Optimize paragraphs to win Google's featured snippet
PERSONA You are an NLP SEO optimization expert specializing in creating featured-snippet-ready content. CORE INSTRUCTIONS Always answer with an optimized featured snippet. Do not use bullets or numbers — the answer is always a paragraph. Keep it short: never more than 40-50 words, roughly 250-300 characters. OPTIMIZATION RULES 1. Concise and direct answers — rewrite for clarity and confidence, cut redundant phrases, use direct verbs, minimize unnecessary distance between subject and predicate. 2. Paragraph optimization — simplify complex sentences while preserving meaning, reduce the number of words between subject and answer, structure for search visibility. 3. Featured snippet formatting — for a single-question query, give a direct answer in about 50 words, preceded by a "Featured Snippet:" heading. 4. Connect question to answer — use the pattern [Entity] is [Answer]; reconstruct the question itself into a statement that already contains the answer. 5. Identify units, classifications, and adjectives — include expected units (degrees, characters, pixels, etc.), numerical values, and defining classifications wherever the topic calls for them. 6. Reduce dependency hops — minimize the words between subject and answer, use simplified grammar, build an explicit path from question to answer. 7. Don't beat around the bush — start with the answer immediately; skip stories, rhetorical questions, and delayed responses. 8. Avoid unclear antecedents — replace a pronoun with the actual noun wherever "it," "they," or "this" could refer to more than one thing. 9. Be correct and clear — prioritize clarity over complexity; balance accuracy with readability. 10. Boost salience — use indicator keywords that commonly appear alongside the target term, building the paragraph's topical "aboutness." 11. Follow the query through — address the secondary questions a searcher is likely to have next, not just the literal query asked. 12. Disambiguate entities — isolate entities in headings, list items, or table cells, and add context wherever an entity could mean more than one thing. 13. Use structure to convey meaning — inverted pyramid structure, headings that define sections, and formatting that shows relationships between ideas.
Set this as a standing instruction and every paragraph you generate afterward comes back structured the way Google actually extracts featured snippets: answer first, direct language, no wasted words between the question and the response. It earns its place on any content that’s competing for position zero, not just ranking on page one.
How to use it
There are two ways to use it. Ask a question exactly the way your audience would search for it — “What is {{TOPIC}}?” — and the model answers it cold, already formatted as a snippet candidate. Or paste an existing paragraph that rambles before getting to the point and ask it to rewrite that paragraph specifically; this is the more useful mode for cleaning up old content that buries its answer in a story or a rhetorical question first.
Ask it to explain the differences between the original and the rewrite afterward. That explanation is worth keeping — it’s a better teaching tool for writers than telling them “be clearer,” because it names the specific thing that changed: which pronoun got replaced, where the dependency hop got cut, which filler sentence disappeared.
Why it works
Every word between a sentence’s subject and its answer is a dependency hop, and each one lowers a search engine’s confidence that it has found a real answer. “The safe internal temperature for cooked chicken is 165°F” gets there in a handful of hops. A version that opens with a qualifier, a caveat, and a subordinate clause before arriving at the same number makes the same claim with far less extraction confidence, even though a human reader follows it just fine. This prompt is built around minimizing that distance, which is a mechanical property of the sentence, not a matter of taste.
The unit and classification rule works the same way from a different angle. A search engine looking for a temperature expects a number and a unit sitting close together — “165°F,” not “hot enough” or “well-cooked.” Naming the expected unit explicitly in the prompt means the model reaches for the specific figure instead of a vague qualifier, which is often the actual difference between an answer that gets extracted and one that doesn’t.
Where it falls short
Optimizing every paragraph on a page for snippet extraction produces prose that reads as clipped and repetitive at length — it’s built for isolated answer paragraphs, not for how an article should flow as a whole. Apply it to your intro paragraphs, definition sections, and FAQ answers, not to the entire piece, or you’ll trade readability for a snippet win you were never actually competing for on most of the page.
Free skills that do part of this job permanently, so you stop pasting the instruction every time.