<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u" -->

---
title: Prompt Tuning Firefox Shake to Summarize for Recipes
description: Mozilla engineers describe how they refined the prompt behind Firefox's Shake to Summarize feature to handle recipe webpages correctly. Early prompts caused...
canonical: https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Prompt Tuning Firefox Shake to Summarize for Recipes | daily.dev
og:description: Mozilla engineers describe how they refined the prompt behind Firefox's Shake to Summarize feature to handle recipe webpages correctly. Early prompts caused...
og:url: https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u
og:image: https://api.daily.dev/og/posts/4eOgiOW9u.png
og:image:alt: Prompt Tuning Firefox Shake to Summarize for Recipes
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompt Tuning Firefox Shake to Summarize for Recipes

**[Mozilla](https://daily.dev/sources/mozilla)** · 7 min read · 2 upvotes · 0 comments

## Summary

Mozilla engineers describe how they refined the prompt behind Firefox's Shake to Summarize feature to handle recipe webpages correctly. Early prompts caused the model to over-summarize recipes, dropping ingredients or steps. The fix was routing recipe pages to a dedicated, more explicit prompt (identified via structured data on the page) rather than overloading a single general-purpose summarization prompt, which more than doubled the rate of complete, accurate recipe summaries. The broader takeaway: reducing ambiguity and narrowing task scope for small models matters as much as model size.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://blog.mozilla.org/en/firefox/firefox-ai/prompt-tuning-firefox-shake-to-summarize-recipes>

## Questions this post answers

### Why did my LLM summarization prompt drop ingredients when summarizing recipe pages?

A single general-purpose summarization prompt tends to over-summarize recipes, omitting key ingredients or skipping the recipe entirely, because smaller models struggle to infer unstated format needs across varied content types. The fix is a dedicated recipe-specific prompt, explicitly listing servings, total time, ingredients, steps, and tips, routed to recipe pages using structured data metadata rather than an extra model inference pass.

_Developers debugging inconsistent LLM summarization can compare prompt-routing approaches like this one on daily.dev._

### How much did routing recipe pages to a dedicated prompt improve summary accuracy compared to a single general prompt?

Testing over a curated set of recipe sites showed the dedicated recipe prompt was more than twice as likely to produce a complete and accurate summary compared to the original single general-purpose prompt that handled all content types including recipes.

_Teams iterating on LLM prompt strategies can track real-world case studies like this on daily.dev._

## Similar posts on daily.dev

- [Firefox’s Shake to Summarize expands to Android and new languages on iOS](https://daily.dev/posts/firefox-s-shake-to-summarize-expands-to-android-and-new-languages-on-ios-91wsjxcyy) · Mozilla · 0 upvotes · 0 comments
- [Shake to Summarize recognized with special mention in TIME’s Best Inventions of 2025](https://daily.dev/posts/shake-to-summarize-recognized-with-special-mention-in-time-s-best-inventions-of-2025-wjo2zzkw2) · Mozilla · 0 upvotes · 0 comments
- [Under the hood: The AI powering Firefox’s Shake to Summarize](https://daily.dev/posts/under-the-hood-the-ai-powering-firefox-s-shake-to-summarize-du6trjo78) · Mozilla · 2 upvotes · 0 comments
- [The duality of language models in the browser](https://daily.dev/posts/the-duality-of-language-models-in-the-browser-ywrqdsevh) · Dave Rupert · 10 upvotes · 0 comments
- [Build new features using built-in AI in Chrome](https://daily.dev/posts/build-new-features-using-built-in-ai-in-chrome-gmmbpgcby) · Chrome Developers · 0 upvotes · 0 comments

---

Tags: [#llm](https://daily.dev/tags/llm), [#prompt-engineering](https://daily.dev/tags/prompt-engineering), [#firefox](https://daily.dev/tags/firefox), [#local-ai](https://daily.dev/tags/local-ai), [#mozilla](https://daily.dev/tags/mozilla)

[View this post on daily.dev](https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Prompt Tuning Firefox Shake to Summarize for Recipes","url":"https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u"},"datePublished":"2026-10-01T16:00:09.366Z","dateModified":"2026-10-01T16:00:28.321Z","description":"Mozilla engineers describe how they refined the prompt behind Firefox's Shake to Summarize feature to handle recipe webpages correctly. Early prompts caused...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/280119f2877eb9e89c4f3b95c8e1b0a2?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/280119f2877eb9e89c4f3b95c8e1b0a2?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Mozilla","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Mozilla","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/0bb80a1805194372922ffd83bbead649","url":"https://daily.dev/sources/mozilla"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,prompt-engineering,firefox,local-ai,mozilla","timeRequired":"PT7M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Mozilla","item":"https://daily.dev/sources/mozilla"},{"@type":"ListItem","position":3,"name":"Prompt Tuning Firefox Shake to Summarize for Recipes"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/prompt-tuning-firefox-shake-to-summarize-for-recipes-4eogiow9u#faq","mainEntity":[{"@type":"Question","name":"Why did my LLM summarization prompt drop ingredients when summarizing recipe pages?","acceptedAnswer":{"@type":"Answer","text":"A single general-purpose summarization prompt tends to over-summarize recipes, omitting key ingredients or skipping the recipe entirely, because smaller models struggle to infer unstated format needs across varied content types. The fix is a dedicated recipe-specific prompt, explicitly listing servings, total time, ingredients, steps, and tips, routed to recipe pages using structured data metadata rather than an extra model inference pass. Developers debugging inconsistent LLM summarization can compare prompt-routing approaches like this one on daily.dev."}},{"@type":"Question","name":"How much did routing recipe pages to a dedicated prompt improve summary accuracy compared to a single general prompt?","acceptedAnswer":{"@type":"Answer","text":"Testing over a curated set of recipe sites showed the dedicated recipe prompt was more than twice as likely to produce a complete and accurate summary compared to the original single general-purpose prompt that handled all content types including recipes. Teams iterating on LLM prompt strategies can track real-world case studies like this on daily.dev."}}]}
```

