Search a definitional query on Google today and compare what sits above the blue links to what sat there two years ago. In 2022 you would often see a featured snippet: a boxed answer lifted verbatim from one ranking page, with that page’s title and URL underneath. In 2026 you are as likely to see an AI Overview, a generated summary stitched from several sources. Featured snippets have not vanished, but their share of the results page has shrunk, and understanding which surface a query produces now matters as much as knowing how to win either one.
A featured snippet is the answer box that appears at the top of some Google results, ahead of the regular listings. The box holds a short answer extracted directly from an indexed page, together with that page’s title and URL. Google selects the content from a page that already ranks well for the query, treating that page as authoritative enough for its text to be lifted out and shown prominently. Featured snippet optimization is the practice of structuring content so the algorithm extracts it, not writing snippet-only pages. The three primary formats (paragraph, list, and table) each reward a different content pattern.
One point worth settling before the tactics. The old shorthand “position zero” implied the snippet was a bonus slot above the ten organic results, an eleventh listing. That has not been true since January 22, 2020, when Google deduplicated snippet results. A page shown as the featured snippet no longer also appears as a separate blue link on page one, and the snippet counts as the page’s position one rather than an extra slot above it. The snippet now consumes an organic ranking position rather than sitting outside the ranked set. Winning it is still valuable for visibility, but it is not free real estate stacked on top of a normal ranking.
Three snippet formats, three content patterns #
The format Google uses depends on what the query asks and what the ranking pages contain. Three formats cover the bulk of displays, and independent studies of snippet inventories have consistently found paragraphs dominant, lists a distant second, and tables least common.
The paragraph snippet is a few sentences answering the query directly. Google pulls these from pages where a paragraph clearly defines or explains the topic. Definitional queries (“what is”, “what does”, “definition of”) and explanatory queries (“why is”, “how does”) produce paragraph snippets most consistently, and paragraphs make up the large majority of all snippets shown.
The list snippet is a bulleted or numbered list extracted from the page. Two sub-types exist: ordered (numbered) lists for sequential content like step-by-step instructions, and unordered (bulleted) lists for sets of related items with no inherent order. Process queries (“how to”, “steps to”), enumeration queries (“types of”, “examples of”), and ranking queries produce list snippets.
The table snippet is a structured comparison or data set extracted from a page. Tables appear for queries about comparisons, specifications, or data points that benefit from columnar presentation. Comparison queries (“vs”, “differences between”) and specification queries (“price of”, “dimensions of”) produce table snippets when ranking pages contain well-structured tables.
Format selection is not fully controllable. Google decides which format fits the query, and the ranking pages must already contain that format for extraction to work. A page with only paragraph content will not produce a table snippet regardless of the query. The strategy that follows: identify what format the target query currently produces (searching the query reveals the format Google has chosen), then structure the page to provide that format in a way at least as extractable as the current winner.
Paragraph snippets: the answer in 40 to 60 words #
Paragraph snippets run roughly 40 to 60 words. The selected text is a self-contained answer, readable without surrounding context, pulled from a single paragraph on the source page and lifted directly into the box.
The pattern that wins them is consistent. A clear definitional or explanatory paragraph sits near a heading that matches the query. The paragraph stands alone as an answer without depending on earlier or later context, and its opening sentence answers the question directly, not after preamble.
A working example. For the query “What is a canonical tag?”, a winning paragraph reads: “A canonical tag is an HTML element placed in the head section of a webpage that tells search engines which URL should be treated as the preferred version of a page. When the same content appears at multiple URLs, the canonical tag points search engines at one specific URL as the version that should rank.” Two sentences, roughly fifty words. The first is a complete definition; the second adds context but is not required for the answer to make sense. Google can extract the first sentence alone or both together, depending on length constraints. The heading directly above (“What is a canonical tag?” as the H2) mirrors the query phrasing and creates a strong signal that the paragraph below answers that exact question.
Three failing patterns account for most missed paragraph snippets. Burying the definition under preamble (“When dealing with website optimization, one important concept is…”) delays the answer past the length limit, so the extractor never reaches it. Splitting the definition across two paragraphs leaves the extractor pulling only the first, incomplete half. Wrapping the definition in qualifiers (“There is some debate about whether…”) makes the extracted text feel uncertain, which the algorithm tends not to favor. The single rule that covers most of this: the first sentence after the heading should be a complete, self-contained answer. Later sentences elaborate.
