You wrote a carefully crafted meta description for a high-traffic landing page. Google ignored it. Instead, it pulled a sentence fragment from the third paragraph of your body copy, one that contained the query terms but lacked the call to action you designed. This is not a bug. Google’s snippet selection system evaluates the declared meta description as one candidate among several, including on-page text passages, DMOZ/directory descriptions (historically), and structured data content. The system selects whichever candidate best matches the specific query the searcher typed. Your meta description competes for display the same way your page competes for ranking: by query-level relevance matching, not by declaration alone.
The Snippet Candidate Pool Google Evaluates Per Query
Google’s snippet generation system draws from multiple text sources on and around the page to build a candidate pool for each search result display. According to Google’s Search Central documentation, updated in January 2024, the content of the page itself is the primary source for snippet generation. The meta description element is one input, but it is not the default or guaranteed selection.
The candidate pool includes the following sources, roughly ordered by evaluation priority. On-page body text constitutes the largest candidate set. Google identifies text passages within the body content that contain the query terms or semantically related phrases, then evaluates each passage for completeness, readability, and contextual relevance. A single page may generate dozens of candidate passages, each corresponding to a different section of the content.
The declared meta description enters the pool as a single candidate. Google’s documentation states that the system will “sometimes use the meta description tag from a page to generate a snippet in search results, if they think it gives users a more accurate description than would be possible purely from the on-page content.” The phrasing is critical: the meta description is used when it outperforms on-page passages, not as a default that on-page passages must beat.
Heading text (H1, H2, H3 elements) provides additional candidates, particularly when headings contain query-relevant terms and the surrounding body text is less directly responsive to the query. Google may combine a heading with adjacent body text to construct a snippet that spans multiple page elements.
Structured data content contributes to the candidate pool indirectly. FAQ schema, product descriptions in schema markup, and review content marked up with structured data can surface as snippet components or replace the traditional snippet format entirely with a rich result.
Historical directory descriptions from DMOZ (the Open Directory Project) were once a significant source. Google discontinued using DMOZ descriptions in 2017, but sites that relied on directory descriptions as their de facto meta descriptions may still see legacy effects if they never created proper meta description tags.
The system evaluates all candidates simultaneously for each query. There is no sequential fallback chain where Google checks the meta description first and only looks elsewhere if it fails. All sources compete on equal footing, with the winner determined by query-specific relevance scoring.
Query-Level Matching Logic for Snippet Candidate Selection
The snippet selection is not a per-page decision. It is a per-query decision. The same page displays different snippets for different search queries because the matching algorithm evaluates each candidate passage against the specific query terms and intent.
The matching logic prioritizes several factors. Query-term presence and bolding is the most visible signal. Google bolds query terms within the displayed snippet, and passages that contain exact or close matches to the query terms score higher in the candidate evaluation. A meta description that contains the primary keyword but not the long-tail modifiers used in a specific search will lose to a body text passage that contains both.
Semantic relevance extends beyond exact term matching. Google’s NLP systems evaluate whether the candidate passage addresses the topic and intent behind the query, not just whether it contains the same words. A passage that answers the implied question behind a query can outperform a passage with higher term overlap but lower semantic alignment.
Passage completeness affects selection. Google prefers snippet candidates that form complete, readable statements rather than fragments. Body text that begins mid-sentence or ends abruptly scores lower than passages that form coherent standalone descriptions. This is one area where well-written meta descriptions have a structural advantage: they are designed as complete, self-contained descriptions, while body text passages are extracted from longer contexts and may not read well in isolation.
Snippet length targeting influences which candidates qualify. While there is no fixed character limit for snippets, Google targets a display length that fits the SERP format, typically 150-160 characters on desktop. Candidates that naturally fall within this range or can be cleanly truncated to fit receive preference over candidates that require heavy editing to display.
The practical implication is that a meta description optimized for the primary target keyword may win snippet selection for that keyword but lose to body text passages for dozens of long-tail variations. This is expected behavior, not a failure of the meta description.
Meta Description Override Rates by Page Type and Context
Multiple large-scale studies have quantified the rate at which Google overrides declared meta descriptions. The data consistently places the override rate between 60-70% across all queries and page types, though the rate varies significantly by context.
Portent’s research found that Google rewrites meta descriptions 71% of the time on mobile and 68% on desktop for first-page results. Ahrefs’ study found that Google rewrote 63% of meta descriptions overall, with 61.46% of descriptions that were too long being rewritten and 63.69% of descriptions within acceptable length ranges also being rewritten. This latter finding is particularly important: even well-formatted, appropriately-lengthed meta descriptions get overridden nearly two-thirds of the time.
The override rate varies by page type and content structure. Product pages with detailed specifications see high override rates because Google frequently pulls specific product attributes (price, dimensions, availability) from body content to match transactional queries, rather than displaying the generic marketing-oriented meta description. Blog posts and articles with strong heading structures see moderate override rates because their body content provides many well-structured candidate passages. Landing pages with minimal body text see lower override rates because fewer body-text candidates compete with the meta description.
Search volume correlates inversely with override rates. Portent’s data suggests that pages ranking for high-volume keywords are less likely to have their meta descriptions overridden. The likely explanation is that SEO practitioners invest more effort in writing query-aligned meta descriptions for high-volume targets, producing descriptions that better match the queries they target.
