, 40 to 60 words, no detours. If the answer comes after a longer intro paragraph, Google can still extract the right sentences, but tends to favor answers placed right after the question. The rest of the paragraph, and the article, can add nuance, examples or exceptions afterward.

List snippet. This shows up for step-by-step questions ("how to") or enumerations ("best X", "types of X"). The most common mistake is writing the list as running prose, items separated by commas inside a single <p>. Google can sometimes reconstruct a list from that, but does it far less often than when the page already uses a real HTML list: <ol> for ordered steps (a recipe, an install process), <ul> for items with no required order (features, benefits, types). Each list item should be a short, self-contained sentence, not a whole paragraph stuffed into an <li>.

Table snippet. Google uses this type for comparison queries and numeric data: prices, specs, deadlines, sizes. The near-mandatory condition here is that the data sit in an actual <table>, with <th> for column headers and one row per compared item. A table built as an image, as a CSS-div layout, or as space-aligned text, isn't readable as a table by the system that assembles this snippet. Fewer columns and clearer values make it more likely that Google will pull the whole table into the box.

The common thread across all three formats: Google doesn't guess the structure from the intent of the text, it reads it from the markup. A page can have perfect content and still lose the snippet to a worse-written but better-structured competitor.

Across all three types, an image from the page can also show up inside the box. The same logic applies here: an image placed right next to or below the relevant paragraph, with a descriptive alt text, stands a better chance of getting pulled in than one sitting elsewhere on the page with no direct tie to the specific answer.

Since Google rolled out AI Overviews broadly in 2024, the relationship between that feature and the classic featured snippet has changed in ways the data now confirm. An Ahrefs analysis of one million US desktop SERPs, between January and June 2025, found the share of searches showing a featured snippet dropped from 15.41% to 5.53% over that half-year, a 64% decline. Over the same period, AI Overview presence rose from 3.93% to 27.43% of searches, a 598% increase. The correlation between the two curves sits at -0.90, a figure that in practice means: where one grows, the other falls, almost as a mirror image. This figure comes from a third-party source citing the Ahrefs study; it wasn't possible to check it against the original Ahrefs report, so it's presented as reported, not independently confirmed.

A separate Semrush study covering ten million keywords, also from 2025, puts AI Overview presence at 13.14% of queries in March of that year, rising to 25% by the last quarter. By industry, healthcare and education queries reached close to 50% AI Overview coverage in 2026. This figure, too, comes through a secondary source citing the Semrush study, not the original report.

What does look consistent across sources is that the two features tend to be mutually exclusive on the same query. When an AI Overview shows up, a classic featured snippet is less likely to appear for that same search. The displacement is strongest for "how to" and "best X" queries, exactly the territory of list snippets, and weaker for short definition questions or navigational queries, where the featured snippet appears to hold up better.

The traffic impact is probably the part that should worry sites that relied on the featured snippet for clicks: according to the same analyses, the first organic position gets roughly 58% fewer clicks when an AI Overview sits above it, and 76% of pages cited inside an AI Overview were already ranking in the top 10 before being cited. Staying in the top 10 and keeping content well structured remains the baseline, both for the featured snippet and for being cited inside an AI Overview, but the actual payoff, the click, is now split differently and, on the whole, shrinking.

• Phrase the question exactly as users search it, as an H2 or H3, followed directly by the answer in two sentences, no filler paragraph before it.
• Use <ol> for ordered steps and <ul> for unordered lists; never bury a list inside one running paragraph if you want the list snippet.
• Build comparisons and figures into a real <table> with a <th> header row, not as an image or a div layout.
• Get a top-10 ranking before polishing the structure. Without that, no format wins the snippet.
• Refresh content regularly; Google re-evaluates who holds position zero when it detects meaningful changes on the topic.
• Pay attention to citability for AI Overviews too: a clear answer backed by original data and verifiable sources has better odds of being cited, even without winning the classic featured snippet.

• Confusing a featured snippet with a rich snippet: the latter depends on schema markup (reviews, prices, FAQ), the former is pulled from the page's text and HTML structure, not from schema.
• Adding structured data on the assumption that this alone triggers a featured snippet; it helps other rich results, but it isn't the mechanism Google uses to assign position zero.
• Opening the answer with a sentence that's too long or too vague instead of stating the specific value right away.
• Writing lists and comparisons as comma-separated running text instead of using <ol>, <ul> or <table>.
• Embedding a comparison table's values as an image, which makes them invisible to the system that assembles the snippet.
• Leaving content stale for years, even after losing position zero to a competitor.