List snippets: ordered steps and unordered sets #
List snippets extract bulleted or numbered lists, typically showing four to eight items with a “more items” link expanding longer lists.
For ordered lists tied to process queries, a numbered or sequential list appears under a matching heading, each item starting with an action verb, each short enough to read at a glance (under about fifteen words). For the query “How to install WordPress on a server”, a winning list runs:
- Download the WordPress installation package from wordpress.org.
- Create a MySQL database for the installation.
- Upload the WordPress files to your server.
- Run the installation script by visiting your domain.
- Enter your database credentials and site information.
- Complete the setup with your administrator account.
Six steps, each a verb-led discrete action, in the order someone would perform them.
For unordered lists tied to enumeration queries, a bulleted list under a matching heading carries related but non-sequential items, each short enough to scan. For “Types of structured data for SEO”, a winning list names Article schema for blog and news content, Product schema for e-commerce listings, LocalBusiness schema for physical locations, Recipe schema for cooking instructions, FAQ schema for question-and-answer content, and Event schema for scheduled occurrences.
Markup matters here more than for any other format. Google extracts actual <ol> and <ul> elements with <li> items far more reliably than visual lists built from line breaks, manual numbering, or styled paragraphs. The semantic markup tells the extractor “this is a list,” which is what makes the items snippet candidates. The common failure is verbose items: each item running three or four sentences makes the list unsuitable, because the box truncates and the answer feels incomplete. Keep items to one or two lines and move any needed explanation into a longer expansion below.
Table snippets: structured comparison in rows and columns #
Table snippets extract HTML table content into the display. The format fits comparison queries (X vs Y), specification queries (dimensions, prices, features), and data queries (statistics, measurements).
The pattern: an HTML table under a matching heading, with clear column headers, consistent rows, and content that genuinely benefits from columns. For “JPG vs PNG vs WebP comparison”, a winning table gives four columns (Format, Best use, File size, Browser support) across three rows: JPG suits photographs, large at high quality, universal support; PNG suits graphics with transparency, larger than JPG, universal; WebP suits both photos and graphics, roughly a quarter to a third smaller than a comparable JPG, supported in all current browsers. Three rows, four columns, clean headers. Google shows the full table or a subset depending on space.
The markup requirement is strict: actual <table>, <thead>, <tbody>, <tr>, <th>, and <td> elements. Tables built from CSS grid or styled divs do not extract reliably because they are not semantically tables. The structure has to be what an HTML table is, not what a table looks like. Tables with merged cells, multi-level headers, or long prose in cells also fail; the extractor handles simple grids well and struggles with complex ones. A table with three header rows and rowspan or colspan attributes rarely appears as a snippet even when the data fits.
The upside is thin competition. Queries that benefit from comparison rarely have many sites offering a clean table answer, so the first site to publish a well-structured comparison table on a topic often captures the snippet because rivals are not using the format.
Question-style headings and the People Also Ask overlap #
Across all three formats, the heading directly above the answer carries weight. Google’s extractor uses heading-content proximity as a signal for which content answers which query. The pattern is an H2 or H3 immediately above snippet-eligible content that matches the query phrasing closely, often as a literal question: “What is a canonical tag?” above a paragraph, “How do I install WordPress?” above a numbered list, “What are the differences between JPG, PNG, and WebP?” above a comparison table.
The discipline is matching how users actually phrase queries. “What is a canonical tag” is a natural query; “Understanding Canonical Tag Implementation” is editorial framing that will not match the same search. Mirror the words users type, not the phrasing a writer might prefer. A page with question-style headings for several related sub-queries can appear as the snippet for one and populate People Also Ask boxes for others, since PAA expands related questions with short answers drawn from ranking pages on the same principle. That multiplies a single well-structured page’s presence on the results surface.
How to displace a snippet, and how to check whether you should #
Most snippets are already held by a specific page when a project begins. Displacing the holder takes more than matching format; it takes a better answer. The sequence: search the query directly to identify the current holder (its source URL sits below the snippet), visit that URL and study what makes the extraction work, then match the format exactly. A paragraph snippet is displaced by a better paragraph, a list by a better list. Then exceed on substance: a more complete or more current paragraph, a more comprehensive or better-ordered list, a table covering more comparison dimensions or presenting the data more clearly. Google has grown better at judging quality, so format matching alone is not enough when the existing answer is genuinely strong. Displacement works best where the incumbent has a visible weakness, an out-of-date answer or an incomplete list, that a new page clearly addresses.