An additional factor: 25% of top-ranking pages lack a meta description entirely, according to Ahrefs’ data. For these pages, Google must generate the snippet exclusively from on-page content, which it does successfully. The system is designed to function without meta descriptions as a primary input.
How Structured Data and Rich Snippets Interact With Meta Description Display
When a page contains structured data markup that qualifies for a rich result, the traditional snippet format may be partially or fully replaced by structured data content. This interaction creates a separate channel through which the meta description can be displaced.
FAQ schema is the most direct replacement mechanism. When FAQ structured data is present and Google displays FAQ rich results, the snippet area shows question-and-answer pairs from the schema markup rather than the meta description or body text. The meta description is not displayed at all in these cases, regardless of its quality or relevance.
Product schema with price, availability, and review data produces rich snippets that include structured information alongside or in place of the text snippet. The meta description may still appear in a reduced form, but the visual prominence of the structured data elements (star ratings, price ranges, stock status) reduces the meta description’s influence on click behavior.
Review schema and how-to schema similarly modify the snippet display format. When these structured data types trigger rich results, the snippet composition changes from a pure text display to a hybrid of structured data elements and text. The meta description competes not only against body text passages but also against these formatted data displays.
The strategic implication is that pages with structured data markup need to account for two separate snippet strategies: the text snippet (controlled by meta description and body content quality) and the rich result display (controlled by structured data completeness and accuracy). Investing heavily in meta description optimization for pages that consistently display rich results produces diminishing returns because the meta description occupies a smaller portion of the search result visual space.
Google’s January 2024 documentation update explicitly clarified that structured data and meta descriptions are “not the primary source” of search snippets, placing on-page body content as the primary source. This positions meta descriptions as a secondary influence that can guide snippet selection but cannot override the system’s preference for query-matched body content.
Practical Implications for Meta Description Writing at Scale
Understanding snippet selection as a competitive, query-level process changes the approach to meta description writing. The goal is not to write one perfect description per page. The goal is to write a description that wins selection for the page’s highest-value queries while ensuring the body content produces acceptable snippets for long-tail queries where the meta description will be overridden.
For high-traffic pages, the meta description should target the primary keyword cluster that drives the most impressions. Include the exact primary keyword phrase within the first 100 characters to maximize the probability of winning snippet selection for that query. Avoid stuffing multiple keyword variations into the meta description, as this dilutes the query-specific relevance signal and produces text that reads as keyword-targeted rather than user-informative.
For pages that serve multiple distinct query intents, accept that the meta description will be displayed for some queries and overridden for others. Structure the body content so that the passages Google extracts for long-tail queries are informative and complete. This means writing clear, self-contained sentences near the top of each content section, rather than burying key information in complex paragraph structures that produce fragmented snippets when extracted.
SearchPilot’s controlled split test provides a cautionary finding. When they used the data-nosnippet attribute to force Google to use the declared meta description instead of body text passages, the result was a statistically significant negative impact on organic traffic. Google’s algorithmic snippet selection outperformed the manually written descriptions. This does not mean meta descriptions are worthless. It means that the system’s ability to select query-specific passages from body content often produces better matches than a single static description can achieve.
For large-scale sites, programmatic meta description generation is both acceptable and recommended by Google. The same structured data attributes used for differentiated title tags can populate meta descriptions. A product page meta description template pulling brand, key feature, and price creates a more query-relevant snippet candidate than a generic category-level description. For parallel insights into how Google’s title selection system operates on similar principles, see Google’s Title Rewriting Algorithm Triggers. For strategies to improve the display rate of declared meta descriptions, see Google’s Title Rewriting Algorithm Triggers.
Does Google evaluate snippet candidates differently for mobile versus desktop search results?
Google displays different snippet lengths on mobile (approximately 120 characters) versus desktop (approximately 150-160 characters). This length difference affects which candidate passages qualify for display on each device type. Portent’s research found that Google rewrites meta descriptions 71% of the time on mobile compared to 68% on desktop, suggesting the narrower mobile display window increases the probability that body text passages outperform the declared description for mobile queries.
Can structured data markup completely replace the meta description as a snippet source?
Structured data does not replace the meta description but can displace it from the visible search result. When FAQ schema, product schema, or review schema triggers a rich result, the snippet area shows structured data content instead of the text snippet. The meta description is not displayed in these cases regardless of its quality. Pages consistently triggering rich results receive diminishing returns from meta description optimization because the description occupies a smaller portion of the search result visual space.
Does the data-nosnippet attribute give full control over which text Google uses as the snippet?
The data-nosnippet attribute prevents Google from using marked content as a snippet source, but it does not guarantee Google will use the meta description instead. SearchPilot’s controlled split test found that restricting Google’s snippet options with data-nosnippet produced a statistically significant negative impact on organic traffic. Google’s algorithmic snippet selection often outperforms manually constrained options, so this attribute should be tested on a small page segment before broad deployment.
Sources
- How to Write Meta Descriptions – Google Search Central
- How Often Does Google Rewrite Meta Descriptions? (New Data Study) – Ahrefs
- How Often Google Ignores Our Meta Descriptions – Portent
- Google Rewrites Meta Descriptions Over 70% of the Time – Search Engine Journal
- SEO Split Test: Forcing Google to Respect Meta Descriptions – SearchPilot
- Google Clarifies How Algorithm Chooses Search Snippets – Search Engine Journal
