Google’s system chooses dynamically between the declared meta description and content extracted directly from the page (and occasionally structured data), selecting whichever it judges best matches and describes the content for the specific query being searched. This is a query-dependent process, meaning the same page can display a different snippet depending on what was searched, since the passage of on-page text that best answers one query may not be the best match for another. One part of the question needs a direct correction first: Google sourcing snippets from third-party directory descriptions, the DMOZ-era practice, is not current. That mechanism was deprecated well over a decade ago and doesn’t factor into how Google generates snippets today. Any framing that treats directory descriptions as an active input is working from outdated information.
How the actual selection process works
Per Google’s Search Central documentation on creating good meta descriptions and on snippet generation, the snippet shown in search results is generated by Google’s systems specifically for the query-result pairing at hand, not fixed once per page. Google looks at the declared meta description as one candidate, and also considers extracting a passage directly from the page’s visible content as an alternative candidate, then displays whichever it judges provides a better, more relevant description of the page for that particular search. Structured data can also factor into what’s shown in some result types, though for standard organic snippet text the comparison is mainly between the declared description and on-page content.
Because this selection happens per query, a single page can show its declared meta description for one search term and a dynamically extracted passage for a different search term that surfaces the same page. This is a deliberate design, not an inconsistency: a static, single description written for one anticipated query can’t be equally well-matched to every query variant that might return that page in results, so Google’s system re-evaluates the best available snippet candidate for each individual search.
The practical implication is that the declared meta description isn’t a guaranteed display, it’s a candidate that competes against Google’s own extraction for each query. When a declared description closely matches, in natural language, what a searcher looking for that content would want to see, it’s more likely to be judged the best candidate. When it doesn’t (because it’s vague, generic, keyword-stuffed, or simply not as descriptive as a directly relevant passage elsewhere on the page), Google’s system will favor the extracted alternative instead.
Why the directory-description part of the question is outdated
In the earlier history of search, some engines did draw on third-party web directory descriptions (most notably from DMOZ, the Open Directory Project) as a fallback source for snippet text when a page lacked its own meta description or had one deemed unsuitable. That mechanism was discontinued industry-wide well over a decade ago as directories like DMOZ shut down and search engines moved toward generating snippets directly from the queried page’s own content and structured data instead. There is no current pathway by which Google pulls a snippet from an external directory listing. If a claim about “directory descriptions” is showing up as a candidate source in current SEO advice, that claim should be treated as obsolete rather than reflecting Google’s present-day mechanism.
What this means for writing descriptions
Since the real competition is between your declared description and Google’s own on-page extraction, the practical goal is writing a description that’s genuinely likely to win that comparison for the queries the page is meant to rank for. That means the description should read as an accurate, specific, natural-language summary of the page’s actual content, closely aligned with how a searcher would phrase what they’re looking for, rather than a generic placeholder or a restatement of the title. A description that’s vague enough to apply to many different pages gives Google’s system little reason to prefer it over pulling a more specific, directly relevant sentence from the body content itself. There’s no way to force Google to always display a given description, since the decision is remade per query, but writing descriptions that are specific, accurate, and well-matched to anticipated query language gives them the best realistic chance of being selected over an on-page extraction.
A worked example of the same page under two different queries
Take a page about home espresso machines that ranks for both “best espresso machine under $500” and “how to descale an espresso machine.” If the declared meta description focuses on machine recommendations and pricing, it’s a strong, specific candidate for the first query, closely matching what that searcher is looking for. For the second query, that same description doesn’t address descaling at all, so even though it’s a well-written, accurate description of the page’s primary focus, Google’s system may judge that a passage from the page’s descaling section (if the page happens to cover that topic further down) is a better match for that specific search, and display that extracted passage instead. The same page, same declared description, produces two different outcomes depending entirely on which query triggered the result. This is the mechanism working as intended, not a sign that the description was poorly written, a single fixed sentence simply can’t be the best-matching candidate for every distinct query a broad page might rank for.
Why this means one description can’t “win” every query a page ranks for
This has a direct practical implication worth stating plainly: if a page is genuinely meant to rank for several substantially different query intents, expecting one declared meta description to be selected for all of them is not a realistic goal, since the comparison is run independently each time and different queries will surface different best-matching passages from the page’s own content. The more actionable response, when a page covers multiple distinct sub-intents, isn’t to write an ever-longer description trying to cover every angle, that tends to produce a vaguer, less competitive description overall, but to make sure the on-page content itself is well-organized enough that whichever passage Google extracts for a less-central query is still accurate and well-written on its own, since that extracted text is effectively serving as the description for that query regardless of what the declared tag says.
A note on structured data and rich snippets
It’s worth distinguishing standard organic snippet text, the subject of this whole comparison, from rich result types that pull specific fields from structured data markup, like review star ratings, recipe cook times, or FAQ answers. Those elements can appear alongside or instead of a traditional text snippet when a page has eligible structured data and Google chooses to display that rich result type for a given query, but they operate under separate documented mechanisms and eligibility requirements specific to each structured data type, not the same declared-description-versus-extraction comparison described above. A page can have excellent meta description practices and still not show a rich result if it lacks the relevant structured data, and conversely a page with strong structured data can display rich elements while its plain-text snippet is still resolved through the same description-versus-extraction process covered here.