Two constraints on timing. The new page must rank well enough to be considered, generally within the top five to ten organic positions, since snippet extraction pulls from pages already ranking near the top. And Google does not switch immediately; displacements tend to happen weeks after publication, not on the next crawl.
The most useful diagnostic is not the snippets a page has won but the ones it has not. For each query the page ranks for in positions two through ten, check whether a snippet appears; if one does and the page ranks nearby, it is competing and losing, and the gap is recoverable by examining what the holder provides that the page lacks. For each query where the page ranks first without holding the snippet, the algorithm evaluated the content and chose not to surface it, meaning the ranking is strong but the structure is not producing extraction. Adding question-style headings, converting prose to proper lists, or building comparison tables in those specific sections converts existing rankings into snippet positions.
Featured snippets in the AI Overview era #
Google’s AI Overviews, which appeared at scale starting in 2024, now sit above featured snippets on many queries. An AI Overview is a generated summary drawn from multiple sources and presented as a unified answer rather than a single-source extraction. The two surfaces relate in a few ways. Some queries that once showed a featured snippet now show an AI Overview instead, and the page that would have held the snippet sits below the Overview among the organic results. Some queries still show a featured snippet alone, typically those with a clear single-source answer, and Google will fall back to a featured snippet when it cannot generate an Overview for a query. Some queries show both.
The share has shifted noticeably. Through 2025, tracking studies recorded featured snippet visibility falling while AI Overview coverage rose across informational queries, so the surface featured snippet optimization targets is smaller than it was, and partly shared with the Overview surface.
The optimization principles overlap heavily, which is the practical reassurance. The patterns that win featured snippets (clear definitions, structured lists, comparison tables, question-style headings) also surface frequently in AI Overviews, because Google’s AI extractors lean on the same structural signals with similar preferences for clean formatting and direct answers. The difference sits at the margins: an AI Overview can synthesize across sources, so a page can contribute one part of a combined answer, whereas a featured snippet needs a single page to carry the complete answer. For a page building topical authority across many related queries, both surfaces reward comprehensive, well-structured content. The format-and-structure work still pays off, sometimes through the Overview surface rather than a traditional box.
Frequently asked questions #
Does my page need to rank on page one before it can win a featured snippet?
Effectively yes. Snippet extraction draws from pages already ranking near the top, generally the top five to ten organic positions. Optimizing content structure without the underlying ranking strength does not produce a snippet, because Google does not consider pages that are not already positioned to rank for the query.
How long is a paragraph snippet supposed to be?
Roughly 40 to 60 words. The extracted text should be a self-contained answer that reads correctly on its own, with the opening sentence carrying the complete answer and any following sentences adding context.
Is “position zero” still an accurate term?
Not really. Since the January 2020 deduplication, a featured snippet counts as the page’s position one rather than an extra slot above the organic results, and the same URL no longer appears both as the snippet and as a separate blue link on page one. The snippet consumes an organic position rather than adding one.
Why does my visually formatted list or table not get extracted?
Because the extractor reads HTML semantics, not appearance. Lists built from line breaks or styled paragraphs, and tables built from CSS grid or divs, are not recognized as lists or tables. Use actual <ol>, <ul>, and <li> elements for lists and real <table> markup for tables; style them with CSS afterward.
Are featured snippets still worth targeting now that AI Overviews exist?
Yes, with realistic expectations. Their share of the results page has declined, but they still appear, especially for definitions and lists, and Google falls back to them when it cannot generate an Overview. The structural work that earns snippets also feeds the AI Overview surface, so the effort is not wasted even on queries where the box itself no longer shows.
What snippet optimization rewards is not writing snippet bait. The answers that hold the position over time are the ones written to genuinely answer the question, in the format that fits the answer, on a page that ranks because the broader content is strong. What it punishes is assuming an existing snippet is secure. Every snippet the site holds is one a competitor is trying to displace, usually not by outranking the page overall but by outwriting the specific section the snippet extracts from. The maintenance discipline is recognizing when that section has aged and refreshing it first. The snippets a page should win but does not are the larger opportunity; the ones it used to win and lost are the warning.