SERP features

SERP features are the elements of a search results page that are not an ordinary title-and-snippet link: featured snippets, knowledge panels, image and video carousels, People Also Ask, local packs, review stars, sitelinks, and the generated answer blocks that now sit above all of them. They are assembled by the search engine from indexed content, which is the whole point — the publisher supplies the material and the engine decides the presentation.

Two families, controlled differently

Almost every practical question about SERP features resolves once you separate the ones with a documented eligibility path from the ones without.

Rich results have one. Google's search gallery documents 29 structured data types — Article, Breadcrumb, Carousel, Dataset, Event, Job posting, Local business, Product, Q&A, Recipe, Review snippet, Video, Fact check and Loyalty program among them — each with its own required properties. Google's framing is that it "uses structured data to understand the content on the page and show that content in a richer appearance in search results, which is called a rich result." Valid markup makes a page eligible. The same documentation warns that "the actual appearance in search results might be different", so eligibility is where publisher control ends. See schema markup.

Algorithmic features have no markup at all. A featured snippet is one of the "special boxes where the format of a regular search result is reversed, showing the descriptive snippet first", in Google's words, and there is no way to nominate a page: "Google systems determine whether a page would make a good featured snippet for a user's search request, and if so, elevates it." The same applies to People Also Ask, to most knowledge panel content, and to AI Overviews. The only documented lever is subtractive.

The subtractive controls

Because the algorithmic family cannot be requested, Google documents how to decline it. This is the part publishers most often get wrong, because the blunt instrument and the precise one look similar.

• `nosnippet` removes text snippets for the page entirely — including the ordinary result snippet. It is rarely what anyone wants.
• `data-nosnippet` marks specific sections of HTML as ineligible, leaving the rest of the page quotable. Use it for the paragraphs you do not want lifted.
• `max-snippet:[n]` caps snippet length. Google notes that featured snippets "will only appear if enough text can be shown to generate a useful featured snippet", so a low cap makes the featured form less likely while keeping a normal listing. Google does not publish the threshold — it varies with the snippet's information, the language and the platform — and warns that a low setting "doesn't guarantee that Google will stop showing featured snippets for your page"; only nosnippet is guaranteed.

There is no positive counterpart to any of these. Nothing in the documentation lets a publisher request a feature.

Why they matter when the reader is an agent

A SERP feature and an AI-assistant citation are the same transaction in different clothing: a passage of yours is extracted, shown in someone else's interface, and answers the question in place. The click may not happen. What survives is the claim and the attribution.

That makes the extraction unit the thing to optimise, and it is a passage rather than a page. A section that carries its own qualifier in the same sentence as its claim can be lifted whole and remain true; one whose caveat lives two paragraphs up arrives stripped of it. Direct-answer openings, one idea per paragraph, headings phrased as the question, and tables for anything comparative are the same craft that gets a passage into a featured snippet and into a generated answer — which is why the SERP-feature skill transferred cleanly to AI search rather than being replaced by it.

The cost side is real and should be stated plainly: features shift traffic toward zero-click outcomes. The right response depends on the query. For a question whose answer is the whole transaction, being the cited source is the win. For a question that should end in a signup, a trial or a purchase, an answer that satisfies in place is a loss, and the page should be written so the useful next step is only available on it.

Failure modes

• Marking up content that is not on the page. Google requires structured data to describe visible page content. Markup describing something a user cannot see is a policy violation, not a shortcut.
• Assuming valid markup means display. Validation proves the syntax; it proves nothing about appearance.
• Reaching for `nosnippet` to suppress a featured snippet. It removes the ordinary snippet too. max-snippet is the targeted control, at the cost of being the unguaranteed one.
• Optimising for a feature that is not shown for the query. Feature presence is query-dependent; a recipe carousel is not available to a B2B documentation page whatever the markup says.
• Writing pages instead of passages. The unit selected is a section, and a section that depends on the surrounding article does not survive extraction.
• Treating the type count as stable. 29 documented types is a snapshot from 3 September 2026. Types are added and retired, and the gallery is the source of truth, not a copy of it.

Frequently asked questions

Can markup guarantee a rich result?

No. It makes a page eligible; Google states the actual appearance in results might be different.

How do I get a featured snippet?

You cannot request one. Google's systems decide; the only documented control is opting out.

How do I opt out of just the featured snippet?

Set a short max-snippet rather than nosnippet, so too little text is available for the featured form. Google says this is not guaranteed; only nosnippet is.

How many rich result types exist?

29 documented in Google's search gallery as of 3 September 2026, each with its own property requirements.